[PATCH 28/59] x86: Use -msse2avx iff compiler supports it

[Adhemerval Zanella Netto]

On 17/10/25 19:30, H.J. Lu wrote:
> On Sat, Oct 18, 2025 at 3:51 AM Adhemerval Zanella
> <adhemerval.zanella@linaro.org> wrote:
>>
>> clang supports it from version 19 and onwards, but it should be gated
>> as an option to assembler (either with -Wa or -Xassembler).  Aadd a
>> test for -Wa as well if -msse2avx fully supported by the compiler.
>> ---
>>  sysdeps/x86_64/configure              | 24 ++++++++++++++++++++++++
>>  sysdeps/x86_64/configure.ac           | 11 +++++++++++
>>  sysdeps/x86_64/fpu/multiarch/Makefile | 14 +++++++-------
>>  3 files changed, 42 insertions(+), 7 deletions(-)
>>
>> diff --git a/sysdeps/x86_64/configure b/sysdeps/x86_64/configure
>> index 32324f62da..9dd2e4a179 100644
>> --- a/sysdeps/x86_64/configure
>> +++ b/sysdeps/x86_64/configure
>> @@ -210,6 +210,30 @@ fi
>>  config_vars="$config_vars
>>  enable-cet = $enable_cet"
>>
>> +{ printf "%s\n" "$as_me:${as_lineno-$LINENO}: checking -msse2avx" >&5
>> +printf %s "checking -msse2avx... " >&6; }
>> +if test ${libc_cv_cc_msse2avx+y}
>> +then :
>> +  printf %s "(cached) " >&6
>> +else case e in #(
>> +  e)   for opt in "-msse2avx" "-Wa,-msse2avx"; do
>> +    if { ac_try='${CC-cc} $CFLAGS -c -xc /dev/null -o /dev/null -c "$opt" 1>&5'
>> +  { { eval echo "\"\$as_me\":${as_lineno-$LINENO}: \"$ac_try\""; } >&5
>> +  (eval $ac_try) 2>&5
>> +  ac_status=$?
>> +  printf "%s\n" "$as_me:${as_lineno-$LINENO}: \$? = $ac_status" >&5
>> +  test $ac_status = 0; }; }; then
>> +      libc_cv_cc_msse2avx="$opt"; break;
>> +    fi
>> +  done
>> + ;;
>> +esac
>> +fi
>> +{ printf "%s\n" "$as_me:${as_lineno-$LINENO}: result: $libc_cv_cc_msse2avx" >&5
>> +printf "%s\n" "$libc_cv_cc_msse2avx" >&6; }
>> +config_vars="$config_vars
>> +config-cflags-msse2avx = $libc_cv_cc_msse2avx"
>> +
>>  # Check if -mamx-tile works properly.
>>
>>  cat > conftest.c <<EOF
>> diff --git a/sysdeps/x86_64/configure.ac b/sysdeps/x86_64/configure.ac
>> index a00958e219..d1bcfd7ca4 100644
>> --- a/sysdeps/x86_64/configure.ac
>> +++ b/sysdeps/x86_64/configure.ac
>> @@ -67,6 +67,17 @@ elif test $enable_cet = permissive; then
>>  fi
>>  LIBC_CONFIG_VAR([enable-cet], [$enable_cet])
>>
>> +dnl Check if compiler supports -msse2avx or -Wa,-msse2avx
>> +AC_CACHE_CHECK(-msse2avx, libc_cv_cc_msse2avx, [dnl
>> +  for opt in "-msse2avx" "-Wa,-msse2avx"; do
>> +    if AC_TRY_COMMAND(${CC-cc} $CFLAGS -c -xc /dev/null -o /dev/null -c "$opt" 1>&AS_MESSAGE_LOG_FD); then
>> +      libc_cv_cc_msse2avx="$opt"; break;
>> +    fi
>> +  done
>> +])
>> +LIBC_CONFIG_VAR([config-cflags-msse2avx],
>> +               [$libc_cv_cc_msse2avx])
>> +
>>  # Check if -mamx-tile works properly.
>>  LIBC_TRY_TEST_CC_COMMAND([whether -mamx-tile works properly],
>>    [#include <x86intrin.h>],
>> diff --git a/sysdeps/x86_64/fpu/multiarch/Makefile b/sysdeps/x86_64/fpu/multiarch/Makefile
>> index 708b14297f..50b66509f9 100644
>> --- a/sysdeps/x86_64/fpu/multiarch/Makefile
>> +++ b/sysdeps/x86_64/fpu/multiarch/Makefile
>> @@ -150,13 +150,13 @@ CFLAGS-s_sin-fma4.c = -mfma4
>>  CFLAGS-s_tan-fma4.c = -mfma4
>>  CFLAGS-s_sincos-fma4.c = -mfma4
>>
>> -CFLAGS-e_atan2-avx.c = -msse2avx -DSSE2AVX
>> -CFLAGS-e_exp-avx.c = -msse2avx -DSSE2AVX
>> -CFLAGS-e_log-avx.c = -msse2avx -DSSE2AVX
>> -CFLAGS-s_atan-avx.c = -msse2avx -DSSE2AVX
>> -CFLAGS-s_sin-avx.c = -msse2avx -DSSE2AVX
>> -CFLAGS-s_tan-avx.c = -msse2avx -DSSE2AVX
>> -CFLAGS-s_sincos-avx.c = -msse2avx -DSSE2AVX
>> +CFLAGS-e_atan2-avx.c = $(config-cflags-msse2avx) -DSSE2AVX
>> +CFLAGS-e_exp-avx.c = $(config-cflags-msse2avx) -DSSE2AVX
>> +CFLAGS-e_log-avx.c = $(config-cflags-msse2avx) -DSSE2AVX
>> +CFLAGS-s_atan-avx.c = $(config-cflags-msse2avx) -DSSE2AVX
>> +CFLAGS-s_sin-avx.c = $(config-cflags-msse2avx) -DSSE2AVX
>> +CFLAGS-s_tan-avx.c = $(config-cflags-msse2avx) -DSSE2AVX
>> +CFLAGS-s_sincos-avx.c = $(config-cflags-msse2avx) -DSSE2AVX
> 
> Why not simply compile them with -mavx2?

Alright, I will change it.

> 
>>  CFLAGS-s_modf-sse4_1.c = -msse4.1 -fno-builtin-modff32x -fno-builtin-modff64
>>  CFLAGS-s_modff-sse4_1.c = -msse4.1 -fno-builtin-modff32
>> --
>> 2.43.0
>>
> 
>