c/mirror_zfs
1
0
Fork 0
mirror of https://git.proxmox.com/git/mirror_zfs.git synced 2024-11-17 10:01:01 +03:00
Code Issues Packages Projects Releases Wiki Activity
5ba4025a8d
mirror_zfs/module/Makefile.bsd
Richard Yao 677c6f8457
btree: Implement faster binary search algorithm 
This implements a binary search algorithm for B-Trees that reduces
branching to the absolute minimum necessary for a binary search
algorithm. It also enables the compiler to inline the comparator to
ensure that the only slowdown when doing binary search is from waiting
for memory accesses. Additionally, it instructs the compiler to unroll
the loop, which gives an additional 40% improve with Clang and 8%
improvement with GCC.

Consumers must opt into using the faster algorithm. At present, only
B-Trees used inside kernel code have been modified to use the faster
algorithm.

Micro-benchmarks suggest that this can improve binary search performance
by up to 3.5 times when compiling with Clang 16 and up to 1.9 times when
compiling with GCC 12.2.

Reviewed-by: Alexander Motin <mav@FreeBSD.org>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Yao <richard.yao@alumni.stonybrook.edu>
Closes #14866
2023-05-26 10:03:12 -07:00

545 lines
13 KiB
Makefile
Raw Blame History

.if !defined(WITH_CTF)
WITH_CTF=1
.endif
.include <bsd.sys.mk>
SRCDIR=${.CURDIR}
INCDIR=${.CURDIR:H}/include
KMOD= openzfs
.PATH: ${SRCDIR}/avl \
${SRCDIR}/lua \
${SRCDIR}/nvpair \
${SRCDIR}/icp/algs/blake3 \
${SRCDIR}/icp/algs/edonr \
${SRCDIR}/icp/algs/sha2 \
${SRCDIR}/icp/asm-aarch64/blake3 \
${SRCDIR}/icp/asm-aarch64/sha2 \
${SRCDIR}/icp/asm-arm/sha2 \
${SRCDIR}/icp/asm-ppc64/sha2 \
${SRCDIR}/icp/asm-ppc64/blake3 \
${SRCDIR}/icp/asm-x86_64/blake3 \
${SRCDIR}/icp/asm-x86_64/sha2 \
${SRCDIR}/os/freebsd/spl \
${SRCDIR}/os/freebsd/zfs \
${SRCDIR}/unicode \
${SRCDIR}/zcommon \
${SRCDIR}/zfs \
${SRCDIR}/zstd \
${SRCDIR}/zstd/lib/common \
${SRCDIR}/zstd/lib/compress \
${SRCDIR}/zstd/lib/decompress
CFLAGS+= -I${INCDIR}
CFLAGS+= -I${SRCDIR}/icp/include
CFLAGS+= -I${INCDIR}/os/freebsd
CFLAGS+= -I${INCDIR}/os/freebsd/spl
CFLAGS+= -I${INCDIR}/os/freebsd/zfs
CFLAGS+= -I${SRCDIR}/zstd/include
CFLAGS+= -include ${INCDIR}/os/freebsd/spl/sys/ccompile.h
CFLAGS+= -I${.CURDIR}
CFLAGS+= -D__KERNEL__ -DFREEBSD_NAMECACHE -DBUILDING_ZFS -D__BSD_VISIBLE=1 \
-DHAVE_UIO_ZEROCOPY -DWITHOUT_NETDUMP -D__KERNEL -D_SYS_CONDVAR_H_ \
-D_SYS_VMEM_H_ -DKDTRACE_HOOKS -DCOMPAT_FREEBSD11
.if ${MACHINE_ARCH} == "amd64"
CFLAGS+= -D__x86_64 -DHAVE_SSE2 -DHAVE_SSSE3 -DHAVE_SSE4_1 -DHAVE_SSE4_2 \
-DHAVE_AVX -DHAVE_AVX2 -DHAVE_AVX512F -DHAVE_AVX512VL -DHAVE_AVX512BW
.endif
.if defined(WITH_DEBUG) && ${WITH_DEBUG} == "true"
CFLAGS+= -DZFS_DEBUG -g
.if defined(WITH_INVARIANTS) && ${WITH_INVARIANTS} == "true"
CFLAGS+= -DINVARIANTS -DWITNESS -DOPENSOLARIS_WITNESS
.endif
.if defined(WITH_O0) && ${WITH_O0} == "true"
CFLAGS+= -O0
.endif
.else
CFLAGS += -DNDEBUG
.endif
.if defined(WITH_VFS_DEBUG) && ${WITH_VFS_DEBUG} == "true"
# kernel must also be built with this option for this to work
CFLAGS+= -DDEBUG_VFS_LOCKS
.endif
.if defined(WITH_GCOV) && ${WITH_GCOV} == "true"
CFLAGS+= -fprofile-arcs -ftest-coverage
.endif
DEBUG_FLAGS=-g
.if ${MACHINE_ARCH} == "i386" || ${MACHINE_ARCH} == "powerpc" || \
${MACHINE_ARCH} == "powerpcspe" || ${MACHINE_ARCH} == "arm"
CFLAGS+= -DBITS_PER_LONG=32
.else
CFLAGS+= -DBITS_PER_LONG=64
.endif
SRCS= vnode_if.h device_if.h bus_if.h
# avl
SRCS+= avl.c
# icp
SRCS+= edonr.c
#icp/algs/blake3
SRCS+= blake3.c \
blake3_generic.c \
blake3_impl.c
#icp/asm-aarch64/blake3
SRCS+= b3_aarch64_sse2.S \
b3_aarch64_sse41.S
#icp/asm-ppc64/blake3
SRCS+= b3_ppc64le_sse2.S \
b3_ppc64le_sse41.S
#icp/asm-x86_64/blake3
SRCS+= blake3_avx2.S \
blake3_avx512.S \
blake3_sse2.S \
blake3_sse41.S
#icp/algs/sha2
SRCS+= sha2_generic.c \
sha256_impl.c \
sha512_impl.c
#icp/asm-arm/sha2
SRCS+= sha256-armv7.S \
sha512-armv7.S
#icp/asm-aarch64/sha2
SRCS+= sha256-armv8.S \
sha512-armv8.S
#icp/asm-ppc64/sha2
SRCS+= sha256-p8.S \
sha512-p8.S \
sha256-ppc.S \
sha512-ppc.S
#icp/asm-x86_64/sha2
SRCS+= sha256-x86_64.S \
sha512-x86_64.S
#lua
SRCS+= lapi.c \
lauxlib.c \
lbaselib.c \
lcode.c \
lcompat.c \
lcorolib.c \
lctype.c \
ldebug.c \
ldo.c \
lfunc.c \
lgc.c \
llex.c \
lmem.c \
lobject.c \
lopcodes.c \
lparser.c \
lstate.c \
lstring.c \
lstrlib.c \
ltable.c \
ltablib.c \
ltm.c \
lvm.c \
lzio.c
#nvpair
SRCS+= nvpair.c \
fnvpair.c \
nvpair_alloc_spl.c \
nvpair_alloc_fixed.c
#os/freebsd/spl
SRCS+= acl_common.c \
callb.c \
list.c \
spl_acl.c \
spl_cmn_err.c \
spl_dtrace.c \
spl_kmem.c \
spl_kstat.c \
spl_misc.c \
spl_policy.c \
spl_procfs_list.c \
spl_string.c \
spl_sunddi.c \
spl_sysevent.c \
spl_taskq.c \
spl_uio.c \
spl_vfs.c \
spl_vm.c \
spl_zlib.c \
spl_zone.c
.if ${MACHINE_ARCH} == "i386" || ${MACHINE_ARCH} == "powerpc" || \
${MACHINE_ARCH} == "powerpcspe" || ${MACHINE_ARCH} == "arm"
SRCS+= spl_atomic.c
.endif
#os/freebsd/zfs
SRCS+= abd_os.c \
arc_os.c \
crypto_os.c \
dmu_os.c \
event_os.c \
hkdf.c \
kmod_core.c \
spa_os.c \
sysctl_os.c \
vdev_file.c \
vdev_geom.c \
vdev_label_os.c \
zfs_acl.c \
zfs_ctldir.c \
zfs_debug.c \
zfs_dir.c \
zfs_ioctl_compat.c \
zfs_ioctl_os.c \
zfs_racct.c \
zfs_vfsops.c \
zfs_vnops_os.c \
zfs_znode.c \
zio_crypt.c \
zvol_os.c
#unicode
SRCS+= uconv.c \
u8_textprep.c
#zcommon
SRCS+= zfeature_common.c \
zfs_comutil.c \
zfs_deleg.c \
zfs_fletcher.c \
zfs_fletcher_avx512.c \
zfs_fletcher_intel.c \
zfs_fletcher_sse.c \
zfs_fletcher_superscalar.c \
zfs_fletcher_superscalar4.c \
zfs_namecheck.c \
zfs_prop.c \
zpool_prop.c \
zprop_common.c
#zfs
SRCS+= abd.c \
aggsum.c \
arc.c \
blake3_zfs.c \
blkptr.c \
bplist.c \
bpobj.c \
brt.c \
btree.c \
cityhash.c \
dbuf.c \
dbuf_stats.c \
bptree.c \
bqueue.c \
dataset_kstats.c \
ddt.c \
ddt_zap.c \
dmu.c \
dmu_diff.c \
dmu_object.c \
dmu_objset.c \
dmu_recv.c \
dmu_redact.c \
dmu_send.c \
dmu_traverse.c \
dmu_tx.c \
dmu_zfetch.c \
dnode.c \
dnode_sync.c \
dsl_dataset.c \
dsl_deadlist.c \
dsl_deleg.c \
dsl_bookmark.c \
dsl_dir.c \
dsl_crypt.c \
dsl_destroy.c \
dsl_pool.c \
dsl_prop.c \
dsl_scan.c \
dsl_synctask.c \
dsl_userhold.c \
edonr_zfs.c \
fm.c \
gzip.c \
lzjb.c \
lz4.c \
lz4_zfs.c \
metaslab.c \
mmp.c \
multilist.c \
objlist.c \
pathname.c \
range_tree.c \
refcount.c \
rrwlock.c \
sa.c \
sha2_zfs.c \
skein_zfs.c \
spa.c \
spa_checkpoint.c \
spa_config.c \
spa_errlog.c \
spa_history.c \
spa_log_spacemap.c \
spa_misc.c \
spa_stats.c \
space_map.c \
space_reftree.c \
txg.c \
uberblock.c \
unique.c \
vdev.c \
vdev_cache.c \
vdev_draid.c \
vdev_draid_rand.c \
vdev_indirect.c \
vdev_indirect_births.c \
vdev_indirect_mapping.c \
vdev_initialize.c \
vdev_label.c \
vdev_mirror.c \
vdev_missing.c \
vdev_queue.c \
vdev_raidz.c \
vdev_raidz_math.c \
vdev_raidz_math_scalar.c \
vdev_raidz_math_avx2.c \
vdev_raidz_math_avx512bw.c \
vdev_raidz_math_avx512f.c \
vdev_raidz_math_sse2.c \
vdev_raidz_math_ssse3.c \
vdev_rebuild.c \
vdev_removal.c \
vdev_root.c \
vdev_trim.c \
zap.c \
zap_leaf.c \
zap_micro.c \
zcp.c \
zcp_get.c \
zcp_global.c \
zcp_iter.c \
zcp_set.c \
zcp_synctask.c \
zfeature.c \
zfs_byteswap.c \
zfs_chksum.c \
zfs_file_os.c \
zfs_fm.c \
zfs_fuid.c \
zfs_impl.c \
zfs_ioctl.c \
zfs_log.c \
zfs_onexit.c \
zfs_quota.c \
zfs_ratelimit.c \
zfs_replay.c \
zfs_rlock.c \
zfs_sa.c \
zfs_vnops.c \
zil.c \
zio.c \
zio_checksum.c \
zio_compress.c \
zio_inject.c \
zle.c \
zrlock.c \
zthr.c \
zvol.c
#zstd
SRCS+= zfs_zstd.c \
entropy_common.c \
error_private.c \
fse_compress.c \
fse_decompress.c \
hist.c \
huf_compress.c \
huf_decompress.c \
pool.c \
xxhash.c \
zstd_common.c \
zstd_compress.c \
zstd_compress_literals.c \
zstd_compress_sequences.c \
zstd_compress_superblock.c \
zstd_ddict.c \
zstd_decompress.c \
zstd_decompress_block.c \
zstd_double_fast.c \
zstd_fast.c \
zstd_lazy.c \
zstd_ldm.c \
zstd_opt.c
beforeinstall:
.if ${MK_DEBUG_FILES} != "no"
mtree -eu \
-f /etc/mtree/BSD.debug.dist \
-p ${DESTDIR}/usr/lib
.endif
.include <bsd.kmod.mk>
# Generated binary search code is particularly bad with this optimization.
# Oddly, range_tree.c is not affected when unrolling is not done and dsl_scan.c
# is not affected when unrolling is done.
# Disable it until the following upstream issue is resolved:
# https://github.com/llvm/llvm-project/issues/62790
.if ${CC} == "clang"
.if ${MACHINE_ARCH} == "i386" || ${MACHINE_ARCH} == "amd64"
CFLAGS.dsl_scan.c= -mllvm -x86-cmov-converter=false
CFLAGS.metaslab.c= -mllvm -x86-cmov-converter=false
CFLAGS.range_tree.c= -mllvm -x86-cmov-converter=false
CFLAGS.zap_micro.c= -mllvm -x86-cmov-converter=false
.endif
.endif
CFLAGS.sysctl_os.c= -include ../zfs_config.h
CFLAGS.xxhash.c+= -include ${SYSDIR}/sys/_null.h
CFLAGS.gcc+= -Wno-pointer-to-int-cast
CFLAGS.abd.c= -Wno-cast-qual
CFLAGS.ddt.c= -Wno-cast-qual
CFLAGS.dmu.c= -Wno-cast-qual
CFLAGS.dmu_traverse.c= -Wno-cast-qual
CFLAGS.dnode.c= ${NO_WUNUSED_BUT_SET_VARIABLE}
CFLAGS.dsl_deadlist.c= -Wno-cast-qual
CFLAGS.dsl_dir.c= -Wno-cast-qual
CFLAGS.dsl_prop.c= -Wno-cast-qual
CFLAGS.edonr.c= -Wno-cast-qual
CFLAGS.fm.c= -Wno-cast-qual
CFLAGS.hist.c= -U__BMI__ -fno-tree-vectorize ${NO_WBITWISE_INSTEAD_OF_LOGICAL}
CFLAGS.lapi.c= -Wno-cast-qual
CFLAGS.lcompat.c= -Wno-cast-qual
CFLAGS.ldo.c= ${NO_WINFINITE_RECURSION}
CFLAGS.lobject.c= -Wno-cast-qual
CFLAGS.ltable.c= -Wno-cast-qual
CFLAGS.lvm.c= -Wno-cast-qual
CFLAGS.lz4.c= -Wno-cast-qual
CFLAGS.lz4_zfs.c= -Wno-cast-qual
CFLAGS.nvpair.c= -Wno-cast-qual -DHAVE_RPC_TYPES ${NO_WSTRINGOP_OVERREAD}
CFLAGS.pool.c+= ${NO_WBITWISE_INSTEAD_OF_LOGICAL}
CFLAGS.pool.c= -U__BMI__ -fno-tree-vectorize
CFLAGS.spa.c= -Wno-cast-qual
CFLAGS.spa_misc.c= -Wno-cast-qual
CFLAGS.spl_string.c= -Wno-cast-qual
CFLAGS.spl_vm.c= -Wno-cast-qual
CFLAGS.spl_zlib.c= -Wno-cast-qual
CFLAGS.u8_textprep.c= -Wno-cast-qual
CFLAGS.vdev_draid.c= -Wno-cast-qual
CFLAGS.vdev_raidz.c= -Wno-cast-qual
CFLAGS.vdev_raidz_math.c= -Wno-cast-qual
CFLAGS.vdev_raidz_math_avx2.c= -Wno-cast-qual -Wno-duplicate-decl-specifier
CFLAGS.vdev_raidz_math_avx512f.c= -Wno-cast-qual -Wno-duplicate-decl-specifier
CFLAGS.vdev_raidz_math_scalar.c= -Wno-cast-qual
CFLAGS.vdev_raidz_math_sse2.c= -Wno-cast-qual -Wno-duplicate-decl-specifier
CFLAGS.zap_leaf.c= -Wno-cast-qual
CFLAGS.zap_micro.c= -Wno-cast-qual
CFLAGS.zcp.c= -Wno-cast-qual
CFLAGS.zfs_fletcher.c= -Wno-cast-qual -Wno-pointer-arith
CFLAGS.zfs_fletcher_avx512.c= -Wno-cast-qual -Wno-pointer-arith
CFLAGS.zfs_fletcher_intel.c= -Wno-cast-qual -Wno-pointer-arith
CFLAGS.zfs_fletcher_sse.c= -Wno-cast-qual -Wno-pointer-arith
CFLAGS.zfs_fm.c= -Wno-cast-qual ${NO_WUNUSED_BUT_SET_VARIABLE}
CFLAGS.zfs_ioctl.c= -Wno-cast-qual
CFLAGS.zfs_log.c= -Wno-cast-qual
CFLAGS.zfs_vnops_os.c= -Wno-pointer-arith
CFLAGS.zfs_zstd.c= -Wno-cast-qual -Wno-pointer-arith
CFLAGS.zil.c= -Wno-cast-qual
CFLAGS.zio.c= -Wno-cast-qual
CFLAGS.zprop_common.c= -Wno-cast-qual
CFLAGS.zrlock.c= -Wno-cast-qual
#zstd
CFLAGS.entropy_common.c= -U__BMI__ -fno-tree-vectorize ${NO_WBITWISE_INSTEAD_OF_LOGICAL}
CFLAGS.error_private.c= -U__BMI__ -fno-tree-vectorize ${NO_WBITWISE_INSTEAD_OF_LOGICAL}
CFLAGS.fse_compress.c= -U__BMI__ -fno-tree-vectorize ${NO_WBITWISE_INSTEAD_OF_LOGICAL} ${NO_WUNUSED_BUT_SET_VARIABLE}
CFLAGS.fse_decompress.c= -U__BMI__ -fno-tree-vectorize ${NO_WBITWISE_INSTEAD_OF_LOGICAL}
CFLAGS.huf_compress.c= -U__BMI__ -fno-tree-vectorize ${NO_WBITWISE_INSTEAD_OF_LOGICAL}
CFLAGS.huf_decompress.c= -U__BMI__ -fno-tree-vectorize ${NO_WBITWISE_INSTEAD_OF_LOGICAL}
CFLAGS.xxhash.c+= -U__BMI__ -fno-tree-vectorize
CFLAGS.xxhash.c+= ${NO_WBITWISE_INSTEAD_OF_LOGICAL}
CFLAGS.zstd.c= -U__BMI__ -fno-tree-vectorize ${NO_WBITWISE_INSTEAD_OF_LOGICAL}
CFLAGS.zstd_common.c= -U__BMI__ -fno-tree-vectorize ${NO_WBITWISE_INSTEAD_OF_LOGICAL}
CFLAGS.zstd_compress.c= -U__BMI__ -fno-tree-vectorize ${NO_WBITWISE_INSTEAD_OF_LOGICAL}
CFLAGS.zstd_compress_literals.c= -U__BMI__ -fno-tree-vectorize ${NO_WBITWISE_INSTEAD_OF_LOGICAL}
CFLAGS.zstd_compress_sequences.c= -U__BMI__ -fno-tree-vectorize ${NO_WBITWISE_INSTEAD_OF_LOGICAL}
CFLAGS.zstd_compress_superblock.c= -U__BMI__ -fno-tree-vectorize ${NO_WBITWISE_INSTEAD_OF_LOGICAL} ${NO_WUNUSED_BUT_SET_VARIABLE}
CFLAGS.zstd_ddict.c= -U__BMI__ -fno-tree-vectorize ${NO_WBITWISE_INSTEAD_OF_LOGICAL}
CFLAGS.zstd_decompress.c= -U__BMI__ -fno-tree-vectorize ${NO_WBITWISE_INSTEAD_OF_LOGICAL}
CFLAGS.zstd_decompress_block.c= -U__BMI__ -fno-tree-vectorize ${NO_WBITWISE_INSTEAD_OF_LOGICAL}
CFLAGS.zstd_double_fast.c= -U__BMI__ -fno-tree-vectorize ${NO_WBITWISE_INSTEAD_OF_LOGICAL}
CFLAGS.zstd_fast.c= -U__BMI__ -fno-tree-vectorize ${NO_WBITWISE_INSTEAD_OF_LOGICAL}
CFLAGS.zstd_lazy.c= -U__BMI__ -fno-tree-vectorize ${NO_WBITWISE_INSTEAD_OF_LOGICAL}
CFLAGS.zstd_ldm.c= -U__BMI__ -fno-tree-vectorize ${NO_WBITWISE_INSTEAD_OF_LOGICAL}
CFLAGS.zstd_opt.c= -U__BMI__ -fno-tree-vectorize ${NO_WBITWISE_INSTEAD_OF_LOGICAL}
.if ${MACHINE_CPUARCH} == "aarch64"
__ZFS_ZSTD_AARCH64_FLAGS= -include ${SRCDIR}/zstd/include/aarch64_compat.h
CFLAGS.zstd.c+= ${__ZFS_ZSTD_AARCH64_FLAGS}
CFLAGS.entropy_common.c+= ${__ZFS_ZSTD_AARCH64_FLAGS}
CFLAGS.error_private.c+= ${__ZFS_ZSTD_AARCH64_FLAGS}
CFLAGS.fse_compress.c+= ${__ZFS_ZSTD_AARCH64_FLAGS}
CFLAGS.fse_decompress.c+= ${__ZFS_ZSTD_AARCH64_FLAGS}
CFLAGS.hist.c+= ${__ZFS_ZSTD_AARCH64_FLAGS}
CFLAGS.huf_compress.c+= ${__ZFS_ZSTD_AARCH64_FLAGS}
CFLAGS.huf_decompress.c+= ${__ZFS_ZSTD_AARCH64_FLAGS}
CFLAGS.pool.c+= ${__ZFS_ZSTD_AARCH64_FLAGS}
CFLAGS.xxhash.c+= ${__ZFS_ZSTD_AARCH64_FLAGS}
CFLAGS.zstd_common.c+= ${__ZFS_ZSTD_AARCH64_FLAGS}
CFLAGS.zstd_compress.c+= ${__ZFS_ZSTD_AARCH64_FLAGS}
CFLAGS.zstd_compress_literals.c+= ${__ZFS_ZSTD_AARCH64_FLAGS}
CFLAGS.zstd_compress_sequences.c+= ${__ZFS_ZSTD_AARCH64_FLAGS}
CFLAGS.zstd_compress_superblock.c+= ${__ZFS_ZSTD_AARCH64_FLAGS}
CFLAGS.zstd_ddict.c+= ${__ZFS_ZSTD_AARCH64_FLAGS}
CFLAGS.zstd_decompress.c+= ${__ZFS_ZSTD_AARCH64_FLAGS}
CFLAGS.zstd_decompress_block.c+= ${__ZFS_ZSTD_AARCH64_FLAGS}
CFLAGS.zstd_double_fast.c+= ${__ZFS_ZSTD_AARCH64_FLAGS}
CFLAGS.zstd_fast.c+= ${__ZFS_ZSTD_AARCH64_FLAGS}
CFLAGS.zstd_lazy.c+= ${__ZFS_ZSTD_AARCH64_FLAGS}
CFLAGS.zstd_ldm.c+= ${__ZFS_ZSTD_AARCH64_FLAGS}
CFLAGS.zstd_opt.c+= ${__ZFS_ZSTD_AARCH64_FLAGS}
sha256-armv8.o: sha256-armv8.S
${CC} -c ${CFLAGS:N-mgeneral-regs-only} ${WERROR} ${.IMPSRC} \
-o ${.TARGET}
${CTFCONVERT_CMD}
sha512-armv8.o: sha512-armv8.S
${CC} -c ${CFLAGS:N-mgeneral-regs-only} ${WERROR} ${.IMPSRC} \
-o ${.TARGET}
${CTFCONVERT_CMD}
b3_aarch64_sse2.o: b3_aarch64_sse2.S
${CC} -c ${CFLAGS:N-mgeneral-regs-only} ${WERROR} ${.IMPSRC} \
-o ${.TARGET}
${CTFCONVERT_CMD}
b3_aarch64_sse41.o: b3_aarch64_sse41.S
${CC} -c ${CFLAGS:N-mgeneral-regs-only} ${WERROR} ${.IMPSRC} \
-o ${.TARGET}
${CTFCONVERT_CMD}
.endif
Reference in New Issue View Git Blame Copy Permalink