RocksDB CRC32c optimization with ARMv8 Intrinsic (#5221)

Summary:
1. Add Arm linear crc32c implemtation for RocksDB.
2. Arm runtime check for crc32
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5221

Differential Revision: D15013685

Pulled By: siying

fbshipit-source-id: 2c2983743d26656d93f212dc7c1a3cf66a1acf12

Makefile

@@ -137,6 +137,12 @@ CFLAGS +=  -DHAVE_POWER8
 HAVE_POWER8=1
 endif
 
+ifeq (,$(shell $(CXX) -fsyntax-only -march=armv8-a+crc -xc /dev/null 2>&1))
+CXXFLAGS += -march=armv8-a+crc
+CFLAGS += -march=armv8-a+crc
+ARMCRC_SOURCE=1
+endif
+
 # if we're compiling for release, compile without debug code (-DNDEBUG)
 ifeq ($(DEBUG_LEVEL),0)
 OPT += -DNDEBUG

src.mk

@@ -216,6 +216,11 @@ LIB_SOURCES =                                                   \
   utilities/write_batch_with_index/write_batch_with_index.cc    \
   utilities/write_batch_with_index/write_batch_with_index_internal.cc    \
 
+ifeq ($(ARMCRC_SOURCE),1)
+LIB_SOURCES +=\
+  util/crc32c_arm64.cc
+endif
+
 ifeq (,$(shell $(CXX) -fsyntax-only -maltivec -xc /dev/null 2>&1))
 LIB_SOURCES_ASM =\
   util/crc32c_ppc_asm.S

util/crc32c.cc

@@ -18,6 +18,8 @@
 #include "util/coding.h"
 #include "util/util.h"
 
+#include "util/crc32c_arm64.h"
+
 #ifdef __powerpc64__
 #include "util/crc32c_ppc.h"
 #include "util/crc32c_ppc_constants.h"
@@ -463,6 +465,11 @@ static bool isAltiVec() {
 }
 #endif
 
+#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
 
 std::string IsFastCrc32Supported() {
   bool has_fast_crc = false;
@@ -478,6 +485,14 @@ std::string IsFastCrc32Supported() {
   has_fast_crc = false;
   arch = "PPC";
 #endif
+#elif defined(__linux__) && defined(HAVE_ARM64_CRC)
+  if (crc32c_runtime_check()) {
+    has_fast_crc = true;
+    arch = "Arm64";
+  } else {
+    has_fast_crc = false;
+    arch = "Arm64";
+  }
 #else
   has_fast_crc = isSSE42();
   arch = "x86";
@@ -1200,7 +1215,15 @@ uint32_t crc32c_3way(uint32_t crc, const char* buf, size_t len) {
 #endif //HAVE_SSE42 && HAVE_PCLMUL
 
 static inline Function Choose_Extend() {
-#ifndef HAVE_POWER8
+#ifdef HAVE_POWER8
+  return isAltiVec() ? ExtendPPCImpl : ExtendImpl<Slow_CRC32>;
+#elif defined(__linux__) && defined(HAVE_ARM64_CRC)
+  if(crc32c_runtime_check()) {
+    return ExtendARMImpl;
+  } else {
+    return ExtendImpl<Slow_CRC32>;
+  }
+#else
   if (isSSE42()) {
     if (isPCLMULQDQ()) {
 #if defined HAVE_SSE42  && defined HAVE_PCLMUL && !defined NO_THREEWAY_CRC32C
@@ -1216,8 +1239,6 @@ static inline Function Choose_Extend() {
   else {
     return ExtendImpl<Slow_CRC32>;
   }
-#else  //HAVE_POWER8
-  return isAltiVec() ? ExtendPPCImpl : ExtendImpl<Slow_CRC32>;
 #endif
 }
 

util/crc32c_arm64.cc

@@ -0,0 +1,56 @@
+//  Copyright (c) 2018, Arm Limited and affiliates. All rights reserved.
+//  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).
+
+#include "util/crc32c_arm64.h"
+
+#if defined(__linux__) && defined(HAVE_ARM64_CRC)
+
+#include <asm/hwcap.h>
+#include <sys/auxv.h>
+#ifndef HWCAP_CRC32
+#define HWCAP_CRC32 (1 << 7)
+#endif
+uint32_t crc32c_runtime_check(void) {
+  uint64_t auxv = getauxval(AT_HWCAP);
+  return (auxv & HWCAP_CRC32) != 0;
+}
+
+uint32_t crc32c_arm64(uint32_t crc, unsigned char const *data,
+                             unsigned len) {
+  const uint8_t *buf1;
+  const uint16_t *buf2;
+  const uint32_t *buf4;
+  const uint64_t *buf8;
+
+  int64_t length = (int64_t)len;
+
+  crc ^= 0xffffffff;
+  buf8 = (const uint64_t *)data;
+  while ((length -= sizeof(uint64_t)) >= 0) {
+    crc = __crc32cd(crc, *buf8++);
+  }
+
+  /* The following is more efficient than the straight loop */
+  buf4 = (const uint32_t *)buf8;
+  if (length & sizeof(uint32_t)) {
+    crc = __crc32cw(crc, *buf4++);
+    length -= 4;
+  }
+
+  buf2 = (const uint16_t *)buf4;
+  if (length & sizeof(uint16_t)) {
+    crc = __crc32ch(crc, *buf2++);
+    length -= 2;
+  }
+
+  buf1 = (const uint8_t *)buf2;
+  if (length & sizeof(uint8_t))
+    crc = __crc32cb(crc, *buf1);
+
+  crc ^= 0xffffffff;
+  return crc;
+}
+
+#endif

util/crc32c_arm64.h

@@ -0,0 +1,21 @@
+//  Copyright (c) 2018, Arm Limited and affiliates. All rights reserved.
+//  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).
+
+#ifndef UTIL_CRC32C_ARM64_H
+#define UTIL_CRC32C_ARM64_H
+
+#include <inttypes.h>
+
+#if defined(__aarch64__) || defined(__AARCH64__)
+#ifdef __ARM_FEATURE_CRC32
+#define HAVE_ARM64_CRC
+#include <arm_acle.h>
+extern uint32_t crc32c_arm64(uint32_t crc, unsigned char const *data, unsigned len);
+extern uint32_t crc32c_runtime_check(void);
+#endif
+#endif
+
+
+#endif