rocksdb/build_tools/fbcode.gcc481.sh

87 lines
4.0 KiB
Bash
Raw Normal View History

#!/bin/sh
#
# Set environment variables so that we can compile rocksdb using
# fbcode settings. It uses the latest g++ compiler and also
# uses jemalloc
TOOLCHAIN_REV=53dc1fe83f84e9145b9ffb81b81aa7f6a49c87cc
CENTOS_VERSION=`rpm -q --qf "%{VERSION}" $(rpm -q --whatprovides redhat-release)`
if [ "$CENTOS_VERSION" = "6" ]; then
TOOLCHAIN_EXECUTABLES="/mnt/gvfs/third-party/$TOOLCHAIN_REV/centos6-native"
else
TOOLCHAIN_EXECUTABLES="/mnt/gvfs/third-party/$TOOLCHAIN_REV/centos5.2-native"
fi
TOOLCHAIN_LIB_BASE="/mnt/gvfs/third-party/$TOOLCHAIN_REV/gcc-4.8.1-glibc-2.17"
# location of libhdfs libraries
if test "$USE_HDFS"; then
JAVA_HOME="/usr/local/jdk-6u22-64"
JINCLUDE="-I$JAVA_HOME/include -I$JAVA_HOME/include/linux"
GLIBC_RUNTIME_PATH="/usr/local/fbcode/gcc-4.8.1-glibc-2.17"
HDFSLIB=" -Wl,--no-whole-archive hdfs/libhdfs.a -L$JAVA_HOME/jre/lib/amd64 "
HDFSLIB+=" -L$JAVA_HOME/jre/lib/amd64/server -L$GLIBC_RUNTIME_PATH/lib "
HDFSLIB+=" -ldl -lverify -ljava -ljvm "
fi
# location of libgcc
LIBGCC_INCLUDE=" -I $TOOLCHAIN_LIB_BASE/libgcc/libgcc-4.8.1/8aac7fc/include"
LIBGCC_LIBS=" -L $TOOLCHAIN_LIB_BASE/libgcc/libgcc-4.8.1/8aac7fc/libs"
# location of glibc
GLIBC_INCLUDE=" -I $TOOLCHAIN_LIB_BASE/glibc/glibc-2.17/99df8fc/include"
GLIBC_LIBS=" -L $TOOLCHAIN_LIB_BASE/glibc/glibc-2.17/99df8fc/lib"
# location of snappy headers and libraries
SNAPPY_INCLUDE=" -I $TOOLCHAIN_LIB_BASE/snappy/snappy-1.0.3/43d84e2/include"
SNAPPY_LIBS=" $TOOLCHAIN_LIB_BASE/snappy/snappy-1.0.3/43d84e2/lib/libsnappy.a"
# location of zlib headers and libraries
ZLIB_INCLUDE=" -I $TOOLCHAIN_LIB_BASE/zlib/zlib-1.2.5/c3f970a/include"
ZLIB_LIBS=" $TOOLCHAIN_LIB_BASE/zlib/zlib-1.2.5/c3f970a/lib/libz.a"
# location of bzip headers and libraries
BZIP_INCLUDE=" -I $TOOLCHAIN_LIB_BASE/bzip2/bzip2-1.0.6/c3f970a/include"
BZIP_LIBS=" $TOOLCHAIN_LIB_BASE/bzip2/bzip2-1.0.6/c3f970a/lib/libbz2.a"
LZ4_REV=065ec7e38fe83329031f6668c43bef83eff5808b
LZ4_INCLUDE=" -I /mnt/gvfs/third-party2/lz4/$LZ4_REV/r108/gcc-4.8.1-glibc-2.17/c3f970a/include"
LZ4_LIBS=" /mnt/gvfs/third-party2/lz4/$LZ4_REV/r108/gcc-4.8.1-glibc-2.17/c3f970a/lib/liblz4.a"
# location of gflags headers and libraries
GFLAGS_INCLUDE=" -I $TOOLCHAIN_LIB_BASE/gflags/gflags-1.6/c3f970a/include"
GFLAGS_LIBS=" $TOOLCHAIN_LIB_BASE/gflags/gflags-1.6/c3f970a/lib/libgflags.a"
# location of jemalloc
JEMALLOC_INCLUDE=" -I $TOOLCHAIN_LIB_BASE/jemalloc/jemalloc-3.4.1/4d53c6f/include/"
JEMALLOC_LIB=" -Wl,--whole-archive $TOOLCHAIN_LIB_BASE/jemalloc/jemalloc-3.4.1/4d53c6f/lib/libjemalloc.a"
Adding NUMA support to db_bench tests Summary: Changes: - Adding numa_aware flag to db_bench.cc - Using numa.h library to bind memory and cpu of threads to a fixed NUMA node Result: There seems to be no significant change in the micros/op time with numa_aware enabled. I also tried this with other implementations, including a combination of pthread_setaffinity_np, sched_setaffinity and set_mempolicy methods. It'd be great if someone could point out where I'm going wrong and if we can achieve a better micors/op. Test Plan: Ran db_bench tests using following command: ./db_bench --db=/mnt/tmp --num_levels=6 --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --block_size=4096 --cache_size=17179869184 --cache_numshardbits=6 --compression_type=none --compression_ratio=1 --min_level_to_compress=-1 --disable_seek_compaction=1 --hard_rate_limit=2 --write_buffer_size=134217728 --max_write_buffer_number=2 --level0_file_num_compaction_trigger=8 --target_file_size_base=134217728 --max_bytes_for_level_base=1073741824 --disable_wal=0 --wal_dir=/mnt/tmp --sync=0 --disable_data_sync=1 --verify_checksum=1 --delete_obsolete_files_period_micros=314572800 --max_grandparent_overlap_factor=10 --max_background_compactions=4 --max_background_flushes=0 --level0_slowdown_writes_trigger=16 --level0_stop_writes_trigger=24 --statistics=0 --stats_per_interval=0 --stats_interval=1048576 --histogram=0 --use_plain_table=1 --open_files=-1 --mmap_read=1 --mmap_write=0 --memtablerep=prefix_hash --bloom_bits=10 --bloom_locality=1 --perf_level=0 --duration=300 --benchmarks=readwhilewriting --use_existing_db=1 --num=157286400 --threads=24 --writes_per_second=10240 --numa_aware=[False/True] The tests were run in private devserver with 24 cores and the db was prepopulated using filluniquerandom test. The tests resulted in 0.145 us/op with numa_aware=False and 0.161 us/op with numa_aware=True. Reviewers: sdong, yhchiang, ljin, igor Reviewed By: ljin, igor Subscribers: igor, leveldb Differential Revision: https://reviews.facebook.net/D19353
2014-07-07 19:53:31 +02:00
# location of numa
NUMA_REV=829d10dac0230f99cd7e1778869d2adf3da24b65
NUMA_INCLUDE=" -I /mnt/gvfs/third-party2/numa/$NUMA_REV/2.0.8/gcc-4.8.1-glibc-2.17/c3f970a/include/"
NUMA_LIB=" /mnt/gvfs/third-party2/numa/$NUMA_REV/2.0.8/gcc-4.8.1-glibc-2.17/c3f970a/lib/libnuma.a"
# use Intel SSE support for checksum calculations
export USE_SSE=" -msse -msse4.2 "
CC="$TOOLCHAIN_EXECUTABLES/gcc/gcc-4.8.1/cc6c9dc/bin/gcc"
Adding NUMA support to db_bench tests Summary: Changes: - Adding numa_aware flag to db_bench.cc - Using numa.h library to bind memory and cpu of threads to a fixed NUMA node Result: There seems to be no significant change in the micros/op time with numa_aware enabled. I also tried this with other implementations, including a combination of pthread_setaffinity_np, sched_setaffinity and set_mempolicy methods. It'd be great if someone could point out where I'm going wrong and if we can achieve a better micors/op. Test Plan: Ran db_bench tests using following command: ./db_bench --db=/mnt/tmp --num_levels=6 --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --block_size=4096 --cache_size=17179869184 --cache_numshardbits=6 --compression_type=none --compression_ratio=1 --min_level_to_compress=-1 --disable_seek_compaction=1 --hard_rate_limit=2 --write_buffer_size=134217728 --max_write_buffer_number=2 --level0_file_num_compaction_trigger=8 --target_file_size_base=134217728 --max_bytes_for_level_base=1073741824 --disable_wal=0 --wal_dir=/mnt/tmp --sync=0 --disable_data_sync=1 --verify_checksum=1 --delete_obsolete_files_period_micros=314572800 --max_grandparent_overlap_factor=10 --max_background_compactions=4 --max_background_flushes=0 --level0_slowdown_writes_trigger=16 --level0_stop_writes_trigger=24 --statistics=0 --stats_per_interval=0 --stats_interval=1048576 --histogram=0 --use_plain_table=1 --open_files=-1 --mmap_read=1 --mmap_write=0 --memtablerep=prefix_hash --bloom_bits=10 --bloom_locality=1 --perf_level=0 --duration=300 --benchmarks=readwhilewriting --use_existing_db=1 --num=157286400 --threads=24 --writes_per_second=10240 --numa_aware=[False/True] The tests were run in private devserver with 24 cores and the db was prepopulated using filluniquerandom test. The tests resulted in 0.145 us/op with numa_aware=False and 0.161 us/op with numa_aware=True. Reviewers: sdong, yhchiang, ljin, igor Reviewed By: ljin, igor Subscribers: igor, leveldb Differential Revision: https://reviews.facebook.net/D19353
2014-07-07 19:53:31 +02:00
CXX="$TOOLCHAIN_EXECUTABLES/gcc/gcc-4.8.1/cc6c9dc/bin/g++ $JINCLUDE $SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP_INCLUDE $LZ4_INCLUDE $GFLAGS_INCLUDE $NUMA_INCLUDE"
AR=$TOOLCHAIN_EXECUTABLES/binutils/binutils-2.21.1/da39a3e/bin/ar
RANLIB=$TOOLCHAIN_EXECUTABLES/binutils/binutils-2.21.1/da39a3e/bin/ranlib
CFLAGS="-B$TOOLCHAIN_EXECUTABLES/binutils/binutils-2.21.1/da39a3e/bin/gold -m64 -mtune=generic"
CFLAGS+=" $LIBGCC_INCLUDE $GLIBC_INCLUDE"
CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_FALLOCATE_PRESENT"
Adding NUMA support to db_bench tests Summary: Changes: - Adding numa_aware flag to db_bench.cc - Using numa.h library to bind memory and cpu of threads to a fixed NUMA node Result: There seems to be no significant change in the micros/op time with numa_aware enabled. I also tried this with other implementations, including a combination of pthread_setaffinity_np, sched_setaffinity and set_mempolicy methods. It'd be great if someone could point out where I'm going wrong and if we can achieve a better micors/op. Test Plan: Ran db_bench tests using following command: ./db_bench --db=/mnt/tmp --num_levels=6 --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --block_size=4096 --cache_size=17179869184 --cache_numshardbits=6 --compression_type=none --compression_ratio=1 --min_level_to_compress=-1 --disable_seek_compaction=1 --hard_rate_limit=2 --write_buffer_size=134217728 --max_write_buffer_number=2 --level0_file_num_compaction_trigger=8 --target_file_size_base=134217728 --max_bytes_for_level_base=1073741824 --disable_wal=0 --wal_dir=/mnt/tmp --sync=0 --disable_data_sync=1 --verify_checksum=1 --delete_obsolete_files_period_micros=314572800 --max_grandparent_overlap_factor=10 --max_background_compactions=4 --max_background_flushes=0 --level0_slowdown_writes_trigger=16 --level0_stop_writes_trigger=24 --statistics=0 --stats_per_interval=0 --stats_interval=1048576 --histogram=0 --use_plain_table=1 --open_files=-1 --mmap_read=1 --mmap_write=0 --memtablerep=prefix_hash --bloom_bits=10 --bloom_locality=1 --perf_level=0 --duration=300 --benchmarks=readwhilewriting --use_existing_db=1 --num=157286400 --threads=24 --writes_per_second=10240 --numa_aware=[False/True] The tests were run in private devserver with 24 cores and the db was prepopulated using filluniquerandom test. The tests resulted in 0.145 us/op with numa_aware=False and 0.161 us/op with numa_aware=True. Reviewers: sdong, yhchiang, ljin, igor Reviewed By: ljin, igor Subscribers: igor, leveldb Differential Revision: https://reviews.facebook.net/D19353
2014-07-07 19:53:31 +02:00
CFLAGS+=" -DSNAPPY -DGFLAGS=google -DZLIB -DBZIP2 -DLZ4 -DNUMA"
EXEC_LDFLAGS="-Wl,--dynamic-linker,/usr/local/fbcode/gcc-4.8.1-glibc-2.17/lib/ld.so"
EXEC_LDFLAGS+=" -Wl,--no-whole-archive $TOOLCHAIN_LIB_BASE/libunwind/libunwind-1.0.1/675d945/lib/libunwind.a"
Adding NUMA support to db_bench tests Summary: Changes: - Adding numa_aware flag to db_bench.cc - Using numa.h library to bind memory and cpu of threads to a fixed NUMA node Result: There seems to be no significant change in the micros/op time with numa_aware enabled. I also tried this with other implementations, including a combination of pthread_setaffinity_np, sched_setaffinity and set_mempolicy methods. It'd be great if someone could point out where I'm going wrong and if we can achieve a better micors/op. Test Plan: Ran db_bench tests using following command: ./db_bench --db=/mnt/tmp --num_levels=6 --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --block_size=4096 --cache_size=17179869184 --cache_numshardbits=6 --compression_type=none --compression_ratio=1 --min_level_to_compress=-1 --disable_seek_compaction=1 --hard_rate_limit=2 --write_buffer_size=134217728 --max_write_buffer_number=2 --level0_file_num_compaction_trigger=8 --target_file_size_base=134217728 --max_bytes_for_level_base=1073741824 --disable_wal=0 --wal_dir=/mnt/tmp --sync=0 --disable_data_sync=1 --verify_checksum=1 --delete_obsolete_files_period_micros=314572800 --max_grandparent_overlap_factor=10 --max_background_compactions=4 --max_background_flushes=0 --level0_slowdown_writes_trigger=16 --level0_stop_writes_trigger=24 --statistics=0 --stats_per_interval=0 --stats_interval=1048576 --histogram=0 --use_plain_table=1 --open_files=-1 --mmap_read=1 --mmap_write=0 --memtablerep=prefix_hash --bloom_bits=10 --bloom_locality=1 --perf_level=0 --duration=300 --benchmarks=readwhilewriting --use_existing_db=1 --num=157286400 --threads=24 --writes_per_second=10240 --numa_aware=[False/True] The tests were run in private devserver with 24 cores and the db was prepopulated using filluniquerandom test. The tests resulted in 0.145 us/op with numa_aware=False and 0.161 us/op with numa_aware=True. Reviewers: sdong, yhchiang, ljin, igor Reviewed By: ljin, igor Subscribers: igor, leveldb Differential Revision: https://reviews.facebook.net/D19353
2014-07-07 19:53:31 +02:00
EXEC_LDFLAGS+=" $HDFSLIB $SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $GFLAGS_LIBS $NUMA_LIB"
PLATFORM_LDFLAGS="$LIBGCC_LIBS $GLIBC_LIBS "
EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $GFLAGS_LIBS"
VALGRIND_VER="$TOOLCHAIN_LIB_BASE/valgrind/valgrind-3.8.1/c3f970a/bin/"
export CC CXX AR RANLIB CFLAGS EXEC_LDFLAGS EXEC_LDFLAGS_SHARED VALGRIND_VER JEMALLOC_LIB JEMALLOC_INCLUDE