2016-02-10 00:12:00 +01:00
|
|
|
// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
|
2017-07-16 01:03:42 +02:00
|
|
|
// This source code is licensed under both the GPLv2 (found in the
|
|
|
|
// COPYING file in the root directory) and Apache 2.0 License
|
|
|
|
// (found in the LICENSE.Apache file in the root directory).
|
2013-10-16 23:59:46 +02:00
|
|
|
//
|
2011-03-18 23:37:00 +01:00
|
|
|
// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
|
|
|
|
// Use of this source code is governed by a BSD-style license that can be
|
|
|
|
// found in the LICENSE file. See the AUTHORS file for names of contributors.
|
|
|
|
//
|
|
|
|
// A portable implementation of crc32c, optimized to handle
|
|
|
|
// four bytes at a time.
|
|
|
|
#include "util/crc32c.h"
|
2012-12-19 01:52:02 +01:00
|
|
|
#include <stdint.h>
|
cross-platform compatibility improvements
Summary:
We've had a couple CockroachDB users fail to build RocksDB on exotic platforms, so I figured I'd try my hand at solving these issues upstream. The problems stem from a) `USE_SSE=1` being too aggressive about turning on SSE4.2, even on toolchains that don't support SSE4.2 and b) RocksDB attempting to detect support for thread-local storage based on OS, even though it can vary by compiler on the same OS.
See the individual commit messages for details. Regarding SSE support, this PR should change virtually nothing for non-CMake based builds. `make`, `PORTABLE=1 make`, `USE_SSE=1 make`, and `PORTABLE=1 USE_SSE=1 make` function exactly as before, except that SSE support will be automatically disabled when a simple SSE4.2-using test program fails to compile, as it does on OpenBSD. (OpenBSD's ports GCC supports SSE4.2, but its binutils do not, so `__SSE_4_2__` is defined but an SSE4.2-using program will fail to assemble.) A warning is emitted in this case. The CMake build is modified to support the same set of options, except that `USE_SSE` is spelled `FORCE_SSE42` because `USE_SSE` is rather useless now that we can automatically detect SSE support, and I figure changing options in the CMake build is less disruptive than changing the non-CMake build.
I've tested these changes on all the platforms I can get my hands on (macOS, Windows MSVC, Windows MinGW, and OpenBSD) and it all works splendidly. Let me know if there's anything you object to—I obviously don't mean to break any of your build pipelines in the process of fixing ours downstream.
Closes https://github.com/facebook/rocksdb/pull/2199
Differential Revision: D5054042
Pulled By: yiwu-arbug
fbshipit-source-id: 938e1fc665c049c02ae15698e1409155b8e72171
2017-05-15 23:42:32 +02:00
|
|
|
#ifdef HAVE_SSE42
|
2012-05-04 03:10:17 +02:00
|
|
|
#include <nmmintrin.h>
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
#include <wmmintrin.h>
|
2012-05-04 03:10:17 +02:00
|
|
|
#endif
|
2020-04-20 22:21:34 +02:00
|
|
|
#include "port/lang.h"
|
2011-03-18 23:37:00 +01:00
|
|
|
#include "util/coding.h"
|
|
|
|
|
2019-04-30 19:56:06 +02:00
|
|
|
#include "util/crc32c_arm64.h"
|
|
|
|
|
2017-08-31 23:02:06 +02:00
|
|
|
#ifdef __powerpc64__
|
|
|
|
#include "util/crc32c_ppc.h"
|
|
|
|
#include "util/crc32c_ppc_constants.h"
|
|
|
|
|
|
|
|
#if __linux__
|
2020-03-04 03:06:56 +01:00
|
|
|
#ifdef ROCKSDB_AUXV_GETAUXVAL_PRESENT
|
2017-08-31 23:02:06 +02:00
|
|
|
#include <sys/auxv.h>
|
2020-03-04 03:06:56 +01:00
|
|
|
#endif
|
2017-08-31 23:02:06 +02:00
|
|
|
|
|
|
|
#ifndef PPC_FEATURE2_VEC_CRYPTO
|
|
|
|
#define PPC_FEATURE2_VEC_CRYPTO 0x02000000
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifndef AT_HWCAP2
|
|
|
|
#define AT_HWCAP2 26
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#endif /* __linux__ */
|
|
|
|
|
|
|
|
#endif
|
|
|
|
|
2020-09-22 19:39:54 +02:00
|
|
|
#if defined(__linux__) && defined(HAVE_ARM64_CRC)
|
|
|
|
bool pmull_runtime_flag = false;
|
|
|
|
#endif
|
|
|
|
|
2020-02-20 21:07:53 +01:00
|
|
|
namespace ROCKSDB_NAMESPACE {
|
2011-03-18 23:37:00 +01:00
|
|
|
namespace crc32c {
|
|
|
|
|
2018-02-09 23:01:10 +01:00
|
|
|
#if defined(HAVE_POWER8) && defined(HAS_ALTIVEC)
|
2017-08-31 23:02:06 +02:00
|
|
|
#ifdef __powerpc64__
|
|
|
|
static int arch_ppc_crc32 = 0;
|
2018-02-09 23:01:10 +01:00
|
|
|
#endif /* __powerpc64__ */
|
2017-08-31 23:02:06 +02:00
|
|
|
#endif
|
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
static const uint32_t table0_[256] = {
|
|
|
|
0x00000000, 0xf26b8303, 0xe13b70f7, 0x1350f3f4,
|
|
|
|
0xc79a971f, 0x35f1141c, 0x26a1e7e8, 0xd4ca64eb,
|
|
|
|
0x8ad958cf, 0x78b2dbcc, 0x6be22838, 0x9989ab3b,
|
|
|
|
0x4d43cfd0, 0xbf284cd3, 0xac78bf27, 0x5e133c24,
|
|
|
|
0x105ec76f, 0xe235446c, 0xf165b798, 0x030e349b,
|
|
|
|
0xd7c45070, 0x25afd373, 0x36ff2087, 0xc494a384,
|
|
|
|
0x9a879fa0, 0x68ec1ca3, 0x7bbcef57, 0x89d76c54,
|
|
|
|
0x5d1d08bf, 0xaf768bbc, 0xbc267848, 0x4e4dfb4b,
|
|
|
|
0x20bd8ede, 0xd2d60ddd, 0xc186fe29, 0x33ed7d2a,
|
|
|
|
0xe72719c1, 0x154c9ac2, 0x061c6936, 0xf477ea35,
|
|
|
|
0xaa64d611, 0x580f5512, 0x4b5fa6e6, 0xb93425e5,
|
|
|
|
0x6dfe410e, 0x9f95c20d, 0x8cc531f9, 0x7eaeb2fa,
|
|
|
|
0x30e349b1, 0xc288cab2, 0xd1d83946, 0x23b3ba45,
|
|
|
|
0xf779deae, 0x05125dad, 0x1642ae59, 0xe4292d5a,
|
|
|
|
0xba3a117e, 0x4851927d, 0x5b016189, 0xa96ae28a,
|
|
|
|
0x7da08661, 0x8fcb0562, 0x9c9bf696, 0x6ef07595,
|
|
|
|
0x417b1dbc, 0xb3109ebf, 0xa0406d4b, 0x522bee48,
|
|
|
|
0x86e18aa3, 0x748a09a0, 0x67dafa54, 0x95b17957,
|
|
|
|
0xcba24573, 0x39c9c670, 0x2a993584, 0xd8f2b687,
|
|
|
|
0x0c38d26c, 0xfe53516f, 0xed03a29b, 0x1f682198,
|
|
|
|
0x5125dad3, 0xa34e59d0, 0xb01eaa24, 0x42752927,
|
|
|
|
0x96bf4dcc, 0x64d4cecf, 0x77843d3b, 0x85efbe38,
|
|
|
|
0xdbfc821c, 0x2997011f, 0x3ac7f2eb, 0xc8ac71e8,
|
|
|
|
0x1c661503, 0xee0d9600, 0xfd5d65f4, 0x0f36e6f7,
|
|
|
|
0x61c69362, 0x93ad1061, 0x80fde395, 0x72966096,
|
|
|
|
0xa65c047d, 0x5437877e, 0x4767748a, 0xb50cf789,
|
|
|
|
0xeb1fcbad, 0x197448ae, 0x0a24bb5a, 0xf84f3859,
|
|
|
|
0x2c855cb2, 0xdeeedfb1, 0xcdbe2c45, 0x3fd5af46,
|
|
|
|
0x7198540d, 0x83f3d70e, 0x90a324fa, 0x62c8a7f9,
|
|
|
|
0xb602c312, 0x44694011, 0x5739b3e5, 0xa55230e6,
|
|
|
|
0xfb410cc2, 0x092a8fc1, 0x1a7a7c35, 0xe811ff36,
|
|
|
|
0x3cdb9bdd, 0xceb018de, 0xdde0eb2a, 0x2f8b6829,
|
|
|
|
0x82f63b78, 0x709db87b, 0x63cd4b8f, 0x91a6c88c,
|
|
|
|
0x456cac67, 0xb7072f64, 0xa457dc90, 0x563c5f93,
|
|
|
|
0x082f63b7, 0xfa44e0b4, 0xe9141340, 0x1b7f9043,
|
|
|
|
0xcfb5f4a8, 0x3dde77ab, 0x2e8e845f, 0xdce5075c,
|
|
|
|
0x92a8fc17, 0x60c37f14, 0x73938ce0, 0x81f80fe3,
|
|
|
|
0x55326b08, 0xa759e80b, 0xb4091bff, 0x466298fc,
|
|
|
|
0x1871a4d8, 0xea1a27db, 0xf94ad42f, 0x0b21572c,
|
|
|
|
0xdfeb33c7, 0x2d80b0c4, 0x3ed04330, 0xccbbc033,
|
|
|
|
0xa24bb5a6, 0x502036a5, 0x4370c551, 0xb11b4652,
|
|
|
|
0x65d122b9, 0x97baa1ba, 0x84ea524e, 0x7681d14d,
|
|
|
|
0x2892ed69, 0xdaf96e6a, 0xc9a99d9e, 0x3bc21e9d,
|
|
|
|
0xef087a76, 0x1d63f975, 0x0e330a81, 0xfc588982,
|
|
|
|
0xb21572c9, 0x407ef1ca, 0x532e023e, 0xa145813d,
|
|
|
|
0x758fe5d6, 0x87e466d5, 0x94b49521, 0x66df1622,
|
|
|
|
0x38cc2a06, 0xcaa7a905, 0xd9f75af1, 0x2b9cd9f2,
|
|
|
|
0xff56bd19, 0x0d3d3e1a, 0x1e6dcdee, 0xec064eed,
|
|
|
|
0xc38d26c4, 0x31e6a5c7, 0x22b65633, 0xd0ddd530,
|
|
|
|
0x0417b1db, 0xf67c32d8, 0xe52cc12c, 0x1747422f,
|
|
|
|
0x49547e0b, 0xbb3ffd08, 0xa86f0efc, 0x5a048dff,
|
|
|
|
0x8ecee914, 0x7ca56a17, 0x6ff599e3, 0x9d9e1ae0,
|
|
|
|
0xd3d3e1ab, 0x21b862a8, 0x32e8915c, 0xc083125f,
|
|
|
|
0x144976b4, 0xe622f5b7, 0xf5720643, 0x07198540,
|
|
|
|
0x590ab964, 0xab613a67, 0xb831c993, 0x4a5a4a90,
|
|
|
|
0x9e902e7b, 0x6cfbad78, 0x7fab5e8c, 0x8dc0dd8f,
|
|
|
|
0xe330a81a, 0x115b2b19, 0x020bd8ed, 0xf0605bee,
|
|
|
|
0x24aa3f05, 0xd6c1bc06, 0xc5914ff2, 0x37faccf1,
|
|
|
|
0x69e9f0d5, 0x9b8273d6, 0x88d28022, 0x7ab90321,
|
|
|
|
0xae7367ca, 0x5c18e4c9, 0x4f48173d, 0xbd23943e,
|
|
|
|
0xf36e6f75, 0x0105ec76, 0x12551f82, 0xe03e9c81,
|
|
|
|
0x34f4f86a, 0xc69f7b69, 0xd5cf889d, 0x27a40b9e,
|
|
|
|
0x79b737ba, 0x8bdcb4b9, 0x988c474d, 0x6ae7c44e,
|
|
|
|
0xbe2da0a5, 0x4c4623a6, 0x5f16d052, 0xad7d5351
|
|
|
|
};
|
|
|
|
static const uint32_t table1_[256] = {
|
|
|
|
0x00000000, 0x13a29877, 0x274530ee, 0x34e7a899,
|
|
|
|
0x4e8a61dc, 0x5d28f9ab, 0x69cf5132, 0x7a6dc945,
|
|
|
|
0x9d14c3b8, 0x8eb65bcf, 0xba51f356, 0xa9f36b21,
|
|
|
|
0xd39ea264, 0xc03c3a13, 0xf4db928a, 0xe7790afd,
|
|
|
|
0x3fc5f181, 0x2c6769f6, 0x1880c16f, 0x0b225918,
|
|
|
|
0x714f905d, 0x62ed082a, 0x560aa0b3, 0x45a838c4,
|
|
|
|
0xa2d13239, 0xb173aa4e, 0x859402d7, 0x96369aa0,
|
|
|
|
0xec5b53e5, 0xfff9cb92, 0xcb1e630b, 0xd8bcfb7c,
|
|
|
|
0x7f8be302, 0x6c297b75, 0x58ced3ec, 0x4b6c4b9b,
|
|
|
|
0x310182de, 0x22a31aa9, 0x1644b230, 0x05e62a47,
|
|
|
|
0xe29f20ba, 0xf13db8cd, 0xc5da1054, 0xd6788823,
|
|
|
|
0xac154166, 0xbfb7d911, 0x8b507188, 0x98f2e9ff,
|
|
|
|
0x404e1283, 0x53ec8af4, 0x670b226d, 0x74a9ba1a,
|
|
|
|
0x0ec4735f, 0x1d66eb28, 0x298143b1, 0x3a23dbc6,
|
|
|
|
0xdd5ad13b, 0xcef8494c, 0xfa1fe1d5, 0xe9bd79a2,
|
|
|
|
0x93d0b0e7, 0x80722890, 0xb4958009, 0xa737187e,
|
|
|
|
0xff17c604, 0xecb55e73, 0xd852f6ea, 0xcbf06e9d,
|
|
|
|
0xb19da7d8, 0xa23f3faf, 0x96d89736, 0x857a0f41,
|
|
|
|
0x620305bc, 0x71a19dcb, 0x45463552, 0x56e4ad25,
|
|
|
|
0x2c896460, 0x3f2bfc17, 0x0bcc548e, 0x186eccf9,
|
|
|
|
0xc0d23785, 0xd370aff2, 0xe797076b, 0xf4359f1c,
|
|
|
|
0x8e585659, 0x9dface2e, 0xa91d66b7, 0xbabffec0,
|
|
|
|
0x5dc6f43d, 0x4e646c4a, 0x7a83c4d3, 0x69215ca4,
|
|
|
|
0x134c95e1, 0x00ee0d96, 0x3409a50f, 0x27ab3d78,
|
|
|
|
0x809c2506, 0x933ebd71, 0xa7d915e8, 0xb47b8d9f,
|
|
|
|
0xce1644da, 0xddb4dcad, 0xe9537434, 0xfaf1ec43,
|
|
|
|
0x1d88e6be, 0x0e2a7ec9, 0x3acdd650, 0x296f4e27,
|
|
|
|
0x53028762, 0x40a01f15, 0x7447b78c, 0x67e52ffb,
|
|
|
|
0xbf59d487, 0xacfb4cf0, 0x981ce469, 0x8bbe7c1e,
|
|
|
|
0xf1d3b55b, 0xe2712d2c, 0xd69685b5, 0xc5341dc2,
|
|
|
|
0x224d173f, 0x31ef8f48, 0x050827d1, 0x16aabfa6,
|
|
|
|
0x6cc776e3, 0x7f65ee94, 0x4b82460d, 0x5820de7a,
|
|
|
|
0xfbc3faf9, 0xe861628e, 0xdc86ca17, 0xcf245260,
|
|
|
|
0xb5499b25, 0xa6eb0352, 0x920cabcb, 0x81ae33bc,
|
|
|
|
0x66d73941, 0x7575a136, 0x419209af, 0x523091d8,
|
|
|
|
0x285d589d, 0x3bffc0ea, 0x0f186873, 0x1cbaf004,
|
|
|
|
0xc4060b78, 0xd7a4930f, 0xe3433b96, 0xf0e1a3e1,
|
|
|
|
0x8a8c6aa4, 0x992ef2d3, 0xadc95a4a, 0xbe6bc23d,
|
|
|
|
0x5912c8c0, 0x4ab050b7, 0x7e57f82e, 0x6df56059,
|
|
|
|
0x1798a91c, 0x043a316b, 0x30dd99f2, 0x237f0185,
|
|
|
|
0x844819fb, 0x97ea818c, 0xa30d2915, 0xb0afb162,
|
|
|
|
0xcac27827, 0xd960e050, 0xed8748c9, 0xfe25d0be,
|
|
|
|
0x195cda43, 0x0afe4234, 0x3e19eaad, 0x2dbb72da,
|
|
|
|
0x57d6bb9f, 0x447423e8, 0x70938b71, 0x63311306,
|
|
|
|
0xbb8de87a, 0xa82f700d, 0x9cc8d894, 0x8f6a40e3,
|
|
|
|
0xf50789a6, 0xe6a511d1, 0xd242b948, 0xc1e0213f,
|
|
|
|
0x26992bc2, 0x353bb3b5, 0x01dc1b2c, 0x127e835b,
|
|
|
|
0x68134a1e, 0x7bb1d269, 0x4f567af0, 0x5cf4e287,
|
|
|
|
0x04d43cfd, 0x1776a48a, 0x23910c13, 0x30339464,
|
|
|
|
0x4a5e5d21, 0x59fcc556, 0x6d1b6dcf, 0x7eb9f5b8,
|
|
|
|
0x99c0ff45, 0x8a626732, 0xbe85cfab, 0xad2757dc,
|
|
|
|
0xd74a9e99, 0xc4e806ee, 0xf00fae77, 0xe3ad3600,
|
|
|
|
0x3b11cd7c, 0x28b3550b, 0x1c54fd92, 0x0ff665e5,
|
|
|
|
0x759baca0, 0x663934d7, 0x52de9c4e, 0x417c0439,
|
|
|
|
0xa6050ec4, 0xb5a796b3, 0x81403e2a, 0x92e2a65d,
|
|
|
|
0xe88f6f18, 0xfb2df76f, 0xcfca5ff6, 0xdc68c781,
|
|
|
|
0x7b5fdfff, 0x68fd4788, 0x5c1aef11, 0x4fb87766,
|
|
|
|
0x35d5be23, 0x26772654, 0x12908ecd, 0x013216ba,
|
|
|
|
0xe64b1c47, 0xf5e98430, 0xc10e2ca9, 0xd2acb4de,
|
|
|
|
0xa8c17d9b, 0xbb63e5ec, 0x8f844d75, 0x9c26d502,
|
|
|
|
0x449a2e7e, 0x5738b609, 0x63df1e90, 0x707d86e7,
|
|
|
|
0x0a104fa2, 0x19b2d7d5, 0x2d557f4c, 0x3ef7e73b,
|
|
|
|
0xd98eedc6, 0xca2c75b1, 0xfecbdd28, 0xed69455f,
|
|
|
|
0x97048c1a, 0x84a6146d, 0xb041bcf4, 0xa3e32483
|
|
|
|
};
|
|
|
|
static const uint32_t table2_[256] = {
|
|
|
|
0x00000000, 0xa541927e, 0x4f6f520d, 0xea2ec073,
|
|
|
|
0x9edea41a, 0x3b9f3664, 0xd1b1f617, 0x74f06469,
|
|
|
|
0x38513ec5, 0x9d10acbb, 0x773e6cc8, 0xd27ffeb6,
|
|
|
|
0xa68f9adf, 0x03ce08a1, 0xe9e0c8d2, 0x4ca15aac,
|
|
|
|
0x70a27d8a, 0xd5e3eff4, 0x3fcd2f87, 0x9a8cbdf9,
|
|
|
|
0xee7cd990, 0x4b3d4bee, 0xa1138b9d, 0x045219e3,
|
|
|
|
0x48f3434f, 0xedb2d131, 0x079c1142, 0xa2dd833c,
|
|
|
|
0xd62de755, 0x736c752b, 0x9942b558, 0x3c032726,
|
|
|
|
0xe144fb14, 0x4405696a, 0xae2ba919, 0x0b6a3b67,
|
|
|
|
0x7f9a5f0e, 0xdadbcd70, 0x30f50d03, 0x95b49f7d,
|
|
|
|
0xd915c5d1, 0x7c5457af, 0x967a97dc, 0x333b05a2,
|
|
|
|
0x47cb61cb, 0xe28af3b5, 0x08a433c6, 0xade5a1b8,
|
|
|
|
0x91e6869e, 0x34a714e0, 0xde89d493, 0x7bc846ed,
|
|
|
|
0x0f382284, 0xaa79b0fa, 0x40577089, 0xe516e2f7,
|
|
|
|
0xa9b7b85b, 0x0cf62a25, 0xe6d8ea56, 0x43997828,
|
|
|
|
0x37691c41, 0x92288e3f, 0x78064e4c, 0xdd47dc32,
|
|
|
|
0xc76580d9, 0x622412a7, 0x880ad2d4, 0x2d4b40aa,
|
|
|
|
0x59bb24c3, 0xfcfab6bd, 0x16d476ce, 0xb395e4b0,
|
|
|
|
0xff34be1c, 0x5a752c62, 0xb05bec11, 0x151a7e6f,
|
|
|
|
0x61ea1a06, 0xc4ab8878, 0x2e85480b, 0x8bc4da75,
|
|
|
|
0xb7c7fd53, 0x12866f2d, 0xf8a8af5e, 0x5de93d20,
|
|
|
|
0x29195949, 0x8c58cb37, 0x66760b44, 0xc337993a,
|
|
|
|
0x8f96c396, 0x2ad751e8, 0xc0f9919b, 0x65b803e5,
|
|
|
|
0x1148678c, 0xb409f5f2, 0x5e273581, 0xfb66a7ff,
|
|
|
|
0x26217bcd, 0x8360e9b3, 0x694e29c0, 0xcc0fbbbe,
|
|
|
|
0xb8ffdfd7, 0x1dbe4da9, 0xf7908dda, 0x52d11fa4,
|
|
|
|
0x1e704508, 0xbb31d776, 0x511f1705, 0xf45e857b,
|
|
|
|
0x80aee112, 0x25ef736c, 0xcfc1b31f, 0x6a802161,
|
|
|
|
0x56830647, 0xf3c29439, 0x19ec544a, 0xbcadc634,
|
|
|
|
0xc85da25d, 0x6d1c3023, 0x8732f050, 0x2273622e,
|
|
|
|
0x6ed23882, 0xcb93aafc, 0x21bd6a8f, 0x84fcf8f1,
|
|
|
|
0xf00c9c98, 0x554d0ee6, 0xbf63ce95, 0x1a225ceb,
|
|
|
|
0x8b277743, 0x2e66e53d, 0xc448254e, 0x6109b730,
|
|
|
|
0x15f9d359, 0xb0b84127, 0x5a968154, 0xffd7132a,
|
|
|
|
0xb3764986, 0x1637dbf8, 0xfc191b8b, 0x595889f5,
|
|
|
|
0x2da8ed9c, 0x88e97fe2, 0x62c7bf91, 0xc7862def,
|
|
|
|
0xfb850ac9, 0x5ec498b7, 0xb4ea58c4, 0x11abcaba,
|
|
|
|
0x655baed3, 0xc01a3cad, 0x2a34fcde, 0x8f756ea0,
|
|
|
|
0xc3d4340c, 0x6695a672, 0x8cbb6601, 0x29faf47f,
|
|
|
|
0x5d0a9016, 0xf84b0268, 0x1265c21b, 0xb7245065,
|
|
|
|
0x6a638c57, 0xcf221e29, 0x250cde5a, 0x804d4c24,
|
|
|
|
0xf4bd284d, 0x51fcba33, 0xbbd27a40, 0x1e93e83e,
|
|
|
|
0x5232b292, 0xf77320ec, 0x1d5de09f, 0xb81c72e1,
|
|
|
|
0xccec1688, 0x69ad84f6, 0x83834485, 0x26c2d6fb,
|
|
|
|
0x1ac1f1dd, 0xbf8063a3, 0x55aea3d0, 0xf0ef31ae,
|
|
|
|
0x841f55c7, 0x215ec7b9, 0xcb7007ca, 0x6e3195b4,
|
|
|
|
0x2290cf18, 0x87d15d66, 0x6dff9d15, 0xc8be0f6b,
|
|
|
|
0xbc4e6b02, 0x190ff97c, 0xf321390f, 0x5660ab71,
|
|
|
|
0x4c42f79a, 0xe90365e4, 0x032da597, 0xa66c37e9,
|
|
|
|
0xd29c5380, 0x77ddc1fe, 0x9df3018d, 0x38b293f3,
|
|
|
|
0x7413c95f, 0xd1525b21, 0x3b7c9b52, 0x9e3d092c,
|
|
|
|
0xeacd6d45, 0x4f8cff3b, 0xa5a23f48, 0x00e3ad36,
|
|
|
|
0x3ce08a10, 0x99a1186e, 0x738fd81d, 0xd6ce4a63,
|
|
|
|
0xa23e2e0a, 0x077fbc74, 0xed517c07, 0x4810ee79,
|
|
|
|
0x04b1b4d5, 0xa1f026ab, 0x4bdee6d8, 0xee9f74a6,
|
|
|
|
0x9a6f10cf, 0x3f2e82b1, 0xd50042c2, 0x7041d0bc,
|
|
|
|
0xad060c8e, 0x08479ef0, 0xe2695e83, 0x4728ccfd,
|
|
|
|
0x33d8a894, 0x96993aea, 0x7cb7fa99, 0xd9f668e7,
|
|
|
|
0x9557324b, 0x3016a035, 0xda386046, 0x7f79f238,
|
|
|
|
0x0b899651, 0xaec8042f, 0x44e6c45c, 0xe1a75622,
|
|
|
|
0xdda47104, 0x78e5e37a, 0x92cb2309, 0x378ab177,
|
|
|
|
0x437ad51e, 0xe63b4760, 0x0c158713, 0xa954156d,
|
|
|
|
0xe5f54fc1, 0x40b4ddbf, 0xaa9a1dcc, 0x0fdb8fb2,
|
|
|
|
0x7b2bebdb, 0xde6a79a5, 0x3444b9d6, 0x91052ba8
|
|
|
|
};
|
|
|
|
static const uint32_t table3_[256] = {
|
|
|
|
0x00000000, 0xdd45aab8, 0xbf672381, 0x62228939,
|
|
|
|
0x7b2231f3, 0xa6679b4b, 0xc4451272, 0x1900b8ca,
|
|
|
|
0xf64463e6, 0x2b01c95e, 0x49234067, 0x9466eadf,
|
|
|
|
0x8d665215, 0x5023f8ad, 0x32017194, 0xef44db2c,
|
|
|
|
0xe964b13d, 0x34211b85, 0x560392bc, 0x8b463804,
|
|
|
|
0x924680ce, 0x4f032a76, 0x2d21a34f, 0xf06409f7,
|
|
|
|
0x1f20d2db, 0xc2657863, 0xa047f15a, 0x7d025be2,
|
|
|
|
0x6402e328, 0xb9474990, 0xdb65c0a9, 0x06206a11,
|
|
|
|
0xd725148b, 0x0a60be33, 0x6842370a, 0xb5079db2,
|
|
|
|
0xac072578, 0x71428fc0, 0x136006f9, 0xce25ac41,
|
|
|
|
0x2161776d, 0xfc24ddd5, 0x9e0654ec, 0x4343fe54,
|
|
|
|
0x5a43469e, 0x8706ec26, 0xe524651f, 0x3861cfa7,
|
|
|
|
0x3e41a5b6, 0xe3040f0e, 0x81268637, 0x5c632c8f,
|
|
|
|
0x45639445, 0x98263efd, 0xfa04b7c4, 0x27411d7c,
|
|
|
|
0xc805c650, 0x15406ce8, 0x7762e5d1, 0xaa274f69,
|
|
|
|
0xb327f7a3, 0x6e625d1b, 0x0c40d422, 0xd1057e9a,
|
|
|
|
0xaba65fe7, 0x76e3f55f, 0x14c17c66, 0xc984d6de,
|
|
|
|
0xd0846e14, 0x0dc1c4ac, 0x6fe34d95, 0xb2a6e72d,
|
|
|
|
0x5de23c01, 0x80a796b9, 0xe2851f80, 0x3fc0b538,
|
|
|
|
0x26c00df2, 0xfb85a74a, 0x99a72e73, 0x44e284cb,
|
|
|
|
0x42c2eeda, 0x9f874462, 0xfda5cd5b, 0x20e067e3,
|
|
|
|
0x39e0df29, 0xe4a57591, 0x8687fca8, 0x5bc25610,
|
|
|
|
0xb4868d3c, 0x69c32784, 0x0be1aebd, 0xd6a40405,
|
|
|
|
0xcfa4bccf, 0x12e11677, 0x70c39f4e, 0xad8635f6,
|
|
|
|
0x7c834b6c, 0xa1c6e1d4, 0xc3e468ed, 0x1ea1c255,
|
|
|
|
0x07a17a9f, 0xdae4d027, 0xb8c6591e, 0x6583f3a6,
|
|
|
|
0x8ac7288a, 0x57828232, 0x35a00b0b, 0xe8e5a1b3,
|
|
|
|
0xf1e51979, 0x2ca0b3c1, 0x4e823af8, 0x93c79040,
|
|
|
|
0x95e7fa51, 0x48a250e9, 0x2a80d9d0, 0xf7c57368,
|
|
|
|
0xeec5cba2, 0x3380611a, 0x51a2e823, 0x8ce7429b,
|
|
|
|
0x63a399b7, 0xbee6330f, 0xdcc4ba36, 0x0181108e,
|
|
|
|
0x1881a844, 0xc5c402fc, 0xa7e68bc5, 0x7aa3217d,
|
|
|
|
0x52a0c93f, 0x8fe56387, 0xedc7eabe, 0x30824006,
|
|
|
|
0x2982f8cc, 0xf4c75274, 0x96e5db4d, 0x4ba071f5,
|
|
|
|
0xa4e4aad9, 0x79a10061, 0x1b838958, 0xc6c623e0,
|
|
|
|
0xdfc69b2a, 0x02833192, 0x60a1b8ab, 0xbde41213,
|
|
|
|
0xbbc47802, 0x6681d2ba, 0x04a35b83, 0xd9e6f13b,
|
|
|
|
0xc0e649f1, 0x1da3e349, 0x7f816a70, 0xa2c4c0c8,
|
|
|
|
0x4d801be4, 0x90c5b15c, 0xf2e73865, 0x2fa292dd,
|
|
|
|
0x36a22a17, 0xebe780af, 0x89c50996, 0x5480a32e,
|
|
|
|
0x8585ddb4, 0x58c0770c, 0x3ae2fe35, 0xe7a7548d,
|
|
|
|
0xfea7ec47, 0x23e246ff, 0x41c0cfc6, 0x9c85657e,
|
|
|
|
0x73c1be52, 0xae8414ea, 0xcca69dd3, 0x11e3376b,
|
|
|
|
0x08e38fa1, 0xd5a62519, 0xb784ac20, 0x6ac10698,
|
|
|
|
0x6ce16c89, 0xb1a4c631, 0xd3864f08, 0x0ec3e5b0,
|
|
|
|
0x17c35d7a, 0xca86f7c2, 0xa8a47efb, 0x75e1d443,
|
|
|
|
0x9aa50f6f, 0x47e0a5d7, 0x25c22cee, 0xf8878656,
|
|
|
|
0xe1873e9c, 0x3cc29424, 0x5ee01d1d, 0x83a5b7a5,
|
|
|
|
0xf90696d8, 0x24433c60, 0x4661b559, 0x9b241fe1,
|
|
|
|
0x8224a72b, 0x5f610d93, 0x3d4384aa, 0xe0062e12,
|
|
|
|
0x0f42f53e, 0xd2075f86, 0xb025d6bf, 0x6d607c07,
|
|
|
|
0x7460c4cd, 0xa9256e75, 0xcb07e74c, 0x16424df4,
|
|
|
|
0x106227e5, 0xcd278d5d, 0xaf050464, 0x7240aedc,
|
|
|
|
0x6b401616, 0xb605bcae, 0xd4273597, 0x09629f2f,
|
|
|
|
0xe6264403, 0x3b63eebb, 0x59416782, 0x8404cd3a,
|
|
|
|
0x9d0475f0, 0x4041df48, 0x22635671, 0xff26fcc9,
|
|
|
|
0x2e238253, 0xf36628eb, 0x9144a1d2, 0x4c010b6a,
|
|
|
|
0x5501b3a0, 0x88441918, 0xea669021, 0x37233a99,
|
|
|
|
0xd867e1b5, 0x05224b0d, 0x6700c234, 0xba45688c,
|
|
|
|
0xa345d046, 0x7e007afe, 0x1c22f3c7, 0xc167597f,
|
|
|
|
0xc747336e, 0x1a0299d6, 0x782010ef, 0xa565ba57,
|
|
|
|
0xbc65029d, 0x6120a825, 0x0302211c, 0xde478ba4,
|
|
|
|
0x31035088, 0xec46fa30, 0x8e647309, 0x5321d9b1,
|
|
|
|
0x4a21617b, 0x9764cbc3, 0xf54642fa, 0x2803e842
|
|
|
|
};
|
|
|
|
|
|
|
|
// Used to fetch a naturally-aligned 32-bit word in little endian byte-order
|
|
|
|
static inline uint32_t LE_LOAD32(const uint8_t *p) {
|
|
|
|
return DecodeFixed32(reinterpret_cast<const char*>(p));
|
|
|
|
}
|
|
|
|
|
cross-platform compatibility improvements
Summary:
We've had a couple CockroachDB users fail to build RocksDB on exotic platforms, so I figured I'd try my hand at solving these issues upstream. The problems stem from a) `USE_SSE=1` being too aggressive about turning on SSE4.2, even on toolchains that don't support SSE4.2 and b) RocksDB attempting to detect support for thread-local storage based on OS, even though it can vary by compiler on the same OS.
See the individual commit messages for details. Regarding SSE support, this PR should change virtually nothing for non-CMake based builds. `make`, `PORTABLE=1 make`, `USE_SSE=1 make`, and `PORTABLE=1 USE_SSE=1 make` function exactly as before, except that SSE support will be automatically disabled when a simple SSE4.2-using test program fails to compile, as it does on OpenBSD. (OpenBSD's ports GCC supports SSE4.2, but its binutils do not, so `__SSE_4_2__` is defined but an SSE4.2-using program will fail to assemble.) A warning is emitted in this case. The CMake build is modified to support the same set of options, except that `USE_SSE` is spelled `FORCE_SSE42` because `USE_SSE` is rather useless now that we can automatically detect SSE support, and I figure changing options in the CMake build is less disruptive than changing the non-CMake build.
I've tested these changes on all the platforms I can get my hands on (macOS, Windows MSVC, Windows MinGW, and OpenBSD) and it all works splendidly. Let me know if there's anything you object to—I obviously don't mean to break any of your build pipelines in the process of fixing ours downstream.
Closes https://github.com/facebook/rocksdb/pull/2199
Differential Revision: D5054042
Pulled By: yiwu-arbug
fbshipit-source-id: 938e1fc665c049c02ae15698e1409155b8e72171
2017-05-15 23:42:32 +02:00
|
|
|
#if defined(HAVE_SSE42) && (defined(__LP64__) || defined(_WIN64))
|
2012-05-04 03:10:17 +02:00
|
|
|
static inline uint64_t LE_LOAD64(const uint8_t *p) {
|
|
|
|
return DecodeFixed64(reinterpret_cast<const char*>(p));
|
|
|
|
}
|
2014-03-14 18:22:37 +01:00
|
|
|
#endif
|
2012-05-04 03:10:17 +02:00
|
|
|
|
2012-12-19 01:52:02 +01:00
|
|
|
static inline void Slow_CRC32(uint64_t* l, uint8_t const **p) {
|
2014-11-11 22:47:22 +01:00
|
|
|
uint32_t c = static_cast<uint32_t>(*l ^ LE_LOAD32(*p));
|
2012-12-19 01:52:02 +01:00
|
|
|
*p += 4;
|
|
|
|
*l = table3_[c & 0xff] ^
|
|
|
|
table2_[(c >> 8) & 0xff] ^
|
|
|
|
table1_[(c >> 16) & 0xff] ^
|
|
|
|
table0_[c >> 24];
|
|
|
|
// DO it twice.
|
2014-11-11 22:47:22 +01:00
|
|
|
c = static_cast<uint32_t>(*l ^ LE_LOAD32(*p));
|
2012-12-19 01:52:02 +01:00
|
|
|
*p += 4;
|
|
|
|
*l = table3_[c & 0xff] ^
|
|
|
|
table2_[(c >> 8) & 0xff] ^
|
|
|
|
table1_[(c >> 16) & 0xff] ^
|
|
|
|
table0_[c >> 24];
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void Fast_CRC32(uint64_t* l, uint8_t const **p) {
|
cross-platform compatibility improvements
Summary:
We've had a couple CockroachDB users fail to build RocksDB on exotic platforms, so I figured I'd try my hand at solving these issues upstream. The problems stem from a) `USE_SSE=1` being too aggressive about turning on SSE4.2, even on toolchains that don't support SSE4.2 and b) RocksDB attempting to detect support for thread-local storage based on OS, even though it can vary by compiler on the same OS.
See the individual commit messages for details. Regarding SSE support, this PR should change virtually nothing for non-CMake based builds. `make`, `PORTABLE=1 make`, `USE_SSE=1 make`, and `PORTABLE=1 USE_SSE=1 make` function exactly as before, except that SSE support will be automatically disabled when a simple SSE4.2-using test program fails to compile, as it does on OpenBSD. (OpenBSD's ports GCC supports SSE4.2, but its binutils do not, so `__SSE_4_2__` is defined but an SSE4.2-using program will fail to assemble.) A warning is emitted in this case. The CMake build is modified to support the same set of options, except that `USE_SSE` is spelled `FORCE_SSE42` because `USE_SSE` is rather useless now that we can automatically detect SSE support, and I figure changing options in the CMake build is less disruptive than changing the non-CMake build.
I've tested these changes on all the platforms I can get my hands on (macOS, Windows MSVC, Windows MinGW, and OpenBSD) and it all works splendidly. Let me know if there's anything you object to—I obviously don't mean to break any of your build pipelines in the process of fixing ours downstream.
Closes https://github.com/facebook/rocksdb/pull/2199
Differential Revision: D5054042
Pulled By: yiwu-arbug
fbshipit-source-id: 938e1fc665c049c02ae15698e1409155b8e72171
2017-05-15 23:42:32 +02:00
|
|
|
#ifndef HAVE_SSE42
|
|
|
|
Slow_CRC32(l, p);
|
|
|
|
#elif defined(__LP64__) || defined(_WIN64)
|
2012-12-19 01:52:02 +01:00
|
|
|
*l = _mm_crc32_u64(*l, LE_LOAD64(*p));
|
|
|
|
*p += 8;
|
2015-12-11 02:26:11 +01:00
|
|
|
#else
|
2016-01-07 19:22:09 +01:00
|
|
|
*l = _mm_crc32_u32(static_cast<unsigned int>(*l), LE_LOAD32(*p));
|
2015-12-11 02:26:11 +01:00
|
|
|
*p += 4;
|
2016-01-07 19:22:09 +01:00
|
|
|
*l = _mm_crc32_u32(static_cast<unsigned int>(*l), LE_LOAD32(*p));
|
2015-12-11 02:26:11 +01:00
|
|
|
*p += 4;
|
|
|
|
#endif
|
2014-04-04 23:03:19 +02:00
|
|
|
}
|
2012-12-19 01:52:02 +01:00
|
|
|
|
2014-04-03 00:15:57 +02:00
|
|
|
template<void (*CRC32)(uint64_t*, uint8_t const**)>
|
|
|
|
uint32_t ExtendImpl(uint32_t crc, const char* buf, size_t size) {
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
const uint8_t *p = reinterpret_cast<const uint8_t *>(buf);
|
|
|
|
const uint8_t *e = p + size;
|
2012-05-04 02:26:14 +02:00
|
|
|
uint64_t l = crc ^ 0xffffffffu;
|
|
|
|
|
|
|
|
// Align n to (1 << m) byte boundary
|
2012-11-29 01:42:36 +01:00
|
|
|
#define ALIGN(n, m) ((n + ((1 << m) - 1)) & ~((1 << m) - 1))
|
2011-03-18 23:37:00 +01:00
|
|
|
|
|
|
|
#define STEP1 do { \
|
|
|
|
int c = (l & 0xff) ^ *p++; \
|
|
|
|
l = table0_[c] ^ (l >> 8); \
|
|
|
|
} while (0)
|
2012-05-04 03:10:17 +02:00
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
|
2012-05-04 02:26:14 +02:00
|
|
|
// Point x at first 16-byte aligned byte in string. This might be
|
2011-03-18 23:37:00 +01:00
|
|
|
// just past the end of the string.
|
|
|
|
const uintptr_t pval = reinterpret_cast<uintptr_t>(p);
|
2012-05-04 02:26:14 +02:00
|
|
|
const uint8_t* x = reinterpret_cast<const uint8_t*>(ALIGN(pval, 4));
|
2011-03-18 23:37:00 +01:00
|
|
|
if (x <= e) {
|
2012-05-04 02:26:14 +02:00
|
|
|
// Process bytes until finished or p is 16-byte aligned
|
2011-03-18 23:37:00 +01:00
|
|
|
while (p != x) {
|
|
|
|
STEP1;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
// Process bytes 16 at a time
|
|
|
|
while ((e-p) >= 16) {
|
2012-12-19 01:52:02 +01:00
|
|
|
CRC32(&l, &p);
|
|
|
|
CRC32(&l, &p);
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
2012-05-04 02:26:14 +02:00
|
|
|
// Process bytes 8 at a time
|
|
|
|
while ((e-p) >= 8) {
|
2012-12-19 01:52:02 +01:00
|
|
|
CRC32(&l, &p);
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
// Process the last few bytes
|
|
|
|
while (p != e) {
|
|
|
|
STEP1;
|
|
|
|
}
|
|
|
|
#undef STEP1
|
2012-05-04 02:26:14 +02:00
|
|
|
#undef ALIGN
|
2014-11-11 22:47:22 +01:00
|
|
|
return static_cast<uint32_t>(l ^ 0xffffffffu);
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
2019-07-15 20:16:55 +02:00
|
|
|
// Detect if ARM64 CRC or not.
|
|
|
|
#ifndef HAVE_ARM64_CRC
|
2014-04-04 23:03:19 +02:00
|
|
|
// Detect if SS42 or not.
|
2017-08-31 23:02:06 +02:00
|
|
|
#ifndef HAVE_POWER8
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
|
2014-04-04 23:03:19 +02:00
|
|
|
static bool isSSE42() {
|
2017-06-29 06:37:55 +02:00
|
|
|
#ifndef HAVE_SSE42
|
|
|
|
return false;
|
|
|
|
#elif defined(__GNUC__) && defined(__x86_64__) && !defined(IOS_CROSS_COMPILE)
|
2014-04-04 23:03:19 +02:00
|
|
|
uint32_t c_;
|
specify SSE42 'target' attribute for Fast_CRC32()
Summary:
if we enable SSE42 globally when compiling the tree for preparing a
portable binary, which could be running on CPU w/o SSE42 instructions
even the GCC on the building host is able to emit SSE42 code, this leads
to illegal instruction errors on machines not supporting SSE42. to solve
this problem, crc32 detects the supported instruction at runtime, and
selects the supported CRC32 implementation according to the result of
`cpuid`. but intrinics like "_mm_crc32_u64()" will not be available
unless the "target" machine is appropriately specified in the command
line, like "-msse42", or using the "target" attribute.
we could pass "-msse42" only when compiling crc32c.cc, and allow the
compiler to generate the SSE42 instructions, but we are still at the
risk of executing illegal instructions on machines does not support
SSE42 if the compiler emits code that is not guarded by our runtime
detection. and we need to do the change in both Makefile and CMakefile.
or, we can use GCC's "target" attribute to enable the machine specific
instructions on certain function. in this way, we have finer grained
control of the used "target". and no need to change the makefiles. so
we don't need to duplicate the changes on both makefile and cmake as
the previous approach.
this problem surfaces when preparing a package for GNU/Linux distribution,
and we only applies to optimization for SSE42, so using a feature
only available on GCC/Clang is not that formidable.
Closes https://github.com/facebook/rocksdb/pull/2807
Differential Revision: D5786084
Pulled By: siying
fbshipit-source-id: bca5c0f877b8d6fb55f58f8f122254a26422843d
2017-09-07 21:32:50 +02:00
|
|
|
__asm__("cpuid" : "=c"(c_) : "a"(1) : "ebx", "edx");
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
return c_ & (1U << 20); // copied from CpuId.h in Folly. Test SSE42
|
2017-03-30 02:34:36 +02:00
|
|
|
#elif defined(_WIN64)
|
|
|
|
int info[4];
|
|
|
|
__cpuidex(info, 0x00000001, 0);
|
|
|
|
return (info[2] & ((int)1 << 20)) != 0;
|
2014-04-04 23:03:19 +02:00
|
|
|
#else
|
|
|
|
return false;
|
|
|
|
#endif
|
|
|
|
}
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
|
|
|
|
static bool isPCLMULQDQ() {
|
|
|
|
#ifndef HAVE_SSE42
|
|
|
|
// in build_detect_platform we set this macro when both SSE42 and PCLMULQDQ are
|
|
|
|
// supported by compiler
|
|
|
|
return false;
|
|
|
|
#elif defined(__GNUC__) && defined(__x86_64__) && !defined(IOS_CROSS_COMPILE)
|
|
|
|
uint32_t c_;
|
|
|
|
__asm__("cpuid" : "=c"(c_) : "a"(1) : "ebx", "edx");
|
|
|
|
return c_ & (1U << 1); // PCLMULQDQ is in bit 1 (not bit 0)
|
|
|
|
#elif defined(_WIN64)
|
|
|
|
int info[4];
|
|
|
|
__cpuidex(info, 0x00000001, 0);
|
|
|
|
return (info[2] & ((int)1 << 1)) != 0;
|
|
|
|
#else
|
|
|
|
return false;
|
2017-08-31 23:02:06 +02:00
|
|
|
#endif
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
#endif // HAVE_POWER8
|
2019-07-15 20:16:55 +02:00
|
|
|
#endif // HAVE_ARM64_CRC
|
2014-04-04 23:03:19 +02:00
|
|
|
|
2014-04-03 00:15:57 +02:00
|
|
|
typedef uint32_t (*Function)(uint32_t, const char*, size_t);
|
|
|
|
|
2017-08-31 23:02:06 +02:00
|
|
|
#if defined(HAVE_POWER8) && defined(HAS_ALTIVEC)
|
|
|
|
uint32_t ExtendPPCImpl(uint32_t crc, const char *buf, size_t size) {
|
|
|
|
return crc32c_ppc(crc, (const unsigned char *)buf, size);
|
|
|
|
}
|
|
|
|
|
|
|
|
#if __linux__
|
|
|
|
static int arch_ppc_probe(void) {
|
|
|
|
arch_ppc_crc32 = 0;
|
|
|
|
|
2020-03-04 03:06:56 +01:00
|
|
|
#if defined(__powerpc64__) && defined(ROCKSDB_AUXV_GETAUXVAL_PRESENT)
|
2017-08-31 23:02:06 +02:00
|
|
|
if (getauxval(AT_HWCAP2) & PPC_FEATURE2_VEC_CRYPTO) arch_ppc_crc32 = 1;
|
|
|
|
#endif /* __powerpc64__ */
|
|
|
|
|
|
|
|
return arch_ppc_crc32;
|
|
|
|
}
|
|
|
|
#endif // __linux__
|
|
|
|
|
|
|
|
static bool isAltiVec() {
|
|
|
|
if (arch_ppc_probe()) {
|
|
|
|
return true;
|
|
|
|
} else {
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2019-04-30 19:56:06 +02:00
|
|
|
#if defined(__linux__) && defined(HAVE_ARM64_CRC)
|
|
|
|
uint32_t ExtendARMImpl(uint32_t crc, const char *buf, size_t size) {
|
|
|
|
return crc32c_arm64(crc, (const unsigned char *)buf, size);
|
|
|
|
}
|
|
|
|
#endif
|
2014-04-03 00:15:57 +02:00
|
|
|
|
2017-08-31 23:02:06 +02:00
|
|
|
std::string IsFastCrc32Supported() {
|
|
|
|
bool has_fast_crc = false;
|
|
|
|
std::string fast_zero_msg;
|
|
|
|
std::string arch;
|
|
|
|
#ifdef HAVE_POWER8
|
|
|
|
#ifdef HAS_ALTIVEC
|
|
|
|
if (arch_ppc_probe()) {
|
|
|
|
has_fast_crc = true;
|
|
|
|
arch = "PPC";
|
|
|
|
}
|
|
|
|
#else
|
|
|
|
has_fast_crc = false;
|
|
|
|
arch = "PPC";
|
|
|
|
#endif
|
2019-04-30 19:56:06 +02:00
|
|
|
#elif defined(__linux__) && defined(HAVE_ARM64_CRC)
|
|
|
|
if (crc32c_runtime_check()) {
|
|
|
|
has_fast_crc = true;
|
|
|
|
arch = "Arm64";
|
2020-09-22 19:39:54 +02:00
|
|
|
pmull_runtime_flag = crc32c_pmull_runtime_check();
|
2019-04-30 19:56:06 +02:00
|
|
|
} else {
|
|
|
|
has_fast_crc = false;
|
|
|
|
arch = "Arm64";
|
|
|
|
}
|
2017-08-31 23:02:06 +02:00
|
|
|
#else
|
|
|
|
has_fast_crc = isSSE42();
|
|
|
|
arch = "x86";
|
|
|
|
#endif
|
2017-09-08 19:57:12 +02:00
|
|
|
if (has_fast_crc) {
|
2017-08-31 23:02:06 +02:00
|
|
|
fast_zero_msg.append("Supported on " + arch);
|
|
|
|
}
|
2017-11-03 06:12:17 +01:00
|
|
|
else {
|
|
|
|
fast_zero_msg.append("Not supported on " + arch);
|
|
|
|
}
|
2017-09-08 19:57:12 +02:00
|
|
|
return fast_zero_msg;
|
2015-07-11 02:35:27 +02:00
|
|
|
}
|
|
|
|
|
2014-04-03 00:15:57 +02:00
|
|
|
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
/*
|
|
|
|
* Copyright 2016 Ferry Toth, Exalon Delft BV, The Netherlands
|
|
|
|
* This software is provided 'as-is', without any express or implied
|
|
|
|
* warranty. In no event will the author be held liable for any damages
|
|
|
|
* arising from the use of this software.
|
|
|
|
* Permission is granted to anyone to use this software for any purpose,
|
|
|
|
* including commercial applications, and to alter it and redistribute it
|
|
|
|
* freely, subject to the following restrictions:
|
|
|
|
* 1. The origin of this software must not be misrepresented; you must not
|
|
|
|
* claim that you wrote the original software. If you use this software
|
|
|
|
* in a product, an acknowledgment in the product documentation would be
|
|
|
|
* appreciated but is not required.
|
|
|
|
* 2. Altered source versions must be plainly marked as such, and must not be
|
|
|
|
* misrepresented as being the original software.
|
|
|
|
* 3. This notice may not be removed or altered from any source distribution.
|
|
|
|
* Ferry Toth
|
|
|
|
* ftoth@exalondelft.nl
|
|
|
|
*
|
|
|
|
* https://github.com/htot/crc32c
|
|
|
|
*
|
|
|
|
* Modified by Facebook
|
|
|
|
*
|
|
|
|
* Original intel whitepaper:
|
|
|
|
* "Fast CRC Computation for iSCSI Polynomial Using CRC32 Instruction"
|
|
|
|
* https://www.intel.com/content/dam/www/public/us/en/documents/white-papers/crc-iscsi-polynomial-crc32-instruction-paper.pdf
|
|
|
|
*
|
|
|
|
* This version is from the folly library, created by Dave Watson <davejwatson@fb.com>
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
#if defined HAVE_SSE42 && defined HAVE_PCLMUL
|
|
|
|
|
|
|
|
#define CRCtriplet(crc, buf, offset) \
|
|
|
|
crc##0 = _mm_crc32_u64(crc##0, *(buf##0 + offset)); \
|
|
|
|
crc##1 = _mm_crc32_u64(crc##1, *(buf##1 + offset)); \
|
|
|
|
crc##2 = _mm_crc32_u64(crc##2, *(buf##2 + offset));
|
|
|
|
|
|
|
|
#define CRCduplet(crc, buf, offset) \
|
|
|
|
crc##0 = _mm_crc32_u64(crc##0, *(buf##0 + offset)); \
|
|
|
|
crc##1 = _mm_crc32_u64(crc##1, *(buf##1 + offset));
|
|
|
|
|
|
|
|
#define CRCsinglet(crc, buf, offset) \
|
|
|
|
crc = _mm_crc32_u64(crc, *(uint64_t*)(buf + offset));
|
|
|
|
|
|
|
|
|
|
|
|
// Numbers taken directly from intel whitepaper.
|
|
|
|
// clang-format off
|
|
|
|
const uint64_t clmul_constants[] = {
|
|
|
|
0x14cd00bd6, 0x105ec76f0, 0x0ba4fc28e, 0x14cd00bd6,
|
|
|
|
0x1d82c63da, 0x0f20c0dfe, 0x09e4addf8, 0x0ba4fc28e,
|
|
|
|
0x039d3b296, 0x1384aa63a, 0x102f9b8a2, 0x1d82c63da,
|
|
|
|
0x14237f5e6, 0x01c291d04, 0x00d3b6092, 0x09e4addf8,
|
|
|
|
0x0c96cfdc0, 0x0740eef02, 0x18266e456, 0x039d3b296,
|
|
|
|
0x0daece73e, 0x0083a6eec, 0x0ab7aff2a, 0x102f9b8a2,
|
|
|
|
0x1248ea574, 0x1c1733996, 0x083348832, 0x14237f5e6,
|
|
|
|
0x12c743124, 0x02ad91c30, 0x0b9e02b86, 0x00d3b6092,
|
|
|
|
0x018b33a4e, 0x06992cea2, 0x1b331e26a, 0x0c96cfdc0,
|
|
|
|
0x17d35ba46, 0x07e908048, 0x1bf2e8b8a, 0x18266e456,
|
|
|
|
0x1a3e0968a, 0x11ed1f9d8, 0x0ce7f39f4, 0x0daece73e,
|
|
|
|
0x061d82e56, 0x0f1d0f55e, 0x0d270f1a2, 0x0ab7aff2a,
|
|
|
|
0x1c3f5f66c, 0x0a87ab8a8, 0x12ed0daac, 0x1248ea574,
|
|
|
|
0x065863b64, 0x08462d800, 0x11eef4f8e, 0x083348832,
|
|
|
|
0x1ee54f54c, 0x071d111a8, 0x0b3e32c28, 0x12c743124,
|
|
|
|
0x0064f7f26, 0x0ffd852c6, 0x0dd7e3b0c, 0x0b9e02b86,
|
|
|
|
0x0f285651c, 0x0dcb17aa4, 0x010746f3c, 0x018b33a4e,
|
|
|
|
0x1c24afea4, 0x0f37c5aee, 0x0271d9844, 0x1b331e26a,
|
|
|
|
0x08e766a0c, 0x06051d5a2, 0x093a5f730, 0x17d35ba46,
|
|
|
|
0x06cb08e5c, 0x11d5ca20e, 0x06b749fb2, 0x1bf2e8b8a,
|
|
|
|
0x1167f94f2, 0x021f3d99c, 0x0cec3662e, 0x1a3e0968a,
|
|
|
|
0x19329634a, 0x08f158014, 0x0e6fc4e6a, 0x0ce7f39f4,
|
|
|
|
0x08227bb8a, 0x1a5e82106, 0x0b0cd4768, 0x061d82e56,
|
|
|
|
0x13c2b89c4, 0x188815ab2, 0x0d7a4825c, 0x0d270f1a2,
|
|
|
|
0x10f5ff2ba, 0x105405f3e, 0x00167d312, 0x1c3f5f66c,
|
|
|
|
0x0f6076544, 0x0e9adf796, 0x026f6a60a, 0x12ed0daac,
|
|
|
|
0x1a2adb74e, 0x096638b34, 0x19d34af3a, 0x065863b64,
|
|
|
|
0x049c3cc9c, 0x1e50585a0, 0x068bce87a, 0x11eef4f8e,
|
|
|
|
0x1524fa6c6, 0x19f1c69dc, 0x16cba8aca, 0x1ee54f54c,
|
|
|
|
0x042d98888, 0x12913343e, 0x1329d9f7e, 0x0b3e32c28,
|
|
|
|
0x1b1c69528, 0x088f25a3a, 0x02178513a, 0x0064f7f26,
|
|
|
|
0x0e0ac139e, 0x04e36f0b0, 0x0170076fa, 0x0dd7e3b0c,
|
|
|
|
0x141a1a2e2, 0x0bd6f81f8, 0x16ad828b4, 0x0f285651c,
|
|
|
|
0x041d17b64, 0x19425cbba, 0x1fae1cc66, 0x010746f3c,
|
|
|
|
0x1a75b4b00, 0x18db37e8a, 0x0f872e54c, 0x1c24afea4,
|
|
|
|
0x01e41e9fc, 0x04c144932, 0x086d8e4d2, 0x0271d9844,
|
|
|
|
0x160f7af7a, 0x052148f02, 0x05bb8f1bc, 0x08e766a0c,
|
|
|
|
0x0a90fd27a, 0x0a3c6f37a, 0x0b3af077a, 0x093a5f730,
|
|
|
|
0x04984d782, 0x1d22c238e, 0x0ca6ef3ac, 0x06cb08e5c,
|
|
|
|
0x0234e0b26, 0x063ded06a, 0x1d88abd4a, 0x06b749fb2,
|
|
|
|
0x04597456a, 0x04d56973c, 0x0e9e28eb4, 0x1167f94f2,
|
|
|
|
0x07b3ff57a, 0x19385bf2e, 0x0c9c8b782, 0x0cec3662e,
|
|
|
|
0x13a9cba9e, 0x0e417f38a, 0x093e106a4, 0x19329634a,
|
|
|
|
0x167001a9c, 0x14e727980, 0x1ddffc5d4, 0x0e6fc4e6a,
|
|
|
|
0x00df04680, 0x0d104b8fc, 0x02342001e, 0x08227bb8a,
|
|
|
|
0x00a2a8d7e, 0x05b397730, 0x168763fa6, 0x0b0cd4768,
|
|
|
|
0x1ed5a407a, 0x0e78eb416, 0x0d2c3ed1a, 0x13c2b89c4,
|
|
|
|
0x0995a5724, 0x1641378f0, 0x19b1afbc4, 0x0d7a4825c,
|
|
|
|
0x109ffedc0, 0x08d96551c, 0x0f2271e60, 0x10f5ff2ba,
|
|
|
|
0x00b0bf8ca, 0x00bf80dd2, 0x123888b7a, 0x00167d312,
|
|
|
|
0x1e888f7dc, 0x18dcddd1c, 0x002ee03b2, 0x0f6076544,
|
|
|
|
0x183e8d8fe, 0x06a45d2b2, 0x133d7a042, 0x026f6a60a,
|
|
|
|
0x116b0f50c, 0x1dd3e10e8, 0x05fabe670, 0x1a2adb74e,
|
|
|
|
0x130004488, 0x0de87806c, 0x000bcf5f6, 0x19d34af3a,
|
|
|
|
0x18f0c7078, 0x014338754, 0x017f27698, 0x049c3cc9c,
|
|
|
|
0x058ca5f00, 0x15e3e77ee, 0x1af900c24, 0x068bce87a,
|
|
|
|
0x0b5cfca28, 0x0dd07448e, 0x0ded288f8, 0x1524fa6c6,
|
|
|
|
0x059f229bc, 0x1d8048348, 0x06d390dec, 0x16cba8aca,
|
|
|
|
0x037170390, 0x0a3e3e02c, 0x06353c1cc, 0x042d98888,
|
|
|
|
0x0c4584f5c, 0x0d73c7bea, 0x1f16a3418, 0x1329d9f7e,
|
|
|
|
0x0531377e2, 0x185137662, 0x1d8d9ca7c, 0x1b1c69528,
|
|
|
|
0x0b25b29f2, 0x18a08b5bc, 0x19fb2a8b0, 0x02178513a,
|
|
|
|
0x1a08fe6ac, 0x1da758ae0, 0x045cddf4e, 0x0e0ac139e,
|
|
|
|
0x1a91647f2, 0x169cf9eb0, 0x1a0f717c4, 0x0170076fa,
|
|
|
|
};
|
|
|
|
|
|
|
|
// Compute the crc32c value for buffer smaller than 8
|
2018-02-13 21:05:36 +01:00
|
|
|
#ifdef ROCKSDB_UBSAN_RUN
|
|
|
|
#if defined(__clang__)
|
|
|
|
__attribute__((__no_sanitize__("alignment")))
|
|
|
|
#elif defined(__GNUC__)
|
|
|
|
__attribute__((__no_sanitize_undefined__))
|
|
|
|
#endif
|
|
|
|
#endif
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
inline void align_to_8(
|
|
|
|
size_t len,
|
|
|
|
uint64_t& crc0, // crc so far, updated on return
|
|
|
|
const unsigned char*& next) { // next data pointer, updated on return
|
|
|
|
uint32_t crc32bit = static_cast<uint32_t>(crc0);
|
|
|
|
if (len & 0x04) {
|
|
|
|
crc32bit = _mm_crc32_u32(crc32bit, *(uint32_t*)next);
|
|
|
|
next += sizeof(uint32_t);
|
|
|
|
}
|
|
|
|
if (len & 0x02) {
|
|
|
|
crc32bit = _mm_crc32_u16(crc32bit, *(uint16_t*)next);
|
|
|
|
next += sizeof(uint16_t);
|
|
|
|
}
|
|
|
|
if (len & 0x01) {
|
|
|
|
crc32bit = _mm_crc32_u8(crc32bit, *(next));
|
|
|
|
next++;
|
|
|
|
}
|
|
|
|
crc0 = crc32bit;
|
|
|
|
}
|
|
|
|
|
|
|
|
//
|
|
|
|
// CombineCRC performs pclmulqdq multiplication of 2 partial CRC's and a well
|
|
|
|
// chosen constant and xor's these with the remaining CRC.
|
|
|
|
//
|
|
|
|
inline uint64_t CombineCRC(
|
|
|
|
size_t block_size,
|
|
|
|
uint64_t crc0,
|
|
|
|
uint64_t crc1,
|
|
|
|
uint64_t crc2,
|
|
|
|
const uint64_t* next2) {
|
|
|
|
const auto multiplier =
|
|
|
|
*(reinterpret_cast<const __m128i*>(clmul_constants) + block_size - 1);
|
|
|
|
const auto crc0_xmm = _mm_set_epi64x(0, crc0);
|
|
|
|
const auto res0 = _mm_clmulepi64_si128(crc0_xmm, multiplier, 0x00);
|
|
|
|
const auto crc1_xmm = _mm_set_epi64x(0, crc1);
|
|
|
|
const auto res1 = _mm_clmulepi64_si128(crc1_xmm, multiplier, 0x10);
|
|
|
|
const auto res = _mm_xor_si128(res0, res1);
|
|
|
|
crc0 = _mm_cvtsi128_si64(res);
|
|
|
|
crc0 = crc0 ^ *((uint64_t*)next2 - 1);
|
|
|
|
crc2 = _mm_crc32_u64(crc2, crc0);
|
|
|
|
return crc2;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Compute CRC-32C using the Intel hardware instruction.
|
2018-02-13 21:05:36 +01:00
|
|
|
#ifdef ROCKSDB_UBSAN_RUN
|
|
|
|
#if defined(__clang__)
|
|
|
|
__attribute__((__no_sanitize__("alignment")))
|
|
|
|
#elif defined(__GNUC__)
|
|
|
|
__attribute__((__no_sanitize_undefined__))
|
|
|
|
#endif
|
|
|
|
#endif
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
uint32_t crc32c_3way(uint32_t crc, const char* buf, size_t len) {
|
|
|
|
const unsigned char* next = (const unsigned char*)buf;
|
|
|
|
uint64_t count;
|
|
|
|
uint64_t crc0, crc1, crc2;
|
|
|
|
crc0 = crc ^ 0xffffffffu;
|
|
|
|
|
|
|
|
|
|
|
|
if (len >= 8) {
|
|
|
|
// if len > 216 then align and use triplets
|
|
|
|
if (len > 216) {
|
|
|
|
{
|
|
|
|
// Work on the bytes (< 8) before the first 8-byte alignment addr starts
|
|
|
|
uint64_t align_bytes = (8 - (uintptr_t)next) & 7;
|
|
|
|
len -= align_bytes;
|
|
|
|
align_to_8(align_bytes, crc0, next);
|
|
|
|
}
|
|
|
|
|
|
|
|
// Now work on the remaining blocks
|
|
|
|
count = len / 24; // number of triplets
|
|
|
|
len %= 24; // bytes remaining
|
|
|
|
uint64_t n = count >> 7; // #blocks = first block + full blocks
|
|
|
|
uint64_t block_size = count & 127;
|
|
|
|
if (block_size == 0) {
|
|
|
|
block_size = 128;
|
|
|
|
} else {
|
|
|
|
n++;
|
|
|
|
}
|
|
|
|
// points to the first byte of the next block
|
|
|
|
const uint64_t* next0 = (uint64_t*)next + block_size;
|
|
|
|
const uint64_t* next1 = next0 + block_size;
|
|
|
|
const uint64_t* next2 = next1 + block_size;
|
|
|
|
|
|
|
|
crc1 = crc2 = 0;
|
|
|
|
// Use Duff's device, a for() loop inside a switch()
|
|
|
|
// statement. This needs to execute at least once, round len
|
|
|
|
// down to nearest triplet multiple
|
|
|
|
switch (block_size) {
|
|
|
|
case 128:
|
|
|
|
do {
|
|
|
|
// jumps here for a full block of len 128
|
|
|
|
CRCtriplet(crc, next, -128);
|
2020-04-20 20:37:36 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 127:
|
|
|
|
// jumps here or below for the first block smaller
|
|
|
|
CRCtriplet(crc, next, -127);
|
2020-04-20 20:37:36 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 126:
|
|
|
|
CRCtriplet(crc, next, -126); // than 128
|
2020-04-20 20:37:36 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 125:
|
|
|
|
CRCtriplet(crc, next, -125);
|
2020-04-20 20:37:36 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 124:
|
|
|
|
CRCtriplet(crc, next, -124);
|
2020-04-20 20:37:36 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 123:
|
|
|
|
CRCtriplet(crc, next, -123);
|
2020-04-20 20:37:36 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 122:
|
|
|
|
CRCtriplet(crc, next, -122);
|
2020-04-20 20:37:36 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 121:
|
|
|
|
CRCtriplet(crc, next, -121);
|
2020-04-20 20:37:36 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 120:
|
|
|
|
CRCtriplet(crc, next, -120);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 119:
|
|
|
|
CRCtriplet(crc, next, -119);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 118:
|
|
|
|
CRCtriplet(crc, next, -118);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 117:
|
|
|
|
CRCtriplet(crc, next, -117);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 116:
|
|
|
|
CRCtriplet(crc, next, -116);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 115:
|
|
|
|
CRCtriplet(crc, next, -115);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 114:
|
|
|
|
CRCtriplet(crc, next, -114);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 113:
|
|
|
|
CRCtriplet(crc, next, -113);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 112:
|
|
|
|
CRCtriplet(crc, next, -112);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 111:
|
|
|
|
CRCtriplet(crc, next, -111);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 110:
|
|
|
|
CRCtriplet(crc, next, -110);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 109:
|
|
|
|
CRCtriplet(crc, next, -109);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 108:
|
|
|
|
CRCtriplet(crc, next, -108);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 107:
|
|
|
|
CRCtriplet(crc, next, -107);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 106:
|
|
|
|
CRCtriplet(crc, next, -106);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 105:
|
|
|
|
CRCtriplet(crc, next, -105);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 104:
|
|
|
|
CRCtriplet(crc, next, -104);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 103:
|
|
|
|
CRCtriplet(crc, next, -103);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 102:
|
|
|
|
CRCtriplet(crc, next, -102);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 101:
|
|
|
|
CRCtriplet(crc, next, -101);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 100:
|
|
|
|
CRCtriplet(crc, next, -100);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 99:
|
|
|
|
CRCtriplet(crc, next, -99);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 98:
|
|
|
|
CRCtriplet(crc, next, -98);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 97:
|
|
|
|
CRCtriplet(crc, next, -97);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 96:
|
|
|
|
CRCtriplet(crc, next, -96);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 95:
|
|
|
|
CRCtriplet(crc, next, -95);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 94:
|
|
|
|
CRCtriplet(crc, next, -94);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 93:
|
|
|
|
CRCtriplet(crc, next, -93);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 92:
|
|
|
|
CRCtriplet(crc, next, -92);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 91:
|
|
|
|
CRCtriplet(crc, next, -91);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 90:
|
|
|
|
CRCtriplet(crc, next, -90);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 89:
|
|
|
|
CRCtriplet(crc, next, -89);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 88:
|
|
|
|
CRCtriplet(crc, next, -88);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 87:
|
|
|
|
CRCtriplet(crc, next, -87);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 86:
|
|
|
|
CRCtriplet(crc, next, -86);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 85:
|
|
|
|
CRCtriplet(crc, next, -85);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 84:
|
|
|
|
CRCtriplet(crc, next, -84);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 83:
|
|
|
|
CRCtriplet(crc, next, -83);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 82:
|
|
|
|
CRCtriplet(crc, next, -82);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 81:
|
|
|
|
CRCtriplet(crc, next, -81);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 80:
|
|
|
|
CRCtriplet(crc, next, -80);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 79:
|
|
|
|
CRCtriplet(crc, next, -79);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 78:
|
|
|
|
CRCtriplet(crc, next, -78);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 77:
|
|
|
|
CRCtriplet(crc, next, -77);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 76:
|
|
|
|
CRCtriplet(crc, next, -76);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 75:
|
|
|
|
CRCtriplet(crc, next, -75);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 74:
|
|
|
|
CRCtriplet(crc, next, -74);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 73:
|
|
|
|
CRCtriplet(crc, next, -73);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 72:
|
|
|
|
CRCtriplet(crc, next, -72);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 71:
|
|
|
|
CRCtriplet(crc, next, -71);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 70:
|
|
|
|
CRCtriplet(crc, next, -70);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 69:
|
|
|
|
CRCtriplet(crc, next, -69);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 68:
|
|
|
|
CRCtriplet(crc, next, -68);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 67:
|
|
|
|
CRCtriplet(crc, next, -67);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 66:
|
|
|
|
CRCtriplet(crc, next, -66);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 65:
|
|
|
|
CRCtriplet(crc, next, -65);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 64:
|
|
|
|
CRCtriplet(crc, next, -64);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 63:
|
|
|
|
CRCtriplet(crc, next, -63);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 62:
|
|
|
|
CRCtriplet(crc, next, -62);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 61:
|
|
|
|
CRCtriplet(crc, next, -61);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 60:
|
|
|
|
CRCtriplet(crc, next, -60);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 59:
|
|
|
|
CRCtriplet(crc, next, -59);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 58:
|
|
|
|
CRCtriplet(crc, next, -58);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 57:
|
|
|
|
CRCtriplet(crc, next, -57);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 56:
|
|
|
|
CRCtriplet(crc, next, -56);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 55:
|
|
|
|
CRCtriplet(crc, next, -55);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 54:
|
|
|
|
CRCtriplet(crc, next, -54);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 53:
|
|
|
|
CRCtriplet(crc, next, -53);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 52:
|
|
|
|
CRCtriplet(crc, next, -52);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 51:
|
|
|
|
CRCtriplet(crc, next, -51);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 50:
|
|
|
|
CRCtriplet(crc, next, -50);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 49:
|
|
|
|
CRCtriplet(crc, next, -49);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 48:
|
|
|
|
CRCtriplet(crc, next, -48);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 47:
|
|
|
|
CRCtriplet(crc, next, -47);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 46:
|
|
|
|
CRCtriplet(crc, next, -46);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 45:
|
|
|
|
CRCtriplet(crc, next, -45);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 44:
|
|
|
|
CRCtriplet(crc, next, -44);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 43:
|
|
|
|
CRCtriplet(crc, next, -43);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 42:
|
|
|
|
CRCtriplet(crc, next, -42);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 41:
|
|
|
|
CRCtriplet(crc, next, -41);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 40:
|
|
|
|
CRCtriplet(crc, next, -40);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 39:
|
|
|
|
CRCtriplet(crc, next, -39);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 38:
|
|
|
|
CRCtriplet(crc, next, -38);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 37:
|
|
|
|
CRCtriplet(crc, next, -37);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 36:
|
|
|
|
CRCtriplet(crc, next, -36);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 35:
|
|
|
|
CRCtriplet(crc, next, -35);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 34:
|
|
|
|
CRCtriplet(crc, next, -34);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 33:
|
|
|
|
CRCtriplet(crc, next, -33);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 32:
|
|
|
|
CRCtriplet(crc, next, -32);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 31:
|
|
|
|
CRCtriplet(crc, next, -31);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 30:
|
|
|
|
CRCtriplet(crc, next, -30);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 29:
|
|
|
|
CRCtriplet(crc, next, -29);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 28:
|
|
|
|
CRCtriplet(crc, next, -28);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 27:
|
|
|
|
CRCtriplet(crc, next, -27);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 26:
|
|
|
|
CRCtriplet(crc, next, -26);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 25:
|
|
|
|
CRCtriplet(crc, next, -25);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 24:
|
|
|
|
CRCtriplet(crc, next, -24);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 23:
|
|
|
|
CRCtriplet(crc, next, -23);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 22:
|
|
|
|
CRCtriplet(crc, next, -22);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 21:
|
|
|
|
CRCtriplet(crc, next, -21);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 20:
|
|
|
|
CRCtriplet(crc, next, -20);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 19:
|
|
|
|
CRCtriplet(crc, next, -19);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 18:
|
|
|
|
CRCtriplet(crc, next, -18);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 17:
|
|
|
|
CRCtriplet(crc, next, -17);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 16:
|
|
|
|
CRCtriplet(crc, next, -16);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 15:
|
|
|
|
CRCtriplet(crc, next, -15);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 14:
|
|
|
|
CRCtriplet(crc, next, -14);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 13:
|
|
|
|
CRCtriplet(crc, next, -13);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 12:
|
|
|
|
CRCtriplet(crc, next, -12);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 11:
|
|
|
|
CRCtriplet(crc, next, -11);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 10:
|
|
|
|
CRCtriplet(crc, next, -10);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 9:
|
|
|
|
CRCtriplet(crc, next, -9);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 8:
|
|
|
|
CRCtriplet(crc, next, -8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 7:
|
|
|
|
CRCtriplet(crc, next, -7);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 6:
|
|
|
|
CRCtriplet(crc, next, -6);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 5:
|
|
|
|
CRCtriplet(crc, next, -5);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 4:
|
|
|
|
CRCtriplet(crc, next, -4);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 3:
|
|
|
|
CRCtriplet(crc, next, -3);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 2:
|
|
|
|
CRCtriplet(crc, next, -2);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 1:
|
|
|
|
CRCduplet(crc, next, -1); // the final triplet is actually only 2
|
|
|
|
//{ CombineCRC(); }
|
|
|
|
crc0 = CombineCRC(block_size, crc0, crc1, crc2, next2);
|
|
|
|
if (--n > 0) {
|
|
|
|
crc1 = crc2 = 0;
|
|
|
|
block_size = 128;
|
|
|
|
// points to the first byte of the next block
|
|
|
|
next0 = next2 + 128;
|
|
|
|
next1 = next0 + 128; // from here on all blocks are 128 long
|
|
|
|
next2 = next1 + 128;
|
|
|
|
}
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 0:;
|
|
|
|
} while (n > 0);
|
|
|
|
}
|
|
|
|
next = (const unsigned char*)next2;
|
|
|
|
}
|
|
|
|
uint64_t count2 = len >> 3; // 216 of less bytes is 27 or less singlets
|
|
|
|
len = len & 7;
|
|
|
|
next += (count2 * 8);
|
|
|
|
switch (count2) {
|
|
|
|
case 27:
|
|
|
|
CRCsinglet(crc0, next, -27 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 26:
|
|
|
|
CRCsinglet(crc0, next, -26 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 25:
|
|
|
|
CRCsinglet(crc0, next, -25 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 24:
|
|
|
|
CRCsinglet(crc0, next, -24 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 23:
|
|
|
|
CRCsinglet(crc0, next, -23 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 22:
|
|
|
|
CRCsinglet(crc0, next, -22 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 21:
|
|
|
|
CRCsinglet(crc0, next, -21 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 20:
|
|
|
|
CRCsinglet(crc0, next, -20 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 19:
|
|
|
|
CRCsinglet(crc0, next, -19 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 18:
|
|
|
|
CRCsinglet(crc0, next, -18 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 17:
|
|
|
|
CRCsinglet(crc0, next, -17 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 16:
|
|
|
|
CRCsinglet(crc0, next, -16 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 15:
|
|
|
|
CRCsinglet(crc0, next, -15 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 14:
|
|
|
|
CRCsinglet(crc0, next, -14 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 13:
|
|
|
|
CRCsinglet(crc0, next, -13 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 12:
|
|
|
|
CRCsinglet(crc0, next, -12 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 11:
|
|
|
|
CRCsinglet(crc0, next, -11 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 10:
|
|
|
|
CRCsinglet(crc0, next, -10 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 9:
|
|
|
|
CRCsinglet(crc0, next, -9 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 8:
|
|
|
|
CRCsinglet(crc0, next, -8 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 7:
|
|
|
|
CRCsinglet(crc0, next, -7 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 6:
|
|
|
|
CRCsinglet(crc0, next, -6 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 5:
|
|
|
|
CRCsinglet(crc0, next, -5 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 4:
|
|
|
|
CRCsinglet(crc0, next, -4 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 3:
|
|
|
|
CRCsinglet(crc0, next, -3 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 2:
|
|
|
|
CRCsinglet(crc0, next, -2 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 1:
|
|
|
|
CRCsinglet(crc0, next, -1 * 8);
|
2018-07-13 19:47:49 +02:00
|
|
|
FALLTHROUGH_INTENDED;
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
case 0:;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
{
|
|
|
|
align_to_8(len, crc0, next);
|
|
|
|
return (uint32_t)crc0 ^ 0xffffffffu;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
#endif //HAVE_SSE42 && HAVE_PCLMUL
|
|
|
|
|
|
|
|
static inline Function Choose_Extend() {
|
2019-04-30 19:56:06 +02:00
|
|
|
#ifdef HAVE_POWER8
|
|
|
|
return isAltiVec() ? ExtendPPCImpl : ExtendImpl<Slow_CRC32>;
|
|
|
|
#elif defined(__linux__) && defined(HAVE_ARM64_CRC)
|
|
|
|
if(crc32c_runtime_check()) {
|
2020-09-22 19:39:54 +02:00
|
|
|
pmull_runtime_flag = crc32c_pmull_runtime_check();
|
2019-04-30 19:56:06 +02:00
|
|
|
return ExtendARMImpl;
|
|
|
|
} else {
|
|
|
|
return ExtendImpl<Slow_CRC32>;
|
|
|
|
}
|
|
|
|
#else
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
if (isSSE42()) {
|
|
|
|
if (isPCLMULQDQ()) {
|
|
|
|
#if defined HAVE_SSE42 && defined HAVE_PCLMUL && !defined NO_THREEWAY_CRC32C
|
|
|
|
return crc32c_3way;
|
|
|
|
#else
|
|
|
|
return ExtendImpl<Fast_CRC32>; // Fast_CRC32 will check HAVE_SSE42 itself
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
else { // no runtime PCLMULQDQ support but has SSE42 support
|
|
|
|
return ExtendImpl<Fast_CRC32>;
|
|
|
|
}
|
|
|
|
} // end of isSSE42()
|
|
|
|
else {
|
|
|
|
return ExtendImpl<Slow_CRC32>;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
|
|
|
static Function ChosenExtend = Choose_Extend();
|
2014-04-03 00:15:57 +02:00
|
|
|
uint32_t Extend(uint32_t crc, const char* buf, size_t size) {
|
|
|
|
return ChosenExtend(crc, buf, size);
|
|
|
|
}
|
|
|
|
|
Port 3 way SSE4.2 crc32c implementation from Folly
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
2017-12-20 03:20:50 +01:00
|
|
|
|
2011-10-31 18:22:06 +01:00
|
|
|
} // namespace crc32c
|
2020-02-20 21:07:53 +01:00
|
|
|
} // namespace ROCKSDB_NAMESPACE
|