hlrs-software-stack/spack
7
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

zlib-ng: build system patches (#39559)

Browse source
This commit is contained in:
Harmen Stoppels 2023-08-24 12:13:22 +02:00 committed by GitHub
parent 36852fe348
commit 46285d9725
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
4 changed files with 871 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

6
var/spack/repos/builtin/packages/zlib-ng/package.py
View file

@ -31,7 +31,11 @@ class ZlibNg(AutotoolsPackage, CMakePackage):
build_system("autotools", "cmake", default="autotools") build_system("autotools", "cmake", default="autotools")
# rpath shenanigans, see https://github.com/zlib-ng/zlib-ng/pull/1546 # rpath shenanigans, see https://github.com/zlib-ng/zlib-ng/pull/1546
patch("pr-1546.patch", when="@2.1.3 platform=darwin") with when("@2.1.3"):
patch("pr-1546.patch", when="platform=darwin")
patch("pr-1542.patch") # fix sse4.2 detection
patch("pr-1561.patch", when="build_system=autotools") # drop bash dependency
patch("pr-1562.patch") # improve intrinsics detection
with when("build_system=cmake"): with when("build_system=cmake"):
depends_on("cmake@3.5.1:", type="build") depends_on("cmake@3.5.1:", type="build")

224
var/spack/repos/builtin/packages/zlib-ng/pr-1542.patch Normal file
View file

@ -0,0 +1,224 @@
From 8c5d5eca51d9e4cd9aa046dba8f939b3f4012256 Mon Sep 17 00:00:00 2001
From: Hans Kristian Rosbach <hk-git@circlestorm.org>
Date: Fri, 21 Jul 2023 13:43:15 +0200
Subject: [PATCH 1/3] Clean up SSE4.2 support, and no longer use asm fallback
or gcc builtin.
Defines changing meaning:
X86_SSE42 used to mean the compiler supports crc asm fallback.
X86_SSE42_CRC_INTRIN used to mean compiler supports SSE4.2 intrinsics.
X86_SSE42 now means compiler supports SSE4.2 intrinsics.
This therefore also fixes the adler32_sse42 checks, since those were depending
on SSE4.2 intrinsics but was mistakenly checking the X86_SSE42 define.
Now the X86_SSE42 define actually means what it appears to.
---
CMakeLists.txt | 5 +----
arch/x86/insert_string_sse42.c | 36 +++++----------------------------
cmake/detect-intrinsics.cmake | 23 +++------------------
configure | 37 ++++++++--------------------------
win32/Makefile.msc | 1 -
5 files changed, 17 insertions(+), 85 deletions(-)
diff --git a/CMakeLists.txt b/CMakeLists.txt
index 420a5c78..1e42239a 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -827,15 +827,12 @@ if(WITH_OPTIM)
endif()
if(WITH_SSE42)
check_sse42_intrinsics()
- if(HAVE_SSE42CRC_INLINE_ASM OR HAVE_SSE42CRC_INTRIN)
+ if(HAVE_SSE42_INTRIN)
add_definitions(-DX86_SSE42)
set(SSE42_SRCS ${ARCHDIR}/adler32_sse42.c ${ARCHDIR}/insert_string_sse42.c)
add_feature_info(SSE42_CRC 1 "Support SSE4.2 optimized CRC hash generation, using \"${SSE42FLAG}\"")
list(APPEND ZLIB_ARCH_SRCS ${SSE42_SRCS})
set_property(SOURCE ${SSE42_SRCS} PROPERTY COMPILE_FLAGS "${SSE42FLAG} ${NOLTOFLAG}")
- if(HAVE_SSE42CRC_INTRIN)
- add_definitions(-DX86_SSE42_CRC_INTRIN)
- endif()
else()
set(WITH_SSE42 OFF)
endif()
diff --git a/arch/x86/insert_string_sse42.c b/arch/x86/insert_string_sse42.c
index 565d92f9..ae092a7e 100644
--- a/arch/x86/insert_string_sse42.c
+++ b/arch/x86/insert_string_sse42.c
@@ -5,38 +5,13 @@
*
*/
+#ifdef X86_SSE42
#include "../../zbuild.h"
-#include <immintrin.h>
-#ifdef _MSC_VER
-# include <nmmintrin.h>
-#endif
+#include <nmmintrin.h>
#include "../../deflate.h"
-#ifdef X86_SSE42_CRC_INTRIN
-# ifdef _MSC_VER
-# define HASH_CALC(s, h, val)\
- h = _mm_crc32_u32(h, val)
-# else
-# define HASH_CALC(s, h, val)\
- h = __builtin_ia32_crc32si(h, val)
-# endif
-#else
-# ifdef _MSC_VER
-# define HASH_CALC(s, h, val) {\
- __asm mov edx, h\
- __asm mov eax, val\
- __asm crc32 eax, edx\
- __asm mov h, eax\
- }
-# else
-# define HASH_CALC(s, h, val) \
- __asm__ __volatile__ (\
- "crc32 %1,%0\n\t"\
- : "+r" (h)\
- : "r" (val)\
- );
-# endif
-#endif
+#define HASH_CALC(s, h, val)\
+ h = _mm_crc32_u32(h, val)
#define HASH_CALC_VAR h
#define HASH_CALC_VAR_INIT uint32_t h = 0
@@ -45,6 +20,5 @@
#define INSERT_STRING insert_string_sse42
#define QUICK_INSERT_STRING quick_insert_string_sse42
-#ifdef X86_SSE42
-# include "../../insert_string_tpl.h"
+#include "../../insert_string_tpl.h"
#endif
diff --git a/cmake/detect-intrinsics.cmake b/cmake/detect-intrinsics.cmake
index 9cbc5908..52c54dc8 100644
--- a/cmake/detect-intrinsics.cmake
+++ b/cmake/detect-intrinsics.cmake
@@ -481,35 +481,18 @@ macro(check_sse42_intrinsics)
set(SSE42FLAG "-msse4.2")
endif()
endif()
- # Check whether compiler supports SSE4.2 CRC inline asm
+ # Check whether compiler supports SSE4.2 intrinsics
set(CMAKE_REQUIRED_FLAGS "${SSE42FLAG} ${NATIVEFLAG}")
check_c_source_compile_or_run(
- "int main(void) {
- unsigned val = 0, h = 0;
- #if defined(_MSC_VER)
- { __asm mov edx, h __asm mov eax, val __asm crc32 eax, edx __asm mov h, eax }
- #else
- __asm__ __volatile__ ( \"crc32 %1,%0\" : \"+r\" (h) : \"r\" (val) );
- #endif
- return (int)h;
- }"
- HAVE_SSE42CRC_INLINE_ASM
- )
- # Check whether compiler supports SSE4.2 CRC intrinsics
- check_c_source_compile_or_run(
- "#include <immintrin.h>
+ "#include <nmmintrin.h>
int main(void) {
unsigned crc = 0;
char c = 'c';
- #if defined(_MSC_VER)
crc = _mm_crc32_u32(crc, c);
- #else
- crc = __builtin_ia32_crc32qi(crc, c);
- #endif
(void)crc;
return 0;
}"
- HAVE_SSE42CRC_INTRIN
+ HAVE_SSE42_INTRIN
)
set(CMAKE_REQUIRED_FLAGS)
endmacro()
diff --git a/configure b/configure
index 8714590e..6b4e7fff 100755
--- a/configure
+++ b/configure
@@ -1431,38 +1431,23 @@ EOF
}
check_sse42_intrinsics() {
- # Check whether compiler supports SSE4.2 CRC inline asm
- cat > $test.c << EOF
-int main(void) {
- unsigned val = 0, h = 0;
- __asm__ __volatile__ ( "crc32 %1,%0" : "+r" (h) : "r" (val) );
- return (int) h;
-}
-EOF
- if try ${CC} ${CFLAGS} ${sse42flag} $test.c; then
- echo "Checking for SSE4.2 CRC inline assembly ... Yes." | tee -a configure.log
- HAVE_SSE42CRC_INLINE_ASM=1
- else
- echo "Checking for SSE4.2 CRC inline assembly ... No." | tee -a configure.log
- HAVE_SSE42CRC_INLINE_ASM=0
- fi
-
- # Check whether compiler supports SSE4.2 CRC intrinsics
+ # Check whether compiler supports SSE4.2 intrinsics
cat > $test.c << EOF
+#include <nmmintrin.h>
int main(void) {
unsigned crc = 0;
char c = 'c';
- crc = __builtin_ia32_crc32qi(crc, c);
+ crc = _mm_crc32_u32(crc, c);
(void)crc;
return 0;
}
EOF
if try ${CC} ${CFLAGS} ${sse42flag} $test.c; then
- echo "Checking for SSE4.2 CRC intrinsics ... Yes." | tee -a configure.log
- HAVE_SSE42CRC_INTRIN=1
+ echo "Checking for SSE4.2 intrinsics ... Yes." | tee -a configure.log
+ HAVE_SSE42_INTRIN=1
else
- echo "Checking for SSE4.2 CRC intrinsics ... No." | tee -a configure.log
- HAVE_SSE42CRC_INTRIN=0
+ echo "Checking for SSE4.2 intrinsics ... No." | tee -a configure.log
+ HAVE_SSE42_INTRIN=0
fi
}
@@ -1606,15 +1591,9 @@ case "${ARCH}" in
check_sse42_intrinsics
- if test ${HAVE_SSE42CRC_INTRIN} -eq 1 || test ${HAVE_SSE42CRC_INLINE_ASM} -eq 1; then
+ if test ${HAVE_SSE42_INTRIN} -eq 1; then
CFLAGS="${CFLAGS} -DX86_SSE42"
SFLAGS="${SFLAGS} -DX86_SSE42"
-
- if test ${HAVE_SSE42CRC_INTRIN} -eq 1; then
- CFLAGS="${CFLAGS} -DX86_SSE42_CRC_INTRIN"
- SFLAGS="${SFLAGS} -DX86_SSE42_CRC_INTRIN"
- fi
-
ARCH_STATIC_OBJS="${ARCH_STATIC_OBJS} adler32_sse42.o insert_string_sse42.o"
ARCH_SHARED_OBJS="${ARCH_SHARED_OBJS} adler32_sse42.lo insert_string_sse42.lo"
fi
diff --git a/win32/Makefile.msc b/win32/Makefile.msc
index 9ed26f28..3035072b 100644
--- a/win32/Makefile.msc
+++ b/win32/Makefile.msc
@@ -31,7 +31,6 @@ WFLAGS = \
-DX86_PCLMULQDQ_CRC \
-DX86_SSE2 \
-DX86_SSE42 \
- -DX86_SSE42_CRC_INTRIN \
-DX86_SSSE3 \
-DX86_AVX2
--
2.39.2

210
var/spack/repos/builtin/packages/zlib-ng/pr-1561.patch Normal file
View file

@ -0,0 +1,210 @@
From f6fb1d350a7b8210cc9c45ed502b3cc34e4dac32 Mon Sep 17 00:00:00 2001
From: Harmen Stoppels <me@harmenstoppels.nl>
Date: Mon, 21 Aug 2023 10:13:10 +0200
Subject: [PATCH 2/3] PR #1561
---
configure | 50 +++++++++++++++++++++++++-------------------------
1 file changed, 25 insertions(+), 25 deletions(-)
diff --git a/configure b/configure
index 6b4e7fff..fc78a135 100755
--- a/configure
+++ b/configure
@@ -1,4 +1,4 @@
-#!/usr/bin/env bash
+#!/bin/sh
# configure script for zlib.
#
# Normally configure builds both a static and a shared library.
@@ -230,13 +230,13 @@ test=ztest$$
# put arguments in log, also put test file in log if used in arguments
show()
{
- case "$*" in
+ case "$@" in
*$test.c*)
echo "=== $test.c ===" >> configure.log
cat $test.c >> configure.log
echo "===" >> configure.log;;
esac
- echo $* >> configure.log
+ echo "$@" >> configure.log
}
# check for gcc vs. cc and set compile and link flags based on the system identified by uname
@@ -246,7 +246,7 @@ int main() {return getchar();}
EOF
cc=${CC-${CROSS_PREFIX}gcc}
-echo -n "Checking for compiler... " | tee -a configure.log
+printf "Checking for compiler... " | tee -a configure.log
case "$cc" in
*gcc*) gcc=1 ;;
*clang*) gcc=1 ;;
@@ -401,7 +401,7 @@ if test "$gcc" -eq 1 && ($cc $CFLAGS -c $test.c) >> configure.log 2>&1; then
SFLAGS="${SFLAGS} -DNDEBUG"
fi
if test -z "$uname"; then
- uname=$((uname -s || echo unknown) 2>/dev/null)
+ uname=$( (uname -s || echo unknown) 2>/dev/null)
fi
case "$uname" in
Linux* | linux* | GNU | GNU/* | solaris*)
@@ -483,7 +483,7 @@ if test "$gcc" -eq 1 && ($cc $CFLAGS -c $test.c) >> configure.log 2>&1; then
fi
RC="${CROSS_PREFIX}windres"
RCFLAGS="-I ${BUILDDIR}"
- if [ "$CC" == "mingw32-gcc" ]; then
+ if [ "$CC" = "mingw32-gcc" ]; then
case $ARCH in
i386 | i486 | i586 | i686) RCFLAGS="${RCFLAGS} -F pe-i386";;
esac;
@@ -498,7 +498,7 @@ if test "$gcc" -eq 1 && ($cc $CFLAGS -c $test.c) >> configure.log 2>&1; then
HP-UX*)
LDSHARED=${LDSHARED-"$cc"}
LDSHAREDFLAGS="-shared"
- case $((uname -m || echo unknown) 2>/dev/null) in
+ case $( (uname -m || echo unknown) 2>/dev/null) in
ia64)
shared_ext='.so'
SHAREDLIB='${LIBNAME}.so' ;;
@@ -539,14 +539,14 @@ else
gcc=0
echo "$CC" | tee -a configure.log
if test -z "$uname"; then
- uname=$((uname -sr || echo unknown) 2>/dev/null)
+ uname=$( (uname -sr || echo unknown) 2>/dev/null)
fi
case "$uname" in
HP-UX*) SFLAGS=${CFLAGS-"-O +z"}
CFLAGS=${CFLAGS-"-O"}
LDSHARED=${LDSHARED-"ld"}
LDSHAREDFLAGS="-b"
- case $((uname -m || echo unknown) 2>/dev/null) in
+ case $( (uname -m || echo unknown) 2>/dev/null) in
ia64)
shared_ext='.so'
SHAREDLIB='${LIBNAME}.so' ;;
@@ -591,15 +591,15 @@ EOF
if ($CC -c $CFLAGS $test.c) 2>/dev/null; then
try()
{
- show $*
- test "$(\( $* \) 2>&1 | tee -a configure.log)" = ""
+ show "$@"
+ test "$( ("$@") 2>&1 | tee -a configure.log)" = ""
}
echo - using any output from compiler to indicate an error >> configure.log
else
try()
{
- show $*
- ( $* ) >> configure.log 2>&1
+ show "$@"
+ ( "$@" ) >> configure.log 2>&1
ret=$?
if test $ret -ne 0; then
echo "(exit code $ret)" >> configure.log
@@ -627,7 +627,7 @@ extern int getchar();
int hello() {return getchar();}
EOF
if test $shared -eq 1; then
- echo -n "Checking for shared library support... " | tee -a configure.log
+ printf "Checking for shared library support... " | tee -a configure.log
# we must test in two steps (cc then ld), required at least on SunOS 4.x
if try $CC -w -c $SFLAGS $test.c &&
try $LDSHARED $LDSHAREDFLAGS $LDFLAGS -o $test$shared_ext $test.o $LDSHAREDLIBC; then
@@ -784,7 +784,7 @@ fi
# Rename @ZLIB_SYMBOL_PREFIX@ to $symbol_prefix in gzread.c, zlib.h and zlib_name_mangling.h
sed < $SRCDIR/gzread.c.in "s/@ZLIB_SYMBOL_PREFIX@/$symbol_prefix/g" > gzread.c
sed < $SRCDIR/zlib${SUFFIX}.h.in "s/@ZLIB_SYMBOL_PREFIX@/$symbol_prefix/g" > zlib${SUFFIX}.h
-if [[ ! -z $symbol_prefix ]]; then
+if [ ! -z "$symbol_prefix" ]; then
sed < $SRCDIR/zlib_name_mangling${SUFFIX}.h.in "s/@ZLIB_SYMBOL_PREFIX@/$symbol_prefix/g" > zlib_name_mangling${SUFFIX}.h
else
# symbol_prefix is not set, copy the empty mangling header
@@ -814,7 +814,7 @@ fi
echo >> configure.log
# check for ptrdiff_t and save result in zconf.h
-echo -n "Checking for ptrdiff_t... " | tee -a configure.log
+printf "Checking for ptrdiff_t... " | tee -a configure.log
cat > $test.c <<EOF
#include <stddef.h>
int fun(ptrdiff_t *a) { (void)a; return 0; }
@@ -826,7 +826,7 @@ else
sed < zconf${SUFFIX}.h "/^#ifdef NEED_PTRDIFF_T.* may be/s/def NEED_PTRDIFF_T\(.*\) may be/ 1\1 was/" > zconf${SUFFIX}.temp.h
mv zconf${SUFFIX}.temp.h zconf${SUFFIX}.h
- echo -n "Checking for sizeof(void *)... " | tee -a configure.log
+ printf "Checking for sizeof(void *)... " | tee -a configure.log
cat > $test.c <<EOF
#include <stdint.h>
#define COMPILE_TIME_ASSERT(pred) struct s { int x: (pred) ? 1 : -1; }
@@ -864,7 +864,7 @@ if test $compat -eq 1; then
esac
fi
-if [[ ! -z $DEFFILE ]]; then
+if [ ! -z "$DEFFILE" ]; then
mkdir -p win32
sed < $SRCDIR/$DEFFILE.in "s/@ZLIB_SYMBOL_PREFIX@/$symbol_prefix/g" > $DEFFILE
fi
@@ -1476,14 +1476,14 @@ EOF
check_vgfma_intrinsics() {
# Check whether "VECTOR GALOIS FIELD MULTIPLY SUM AND ACCUMULATE" intrinsic is available
- echo -n "Checking for -mzarch... " | tee -a configure.log
+ printf "Checking for -mzarch... " | tee -a configure.log
if try $CC -x c -c /dev/null -o /dev/null -mzarch; then
echo Yes. | tee -a configure.log
vgfmaflag="${vgfmaflag} -mzarch"
else
echo No. | tee -a configure.log
fi
- echo -n "Checking for -fzvector... " | tee -a configure.log
+ printf "Checking for -fzvector... " | tee -a configure.log
if try $CC -x c -c /dev/null -o /dev/null -fzvector; then
echo Yes. | tee -a configure.log
vgfmaflag="${vgfmaflag} -fzvector"
@@ -1500,7 +1500,7 @@ int main(void) {
return c[0];
}
EOF
- echo -n "Checking for VGFMA support... " | tee -a configure.log
+ printf "Checking for VGFMA support... " | tee -a configure.log
if try $CC -c $CFLAGS $vgfmaflag $test.c; then
HAVE_VGFMA_INTRIN=1
echo "Yes." | tee -a configure.log
@@ -2156,11 +2156,11 @@ for file in $SRCDIR/*.c $SRCDIR/test/*.c $SRCDIR/test/fuzz/*.c $SRCDIR/$ARCHDIR/
# Check that the include file exists in the current dir,
# otherwise it may be one of the system include header.
if test -e $SRCDIR/$i; then
- echo -n " \$(SRCDIR)/$i"
+ printf " \$(SRCDIR)/$i"
fi
# We also need to check whether the include file is in the ARCHDIR.
if test -e $SRCDIR/$ARCHDIR/$i; then
- echo -n " \$(SRCDIR)/$ARCHDIR/$i"
+ printf " \$(SRCDIR)/$ARCHDIR/$i"
fi
done)
obj=$(basename $(echo $file | sed -e 's/\.c/\.o/g' -e 's#^\./##g'))
@@ -2233,11 +2233,11 @@ for file in $SRCDIR/$ARCHDIR/*.c; do
# Check that the include file exists in the current dir,
# otherwise it may be one of the system include header.
if test -e $SRCDIR/$i; then
- echo -n " \$(SRCTOP)/$i"
+ printf " \$(SRCTOP)/$i"
fi
# We also need to check whether the include file is in the ARCHDIR.
if test -e $SRCDIR/$ARCHDIR/$i; then
- echo -n " \$(SRCDIR)/$i"
+ printf " \$(SRCDIR)/$i"
fi
done)
obj=$(basename $(echo $file | sed -e 's/\.c/\.o/g' -e 's#^\./##g'))
--
2.39.2

432
var/spack/repos/builtin/packages/zlib-ng/pr-1562.patch Normal file
View file

@ -0,0 +1,432 @@
From 13df84c54aaf06cc7aeb1813ef60b17591d29ea3 Mon Sep 17 00:00:00 2001
From: Harmen Stoppels <me@harmenstoppels.nl>
Date: Mon, 21 Aug 2023 11:10:29 +0200
Subject: [PATCH 3/3] PR #1562
---
cmake/detect-intrinsics.cmake | 118 ++++++++++++----------------------
configure | 115 ++++++++++-----------------------
2 files changed, 73 insertions(+), 160 deletions(-)
diff --git a/cmake/detect-intrinsics.cmake b/cmake/detect-intrinsics.cmake
index 52c54dc8..d476093f 100644
--- a/cmake/detect-intrinsics.cmake
+++ b/cmake/detect-intrinsics.cmake
@@ -62,28 +62,19 @@ macro(check_avx512_intrinsics)
set(CMAKE_REQUIRED_FLAGS "${AVX512FLAG} ${NATIVEFLAG}")
check_c_source_compile_or_run(
"#include <immintrin.h>
- int main(void) {
- __m512i x = _mm512_set1_epi8(2);
- const __m512i y = _mm512_set_epi32(0x1020304, 0x5060708, 0x90a0b0c, 0xd0e0f10,
- 0x11121314, 0x15161718, 0x191a1b1c, 0x1d1e1f20,
- 0x21222324, 0x25262728, 0x292a2b2c, 0x2d2e2f30,
- 0x31323334, 0x35363738, 0x393a3b3c, 0x3d3e3f40);
- x = _mm512_sub_epi8(x, y);
- (void)x;
- return 0;
- }"
+ __m512i f(__m512i y) {
+ __m512i x = _mm512_set1_epi8(2);
+ return _mm512_sub_epi8(x, y);
+ }
+ int main(void) { return 0; }"
HAVE_AVX512_INTRIN
)
# Evidently both GCC and clang were late to implementing these
check_c_source_compile_or_run(
"#include <immintrin.h>
- int main(void) {
- __mmask16 a = 0xFF;
- a = _knot_mask16(a);
- (void)a;
- return 0;
- }"
+ __mmask16 f(__mmask16 x) { return _knot_mask16(x); }
+ int main(void) { return 0; }"
HAVE_MASK_INTRIN
)
set(CMAKE_REQUIRED_FLAGS)
@@ -117,17 +108,11 @@ macro(check_avx512vnni_intrinsics)
set(CMAKE_REQUIRED_FLAGS "${AVX512VNNIFLAG} ${NATIVEFLAG}")
check_c_source_compile_or_run(
"#include <immintrin.h>
- int main(void) {
- __m512i x = _mm512_set1_epi8(2);
- const __m512i y = _mm512_set_epi8(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19,
- 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37,
- 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55,
- 56, 57, 58, 59, 60, 61, 62, 63, 64);
+ __m512i f(__m512i x, __m512i y) {
__m512i z = _mm512_setzero_epi32();
- z = _mm512_dpbusd_epi32(z, x, y);
- (void)z;
- return 0;
- }"
+ return _mm512_dpbusd_epi32(z, x, y);
+ }
+ int main(void) { return 0; }"
HAVE_AVX512VNNI_INTRIN
)
set(CMAKE_REQUIRED_FLAGS)
@@ -151,13 +136,11 @@ macro(check_avx2_intrinsics)
set(CMAKE_REQUIRED_FLAGS "${AVX2FLAG} ${NATIVEFLAG}")
check_c_source_compile_or_run(
"#include <immintrin.h>
- int main(void) {
- __m256i x = _mm256_set1_epi16(2);
+ __m256i f(__m256i x) {
const __m256i y = _mm256_set1_epi16(1);
- x = _mm256_subs_epu16(x, y);
- (void)x;
- return 0;
- }"
+ return _mm256_subs_epu16(x, y);
+ }
+ int main(void) { return 0; }"
HAVE_AVX2_INTRIN
)
set(CMAKE_REQUIRED_FLAGS)
@@ -204,12 +187,8 @@ macro(check_neon_ld4_intrinsics)
#else
# include <arm_neon.h>
#endif
- int main(void) {
- int stack_var[16];
- int32x4x4_t v = vld1q_s32_x4(stack_var);
- (void)v;
- return 0;
- }"
+ int32x4x4_t f(int var[16]) { return vld1q_s32_x4(var); }
+ int main(void) { return 0; }"
NEON_HAS_LD4)
set(CMAKE_REQUIRED_FLAGS)
endmacro()
@@ -226,13 +205,9 @@ macro(check_pclmulqdq_intrinsics)
set(CMAKE_REQUIRED_FLAGS "${PCLMULFLAG} ${NATIVEFLAG}")
check_c_source_compile_or_run(
"#include <immintrin.h>
- int main(void) {
- __m128i a = _mm_setzero_si128();
- __m128i b = _mm_setzero_si128();
- __m128i c = _mm_clmulepi64_si128(a, b, 0x10);
- (void)c;
- return 0;
- }"
+ #include <wmmintrin.h>
+ __m128i f(__m128i a, __m128i b) { return _mm_clmulepi64_si128(a, b, 0x10); }
+ int main(void) { return 0; }"
HAVE_PCLMULQDQ_INTRIN
)
set(CMAKE_REQUIRED_FLAGS)
@@ -252,13 +227,12 @@ macro(check_vpclmulqdq_intrinsics)
set(CMAKE_REQUIRED_FLAGS "${VPCLMULFLAG} ${NATIVEFLAG}")
check_c_source_compile_or_run(
"#include <immintrin.h>
- int main(void) {
- __m512i a = _mm512_setzero_si512();
+ #include <wmmintrin.h>
+ __m512i f(__m512i a) {
__m512i b = _mm512_setzero_si512();
- __m512i c = _mm512_clmulepi64_epi128(a, b, 0x10);
- (void)c;
- return 0;
- }"
+ return _mm512_clmulepi64_epi128(a, b, 0x10);
+ }
+ int main(void) { return 0; }"
HAVE_VPCLMULQDQ_INTRIN
)
set(CMAKE_REQUIRED_FLAGS)
@@ -431,11 +405,8 @@ macro(check_sse2_intrinsics)
set(CMAKE_REQUIRED_FLAGS "${SSE2FLAG} ${NATIVEFLAG}")
check_c_source_compile_or_run(
"#include <immintrin.h>
- int main(void) {
- __m128i zero = _mm_setzero_si128();
- (void)zero;
- return 0;
- }"
+ __m128i f(__m128i x, __m128i y) { return _mm_sad_epu8(x, y); }
+ int main(void) { return 0; }"
HAVE_SSE2_INTRIN
)
set(CMAKE_REQUIRED_FLAGS)
@@ -457,14 +428,11 @@ macro(check_ssse3_intrinsics)
set(CMAKE_REQUIRED_FLAGS "${SSSE3FLAG} ${NATIVEFLAG}")
check_c_source_compile_or_run(
"#include <immintrin.h>
- int main(void) {
- __m128i u, v, w;
- u = _mm_set1_epi32(1);
- v = _mm_set1_epi32(2);
- w = _mm_hadd_epi32(u, v);
- (void)w;
- return 0;
- }"
+ __m128i f(__m128i u) {
+ __m128i v = _mm_set1_epi32(1);
+ return _mm_hadd_epi32(u, v);
+ }
+ int main(void) { return 0; }"
HAVE_SSSE3_INTRIN
)
endmacro()
@@ -485,13 +453,8 @@ macro(check_sse42_intrinsics)
set(CMAKE_REQUIRED_FLAGS "${SSE42FLAG} ${NATIVEFLAG}")
check_c_source_compile_or_run(
"#include <nmmintrin.h>
- int main(void) {
- unsigned crc = 0;
- char c = 'c';
- crc = _mm_crc32_u32(crc, c);
- (void)crc;
- return 0;
- }"
+ unsigned int f(unsigned int a, unsigned int b) { return _mm_crc32_u32(a, b); }
+ int main(void) { return 0; }"
HAVE_SSE42_INTRIN
)
set(CMAKE_REQUIRED_FLAGS)
@@ -529,13 +492,12 @@ macro(check_xsave_intrinsics)
set(CMAKE_REQUIRED_FLAGS "${XSAVEFLAG} ${NATIVEFLAG}")
check_c_source_compiles(
"#ifdef _WIN32
- # include <intrin.h>
- #else
- # include <x86gprintrin.h>
- #endif
- int main(void) {
- return _xgetbv(0);
- }"
+ # include <intrin.h>
+ #else
+ # include <x86gprintrin.h>
+ #endif
+ unsigned int f(unsigned int a) { return _xgetbv(a); }
+ int main(void) { return 0; }"
HAVE_XSAVE_INTRIN FAIL_REGEX "not supported")
set(CMAKE_REQUIRED_FLAGS)
endmacro()
diff --git a/configure b/configure
index fc78a135..e5a1965f 100755
--- a/configure
+++ b/configure
@@ -1023,12 +1023,8 @@ fi
# Check for __builtin_ctz() support in compiler
cat > $test.c << EOF
-int main(void) {
- unsigned int zero = 0;
- long test = __builtin_ctz(zero);
- (void)test;
- return 0;
-}
+long f(unsigned int x) { return __builtin_ctz(x); }
+int main(void) { return 0; }
EOF
if try ${CC} ${CFLAGS} $test.c $LDSHAREDLIBC; then
echo "Checking for __builtin_ctz ... Yes." | tee -a configure.log
@@ -1040,12 +1036,8 @@ fi
# Check for __builtin_ctzll() support in compiler
cat > $test.c << EOF
-int main(void) {
- unsigned long long zero = 0;
- long test = __builtin_ctzll(zero);
- (void)test;
- return 0;
-}
+long f(unsigned long long x) { return __builtin_ctzll(x); }
+int main(void) { return 0; }
EOF
if try ${CC} ${CFLAGS} $test.c $LDSHAREDLIBC; then
echo "Checking for __builtin_ctzll ... Yes." | tee -a configure.log
@@ -1059,13 +1051,11 @@ check_avx2_intrinsics() {
# Check whether compiler supports AVX2 intrinsics
cat > $test.c << EOF
#include <immintrin.h>
-int main(void) {
- __m256i x = _mm256_set1_epi16(2);
+__m256i f(__m256i x) {
const __m256i y = _mm256_set1_epi16(1);
- x = _mm256_subs_epu16(x, y);
- (void)x;
- return 0;
+ return _mm256_subs_epu16(x, y);
}
+int main(void) { return 0; }
EOF
if try ${CC} ${CFLAGS} ${avx2flag} $test.c; then
echo "Checking for AVX2 intrinsics ... Yes." | tee -a configure.log
@@ -1080,16 +1070,11 @@ check_avx512_intrinsics() {
# Check whether compiler supports AVX512 intrinsics
cat > $test.c << EOF
#include <immintrin.h>
-int main(void) {
- __m512i x = _mm512_set1_epi8(2);
- const __m512i y = _mm512_set_epi32(0x1020304, 0x5060708, 0x90a0b0c, 0xd0e0f10,
- 0x11121314, 0x15161718, 0x191a1b1c, 0x1d1e1f20,
- 0x21222324, 0x25262728, 0x292a2b2c, 0x2d2e2f30,
- 0x31323334, 0x35363738, 0x393a3b3c, 0x3d3e3f40);
- x = _mm512_sub_epi8(x, y);
- (void)x;
- return 0;
+__m512i f(__m512i y) {
+ __m512i x = _mm512_set1_epi8(2);
+ return _mm512_sub_epi8(x, y);
}
+int main(void) { return 0; }
EOF
if try ${CC} ${CFLAGS} ${avx512flag} $test.c; then
echo "Checking for AVX512 intrinsics ... Yes." | tee -a configure.log
@@ -1133,17 +1118,11 @@ check_avx512vnni_intrinsics() {
# Check whether compiler supports AVX512-VNNI intrinsics
cat > $test.c << EOF
#include <immintrin.h>
-int main(void) {
- __m512i x = _mm512_set1_epi8(2);
- const __m512i y = _mm512_set_epi8(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19,
- 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37,
- 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55,
- 56, 57, 58, 59, 60, 61, 62, 63, 64);
+__m512i f(__m512i x, __m512i y) {
__m512i z = _mm512_setzero_epi32();
- z = _mm512_dpbusd_epi32(z, x, y);
- (void)z;
- return 0;
+ return _mm512_dpbusd_epi32(z, x, y);
}
+int main(void) { return 0; }
EOF
if try ${CC} ${CFLAGS} ${avx512vnniflag} $test.c; then
echo "Checking for AVX512VNNI intrinsics ... Yes." | tee -a configure.log
@@ -1158,12 +1137,8 @@ check_mask_intrinsics() {
# Check whether compiler supports AVX512 k-mask intrinsics
cat > $test.c << EOF
#include <immintrin.h>
-int main(void) {
- __mmask16 a = 0xFF;
- a = _knot_mask16(a);
- (void)a;
- return 0;
-}
+__mmask16 f(__mmask16 x) { return _knot_mask16(x); }
+int main(void) { return 0; }
EOF
if try ${CC} ${CFLAGS} ${avx512flag} $test.c; then
echo "Checking for AVX512 k-mask intrinsics ... Yes." | tee -a configure.log
@@ -1230,12 +1205,8 @@ check_neon_ld4_intrinsics() {
#else
# include <arm_neon.h>
#endif
-int main(void) {
- int stack_var[16];
- int32x4x4_t v = vld1q_s32_x4(stack_var);
- (void)v;
- return 0;
-}
+int32x4x4_t f(int var[16]) { return vld1q_s32_x4(var); }
+int main(void) { return 0; }
EOF
if try $CC -c $CFLAGS $neonflag $test.c; then
NEON_HAS_LD4=1
@@ -1251,13 +1222,8 @@ check_pclmulqdq_intrinsics() {
cat > $test.c << EOF
#include <immintrin.h>
#include <wmmintrin.h>
-int main(void) {
- __m128i a = _mm_setzero_si128();
- __m128i b = _mm_setzero_si128();
- __m128i c = _mm_clmulepi64_si128(a, b, 0x10);
- (void)c;
- return 0;
-}
+__m128i f(__m128i a, __m128i b) { return _mm_clmulepi64_si128(a, b, 0x10); }
+int main(void) { return 0; }
EOF
if try ${CC} ${CFLAGS} ${pclmulflag} $test.c; then
echo "Checking for PCLMULQDQ intrinsics ... Yes." | tee -a configure.log
@@ -1273,13 +1239,11 @@ check_vpclmulqdq_intrinsics() {
cat > $test.c << EOF
#include <immintrin.h>
#include <wmmintrin.h>
-int main(void) {
- __m512i a = _mm512_setzero_si512();
+__m512i f(__m512i a) {
__m512i b = _mm512_setzero_si512();
- __m512i c = _mm512_clmulepi64_epi128(a, b, 0x10);
- (void)c;
- return 0;
+ return _mm512_clmulepi64_epi128(a, b, 0x10);
}
+int main(void) { return 0; }
EOF
if try ${CC} ${CFLAGS} ${vpclmulflag} $test.c; then
echo "Checking for VPCLMULQDQ intrinsics ... Yes." | tee -a configure.log
@@ -1298,9 +1262,8 @@ check_xsave_intrinsics() {
#else
# include <x86gprintrin.h>
#endif
-int main(void) {
- return _xgetbv(0);
-}
+unsigned int f(unsigned int a) { return _xgetbv(a); }
+int main(void) { return 0; }
EOF
if try ${CC} ${CFLAGS} ${xsaveflag} $test.c; then
echo "Checking for XSAVE intrinsics ... Yes." | tee -a configure.log
@@ -1415,11 +1378,8 @@ check_sse2_intrinsics() {
# Check whether compiler supports SSE2 intrinsics
cat > $test.c << EOF
#include <immintrin.h>
-int main(void) {
- __m128i zero = _mm_setzero_si128();
- (void)zero;
- return 0;
-}
+__m128i f(__m128i x, __m128i y) { return _mm_sad_epu8(x, y); }
+int main(void) { return 0; }
EOF
if try ${CC} ${CFLAGS} ${sse2flag} $test.c; then
echo "Checking for SSE2 intrinsics ... Yes." | tee -a configure.log
@@ -1434,13 +1394,8 @@ check_sse42_intrinsics() {
# Check whether compiler supports SSE4.2 intrinsics
cat > $test.c << EOF
#include <nmmintrin.h>
-int main(void) {
- unsigned crc = 0;
- char c = 'c';
- crc = _mm_crc32_u32(crc, c);
- (void)crc;
- return 0;
-}
+unsigned int f(unsigned int a, unsigned int b) { return _mm_crc32_u32(a, b); }
+int main(void) { return 0; }
EOF
if try ${CC} ${CFLAGS} ${sse42flag} $test.c; then
echo "Checking for SSE4.2 intrinsics ... Yes." | tee -a configure.log
@@ -1455,15 +1410,11 @@ check_ssse3_intrinsics() {
# Check whether compiler supports SSSE3 intrinsics
cat > $test.c << EOF
#include <immintrin.h>
-int main(void)
-{
- __m128i u, v, w;
- u = _mm_set1_epi32(1);
- v = _mm_set1_epi32(2);
- w = _mm_hadd_epi32(u, v);
- (void)w;
- return 0;
+__m128i f(__m128i u) {
+ __m128i v = _mm_set1_epi32(1);
+ return _mm_hadd_epi32(u, v);
}
+int main(void) { return 0; }
EOF
if try ${CC} ${CFLAGS} ${ssse3flag} $test.c; then
echo "Checking for SSSE3 intrinsics ... Yes." | tee -a configure.log
--
2.39.2