>From 5bb63d463c17b68b022d5375b3e3a4ac643e086c Mon Sep 17 00:00:00 2001 From: Abhijit Menon-Sen Date: Tue, 30 Dec 2014 12:55:53 +0530 Subject: Use the SSE4.2 CRC instructions where available We execute cpuid at startup to determine if the processor supports SSE4.2 instructions, and use the crc32* instructions instead of the default slice-by-8 code (both produce identical results). Reviewers: Andres Freund, Heikki Linnakangas Author: Abhijit Menon-Sen --- configure | 2 +- configure.in | 2 +- src/backend/main/main.c | 7 ++++ src/include/pg_config.h.in | 3 ++ src/include/pg_config.h.win32 | 3 ++ src/include/utils/pg_crc.h | 3 +- src/port/pg_crc.c | 97 ++++++++++++++++++++++++++++++++++++++++++- 7 files changed, 112 insertions(+), 5 deletions(-) diff --git a/configure b/configure index 27092ec..f76cf4b 100755 --- a/configure +++ b/configure @@ -9195,7 +9195,7 @@ fi done -for ac_header in atomic.h crypt.h dld.h fp_class.h getopt.h ieeefp.h ifaddrs.h langinfo.h mbarrier.h poll.h pwd.h sys/ioctl.h sys/ipc.h sys/poll.h sys/pstat.h sys/resource.h sys/select.h sys/sem.h sys/shm.h sys/socket.h sys/sockio.h sys/tas.h sys/time.h sys/un.h termios.h ucred.h utime.h wchar.h wctype.h +for ac_header in atomic.h cpuid.h crypt.h dld.h fp_class.h getopt.h ieeefp.h ifaddrs.h langinfo.h mbarrier.h poll.h pwd.h sys/ioctl.h sys/ipc.h sys/poll.h sys/pstat.h sys/resource.h sys/select.h sys/sem.h sys/shm.h sys/socket.h sys/sockio.h sys/tas.h sys/time.h sys/un.h termios.h ucred.h utime.h wchar.h wctype.h do : as_ac_Header=`$as_echo "ac_cv_header_$ac_header" | $as_tr_sh` ac_fn_c_check_header_mongrel "$LINENO" "$ac_header" "$as_ac_Header" "$ac_includes_default" diff --git a/configure.in b/configure.in index 0206836..6005edc 100644 --- a/configure.in +++ b/configure.in @@ -1023,7 +1023,7 @@ AC_SUBST(UUID_LIBS) ## dnl sys/socket.h is required by AC_FUNC_ACCEPT_ARGTYPES -AC_CHECK_HEADERS([atomic.h crypt.h dld.h fp_class.h getopt.h ieeefp.h ifaddrs.h langinfo.h mbarrier.h poll.h pwd.h sys/ioctl.h sys/ipc.h sys/poll.h sys/pstat.h sys/resource.h sys/select.h sys/sem.h sys/shm.h sys/socket.h sys/sockio.h sys/tas.h sys/time.h sys/un.h termios.h ucred.h utime.h wchar.h wctype.h]) +AC_CHECK_HEADERS([atomic.h cpuid.h crypt.h dld.h fp_class.h getopt.h ieeefp.h ifaddrs.h langinfo.h mbarrier.h poll.h pwd.h sys/ioctl.h sys/ipc.h sys/poll.h sys/pstat.h sys/resource.h sys/select.h sys/sem.h sys/shm.h sys/socket.h sys/sockio.h sys/tas.h sys/time.h sys/un.h termios.h ucred.h utime.h wchar.h wctype.h]) # On BSD, test for net/if.h will fail unless sys/socket.h # is included first. diff --git a/src/backend/main/main.c b/src/backend/main/main.c index 582198f..153a97d 100644 --- a/src/backend/main/main.c +++ b/src/backend/main/main.c @@ -37,6 +37,7 @@ #include "utils/memutils.h" #include "utils/pg_locale.h" #include "utils/ps_status.h" +#include "utils/pg_crc.h" const char *progname; @@ -77,6 +78,12 @@ main(int argc, char *argv[]) argv = save_ps_display_args(argc, argv); /* + * Select the fastest available CRC32 implementation for the + * platform. + */ + pg_choose_crc_impl(); + + /* * If supported on the current platform, set up a handler to be called if * the backend/postmaster crashes with a fatal signal or exception. */ diff --git a/src/include/pg_config.h.in b/src/include/pg_config.h.in index 7962757..e12c4c9 100644 --- a/src/include/pg_config.h.in +++ b/src/include/pg_config.h.in @@ -99,6 +99,9 @@ /* Define to 1 if you have the `class' function. */ #undef HAVE_CLASS +/* Define to 1 if you have the header file. */ +#undef HAVE_CPUID_H + /* Define to 1 if you have the header file. */ #undef HAVE_CRTDEFS_H diff --git a/src/include/pg_config.h.win32 b/src/include/pg_config.h.win32 index 18da922..e2a5e21 100644 --- a/src/include/pg_config.h.win32 +++ b/src/include/pg_config.h.win32 @@ -78,6 +78,9 @@ /* Define to 1 if you have the `class' function. */ /* #undef HAVE_CLASS */ +/* Define to 1 if you have the header file. */ +/* #undef HAVE_CPUID_H */ + /* Define to 1 if you have the `crypt' function. */ /* #undef HAVE_CRYPT */ diff --git a/src/include/utils/pg_crc.h b/src/include/utils/pg_crc.h index 4d35601..e5d9668 100644 --- a/src/include/utils/pg_crc.h +++ b/src/include/utils/pg_crc.h @@ -42,7 +42,8 @@ typedef uint32 pg_crc32; extern uint32 bswap32(uint32 x); -extern pg_crc32 pg_comp_crc32c(pg_crc32 crc, const void *data, size_t len); +extern void pg_choose_crc_impl(void); +extern pg_crc32 (*pg_comp_crc32c)(pg_crc32 crc, const void *data, size_t len); /* * CRC calculation using the CRC-32C (Castagnoli) polynomial. diff --git a/src/port/pg_crc.c b/src/port/pg_crc.c index b1bdd1e..e57945b 100644 --- a/src/port/pg_crc.c +++ b/src/port/pg_crc.c @@ -20,6 +20,13 @@ #include "utils/pg_crc.h" #include "utils/pg_crc_tables.h" +#if defined(HAVE_CPUID_H) +#include +#elif defined(_MSC_VER) +#include +#include +#endif + #ifndef WORDS_BIGENDIAN #define CRC8(x) pg_crc32c_table[0][(crc ^ (x)) & 0xFF] ^ (crc >> 8) #else @@ -51,8 +58,8 @@ bswap32(uint32 x) * pp. 1550-1560, November 2008, doi:10.1109/TC.2008.85 */ -pg_crc32 -pg_comp_crc32c(pg_crc32 crc, const void *data, size_t len) +static pg_crc32 +pg_comp_crc32c_sb8(pg_crc32 crc, const void *data, size_t len) { const unsigned char *p = data; const uint32 *p8; @@ -120,3 +127,89 @@ pg_comp_crc32c(pg_crc32 crc, const void *data, size_t len) return crc; } + +static inline pg_crc32 +pg_asm_crc32b(pg_crc32 crc, unsigned char data) +{ +#if defined(__GNUC__) && defined(__x86_64__) + __asm__ ("crc32b %[data], %[crc]\n" : [crc] "+r" (crc) : [data] "rm" (data)); + return crc; +#elif defined(_MSC_VER) + return _mm_crc32_u8(crc, data); +#else + /* Can't generate crc32b, but keep the compiler quiet. */ + return 0; +#endif +} + +static inline pg_crc32 +pg_asm_crc32q(uint64 crc, unsigned long long data) +{ +#if defined(__GNUC__) && defined(__x86_64__) + __asm__ ("crc32q %[data], %[crc]\n" : [crc] "+r" (crc) : [data] "rm" (data)); + return crc; +#elif defined(_MSC_VER) + return _mm_crc32_u64(crc, data); +#else + /* Can't generate crc32q, but keep the compiler quiet. */ + return 0; +#endif +} + +/* + * This function computes a CRC using the SSE4.2 CRC32B and CRC32Q + * instructions. + */ + +static pg_crc32 +pg_comp_crc32c_sse(pg_crc32 crc, const void *data, size_t len) +{ + const unsigned char *p = data; + const uint64 *p8; + + /* + * Process eight bytes of data at a time. + */ + + p8 = (const uint64 *) p; + while (len >= 8) + { + crc = pg_asm_crc32q(crc, *p8++); + len -= 8; + } + + /* + * Handle any remaining bytes one at a time. + */ + + p = (const unsigned char *) p8; + while (len > 0) + { + crc = pg_asm_crc32b(crc, *p++); + len--; + } + + return crc; +} + +/* + * If (we can tell that) the CPU supports SSE4.2, we can use the CRC + * instructions, otherwise we fall back to slice-by-8 in software. + */ + +void +pg_choose_crc_impl(void) +{ + unsigned int exx[4] = {0, 0, 0, 0}; + +#if defined(__GNUC__) && defined(HAVE_CPUID_H) + __get_cpuid(1, &exx[0], &exx[1], &exx[2], &exx[3]); +#elif defined(_MSC_VER) + __cpuid(exx, 1); +#endif + + if (exx[2] & (1 << 20)) + pg_comp_crc32c = pg_comp_crc32c_sse; +} + +pg_crc32 (*pg_comp_crc32c)(pg_crc32 crc, const void *data, size_t len) = pg_comp_crc32c_sb8; -- 1.9.1