On Mon, Apr 27, 2026 at 02:43:20PM -0500, Nathan Bossart wrote: > On Sun, Apr 26, 2026 at 09:31:37PM -0500, Nathan Bossart wrote: >> On Sun, Apr 26, 2026 at 10:18:33PM -0400, Tom Lane wrote: >>> I previously suggested that we should teach the build systems not >>> to build the foo_x86 and foo_aarch64 modules when not on those >>> architectures. That was shot down for reasons that made no great >>> amount of sense to me, but I think it'd be fairly easy and clean. >> >> Yeah, AFAICT that's our best bet. Will write it up in the morning. > > Here's what I have so far. Two notes: > > * Since pg_popcount_aarch64.c only builds symbols when USE_NEON is defined, > I needed to teach the build code about that #define. So, this patch > effectively moves USE_NEON and USE_SSE2 from c.h to pg_config.h, which > happens to be the first #include within c.h. The reason that I bring this > up is because back-patching it seems a little scary, although I don't see > any concrete reasons it would be unsafe. > > * pg_cpu_x86.c has a similar problem, which I haven't fixed yet. > Presumably a similar approach will work there. I've added John Naylor to > this thread for his thoughts.
Sorry, I noticed I was using USE_SSE2 for choosing whether to build pg_popcount_x86.c, but the code in that file is actually surrounded by HAVE_X86_64_POPCNTQ. New patch attached. -- nathan
>From c2060ada9078b2dd22893c678c77ebaa2962fb05 Mon Sep 17 00:00:00 2001 From: Nathan Bossart <[email protected]> Date: Mon, 27 Apr 2026 14:32:12 -0500 Subject: [PATCH v2 1/1] avoid ranlib warnings on macOS --- configure | 83 ++++++++++++++++++++++++++++++++++++-- configure.ac | 44 ++++++++++++++++++-- meson.build | 48 ++++++++++++++++++++++ src/Makefile.global.in | 4 ++ src/include/c.h | 19 --------- src/include/pg_config.h.in | 6 +++ src/makefiles/meson.build | 2 +- src/port/Makefile | 10 ++++- src/port/meson.build | 4 +- 9 files changed, 189 insertions(+), 31 deletions(-) diff --git a/configure b/configure index f66c1054a7a..79c7513782b 100755 --- a/configure +++ b/configure @@ -647,7 +647,9 @@ MSGFMT_FLAGS MSGFMT PG_CRC32C_OBJS CFLAGS_CRC +USE_NEON LIBOBJS +HAVE_X86_64_POPCNTQ OPENSSL ZSTD LZ4 @@ -15469,6 +15471,8 @@ _ACEOF fi +HAVE_X86_64_POPCNTQ=0 + case $host_cpu in x86_64) # On x86_64, check if we can compile a popcntq instruction @@ -15490,15 +15494,13 @@ long long x = 1; long long r; } _ACEOF if ac_fn_c_try_compile "$LINENO"; then : - pgac_cv_have_x86_64_popcntq=yes -else - pgac_cv_have_x86_64_popcntq=no + HAVE_X86_64_POPCNTQ=1 fi rm -f core conftest.err conftest.$ac_objext conftest.$ac_ext fi { $as_echo "$as_me:${as_lineno-$LINENO}: result: $pgac_cv_have_x86_64_popcntq" >&5 $as_echo "$pgac_cv_have_x86_64_popcntq" >&6; } - if test x"$pgac_cv_have_x86_64_popcntq" = xyes ; then + if test x"$HAVE_X86_64_POPCNTQ" = x"1" ; then $as_echo "#define HAVE_X86_64_POPCNTQ 1" >>confdefs.h @@ -15548,6 +15550,8 @@ $as_echo "#define HAVE_I_CONSTRAINT__BUILTIN_CONSTANT_P 1" >>confdefs.h ;; esac + + # Check largefile support. You might think this is a system service not a # compiler characteristic, but you'd be wrong. We must check this before # probing existence of related functions such as fseeko, since the largefile @@ -17906,6 +17910,77 @@ $as_echo "#define USE_AVX2_WITH_RUNTIME_CHECK 1" >>confdefs.h fi fi +# Check for the availability of NEON intrinsics. +# +# We use the Neon instructions if the compiler provides access +# to them (as indicated by __ARM_NEON) and we are on aarch64. +# While Neon support is technically optional for aarch64, it +# appears that all available 64-bit hardware does have it. +# Neon exists in some 32-bit hardware too, but we could not +# realistically use it there without a run-time check, which +# seems not worth the trouble for now. +# +cat confdefs.h - <<_ACEOF >conftest.$ac_ext +/* end confdefs.h. */ + +int +main () +{ + +#if !defined(__aarch64__) || !defined(__ARM_NEON) +#error compiler does not advertise NEON intrinsics +#endif + + ; + return 0; +} +_ACEOF +if ac_fn_c_try_compile "$LINENO"; then : + USE_NEON=1 +else + USE_NEON=0 +fi +rm -f core conftest.err conftest.$ac_objext conftest.$ac_ext +if test x"$USE_NEON" == x"1"; then + +$as_echo "#define USE_NEON 1" >>confdefs.h + +fi + + +# Check for SSE2 intrinsics +# +# SSE2 instructions are part of the spec for the 64-bit x86 +# ISA. We assume that compilers targeting this architecture +# understand SSE2 intrinsics. +# +cat confdefs.h - <<_ACEOF >conftest.$ac_ext +/* end confdefs.h. */ + +int +main () +{ + +#if !defined(__x86_64__) && !defined(_M_AMD64) +#error compiler does not advertise SSE2 intrinsics +#endif + + ; + return 0; +} +_ACEOF +if ac_fn_c_try_compile "$LINENO"; then : + USE_SSE2=1 +else + USE_SSE2=0 +fi +rm -f core conftest.err conftest.$ac_objext conftest.$ac_ext +if test x"$USE_SSE2" == x"1"; then + +$as_echo "#define USE_SSE2 1" >>confdefs.h + +fi + # Check for XSAVE intrinsics # { $as_echo "$as_me:${as_lineno-$LINENO}: checking for _xgetbv" >&5 diff --git a/configure.ac b/configure.ac index 8d176bd3468..c77b1819f78 100644 --- a/configure.ac +++ b/configure.ac @@ -1756,6 +1756,8 @@ AC_CHECK_TYPES([struct option], [], [], #include <getopt.h> #endif]) +HAVE_X86_64_POPCNTQ=0 + case $host_cpu in x86_64) # On x86_64, check if we can compile a popcntq instruction @@ -1764,9 +1766,8 @@ case $host_cpu in [AC_COMPILE_IFELSE([AC_LANG_PROGRAM([], [long long x = 1; long long r; __asm__ __volatile__ (" popcntq %1,%0\n" : "=q"(r) : "rm"(x));])], - [pgac_cv_have_x86_64_popcntq=yes], - [pgac_cv_have_x86_64_popcntq=no])]) - if test x"$pgac_cv_have_x86_64_popcntq" = xyes ; then + [HAVE_X86_64_POPCNTQ=1])]) + if test x"$HAVE_X86_64_POPCNTQ" = x"1" ; then AC_DEFINE(HAVE_X86_64_POPCNTQ, 1, [Define to 1 if the assembler supports X86_64's POPCNTQ instruction.]) fi ;; @@ -1794,6 +1795,8 @@ case $host_cpu in ;; esac +AC_SUBST(HAVE_X86_64_POPCNTQ) + # Check largefile support. You might think this is a system service not a # compiler characteristic, but you'd be wrong. We must check this before # probing existence of related functions such as fseeko, since the largefile @@ -2162,6 +2165,41 @@ if test x"$host_cpu" = x"x86_64"; then fi fi +# Check for the availability of NEON intrinsics. +# +# We use the Neon instructions if the compiler provides access +# to them (as indicated by __ARM_NEON) and we are on aarch64. +# While Neon support is technically optional for aarch64, it +# appears that all available 64-bit hardware does have it. +# Neon exists in some 32-bit hardware too, but we could not +# realistically use it there without a run-time check, which +# seems not worth the trouble for now. +# +AC_COMPILE_IFELSE([AC_LANG_PROGRAM([], [ +#if !defined(__aarch64__) || !defined(__ARM_NEON) +#error compiler does not advertise NEON intrinsics +#endif +])], [USE_NEON=1], [USE_NEON=0]) +if test x"$USE_NEON" == x"1"; then + AC_DEFINE(USE_NEON, 1, [Define to 1 if compiler understands AArch64 NEON intrinsics.]) +fi +AC_SUBST(USE_NEON) + +# Check for SSE2 intrinsics +# +# SSE2 instructions are part of the spec for the 64-bit x86 +# ISA. We assume that compilers targeting this architecture +# understand SSE2 intrinsics. +# +AC_COMPILE_IFELSE([AC_LANG_PROGRAM([], [ +#if !defined(__x86_64__) && !defined(_M_AMD64) +#error compiler does not advertise SSE2 intrinsics +#endif +])], [USE_SSE2=1], [USE_SSE2=0]) +if test x"$USE_SSE2" == x"1"; then + AC_DEFINE(USE_SSE2, 1, [Define to 1 if compiler understands x86_64 SSE2 intrinsics.]) +fi + # Check for XSAVE intrinsics # PGAC_XSAVE_INTRINSICS() diff --git a/meson.build b/meson.build index 20b887f1a1b..6876999b3ee 100644 --- a/meson.build +++ b/meson.build @@ -2468,6 +2468,54 @@ int main(void) endforeach +############################################################### +# Check for the availability of NEON intrinsics. +# +# We use the Neon instructions if the compiler provides access +# to them (as indicated by __ARM_NEON) and we are on aarch64. +# While Neon support is technically optional for aarch64, it +# appears that all available 64-bit hardware does have it. +# Neon exists in some 32-bit hardware too, but we could not +# realistically use it there without a run-time check, which +# seems not worth the trouble for now. +############################################################### + +if host_cpu == 'aarch64' + + neon_test = ''' +#if !defined(__aarch64__) || !defined(__ARM_NEON) +#error "Compiler does not advertise NEON intrinsics" +#endif +''' + + if cc.compiles(neon_test, name: 'AArch64 NEON intrinsics') + cdata.set('USE_NEON', 1) + endif +endif + + +############################################################### +# Check for the availability of SSE2 intrinsics. +# +# SSE2 instructions are part of the spec for the 64-bit x86 +# ISA. We assume that compilers targeting this architecture +# understand SSE2 intrinsics. +############################################################### + +if host_cpu == 'x86_64' + + sse2_test = ''' +#if !defined(__x86_64__) && !defined(_M_AMD64) +#error "Compiler does not advertise SSE2 intrinsics" +#endif +''' + + if cc.compiles(sse2_test, name: 'x86_64 SSE2 intrinsics') + cdata.set('USE_SSE2', 1) + endif +endif + + ############################################################### # Check for the availability of XSAVE intrinsics. ############################################################### diff --git a/src/Makefile.global.in b/src/Makefile.global.in index cef1ad7f87d..c7d797e452a 100644 --- a/src/Makefile.global.in +++ b/src/Makefile.global.in @@ -588,6 +588,10 @@ endif show_dl_suffix: @echo $(DLSUFFIX) +# AArch64/NEON and x86_64/POPCNTQ intrinsics +USE_NEON = @USE_NEON@ +HAVE_X86_64_POPCNTQ = @HAVE_X86_64_POPCNTQ@ + ########################################################################## # diff --git a/src/include/c.h b/src/include/c.h index 97ed8c63f5e..aa43d302057 100644 --- a/src/include/c.h +++ b/src/include/c.h @@ -1333,25 +1333,6 @@ typedef struct PGAlignedXLogBlock PGAlignedXLogBlock; (underlying_type) (expr)) #endif -/* - * SSE2 instructions are part of the spec for the 64-bit x86 ISA. We assume - * that compilers targeting this architecture understand SSE2 intrinsics. - */ -#if (defined(__x86_64__) || defined(_M_AMD64)) -#define USE_SSE2 - -/* - * We use the Neon instructions if the compiler provides access to them (as - * indicated by __ARM_NEON) and we are on aarch64. While Neon support is - * technically optional for aarch64, it appears that all available 64-bit - * hardware does have it. Neon exists in some 32-bit hardware too, but we - * could not realistically use it there without a run-time check, which seems - * not worth the trouble for now. - */ -#elif defined(__aarch64__) && defined(__ARM_NEON) -#define USE_NEON -#endif - /* ---------------------------------------------------------------- * Section 9: system-specific hacks * diff --git a/src/include/pg_config.h.in b/src/include/pg_config.h.in index 4f8113c144b..3e4aa1e1e8f 100644 --- a/src/include/pg_config.h.in +++ b/src/include/pg_config.h.in @@ -729,6 +729,9 @@ /* Define to select named POSIX semaphores. */ #undef USE_NAMED_POSIX_SEMAPHORES +/* Define to 1 if compiler understands AArch64 NEON intrinsics. */ +#undef USE_NEON + /* Define to 1 to build with OpenSSL support. (--with-ssl=openssl) */ #undef USE_OPENSSL @@ -741,6 +744,9 @@ /* Define to 1 to use software CRC-32C implementation (slicing-by-8). */ #undef USE_SLICING_BY_8_CRC32C +/* Define to 1 if compiler understands x86_64 SSE2 intrinsics. */ +#undef USE_SSE2 + /* Define to 1 use Intel SSE 4.2 CRC instructions. */ #undef USE_SSE42_CRC32C diff --git a/src/makefiles/meson.build b/src/makefiles/meson.build index 2401025d1cd..a8a041f8b42 100644 --- a/src/makefiles/meson.build +++ b/src/makefiles/meson.build @@ -182,7 +182,7 @@ pgxs_empty = [ 'WANTED_LANGUAGES', # Not needed because we don't build the server / PLs with the generated makefile - 'LIBOBJS', 'PG_CRC32C_OBJS', + 'LIBOBJS', 'PG_CRC32C_OBJS', 'USE_NEON', 'HAVE_X86_64_POPCNTQ', 'PG_TEST_EXTRA', 'DTRACEFLAGS', # only server has dtrace probes diff --git a/src/port/Makefile b/src/port/Makefile index 7e9b5877652..c30c6bfe6c0 100644 --- a/src/port/Makefile +++ b/src/port/Makefile @@ -48,8 +48,6 @@ OBJS = \ pg_getopt_ctx.o \ pg_localeconv_r.o \ pg_numa.o \ - pg_popcount_aarch64.o \ - pg_popcount_x86.o \ pg_strong_random.o \ pgcheckdir.o \ pgmkdirp.o \ @@ -64,6 +62,14 @@ OBJS = \ strerror.o \ tar.o +ifeq ($(USE_NEON),1) +OBJS += pg_popcount_aarch64.o +endif + +ifeq ($(HAVE_X86_64_POPCNTQ),1) +OBJS += pg_popcount_x86.o +endif + # libpgport.a, libpgport_shlib.a, and libpgport_srv.a contain the same files # foo.o, foo_shlib.o, and foo_srv.o are all built from foo.c OBJS_SHLIB = $(OBJS:%.o=%_shlib.o) diff --git a/src/port/meson.build b/src/port/meson.build index 922b3f64676..b5ff4d0c31f 100644 --- a/src/port/meson.build +++ b/src/port/meson.build @@ -11,8 +11,6 @@ pgport_sources = [ 'pg_getopt_ctx.c', 'pg_localeconv_r.c', 'pg_numa.c', - 'pg_popcount_aarch64.c', - 'pg_popcount_x86.c', 'pg_strong_random.c', 'pgcheckdir.c', 'pgmkdirp.c', @@ -89,6 +87,7 @@ replace_funcs_pos = [ ['pg_crc32c_sse42', 'USE_SSE42_CRC32C'], ['pg_crc32c_sse42', 'USE_SSE42_CRC32C_WITH_RUNTIME_CHECK'], ['pg_crc32c_sb8', 'USE_SSE42_CRC32C_WITH_RUNTIME_CHECK'], + ['pg_popcount_x86', 'HAVE_X86_64_POPCNTQ'], # arm / aarch64 ['pg_crc32c_armv8', 'USE_ARMV8_CRC32C'], @@ -96,6 +95,7 @@ replace_funcs_pos = [ ['pg_crc32c_armv8_choose', 'USE_ARMV8_CRC32C'], ['pg_crc32c_armv8_choose', 'USE_ARMV8_CRC32C_WITH_RUNTIME_CHECK'], ['pg_crc32c_sb8', 'USE_ARMV8_CRC32C_WITH_RUNTIME_CHECK'], + ['pg_popcount_aarch64', 'USE_NEON'], # loongarch ['pg_crc32c_loongarch', 'USE_LOONGARCH_CRC32C'], -- 2.50.1 (Apple Git-155)
