[Mesa-dev] [PATCH] swr: Refactor checks for compiler feature flags
Chuck Atkins
chuck.atkins at kitware.com
Tue Jun 28 18:14:15 UTC 2016
The only guaranteed way I can think of to ensure compiler support is to try
compiling source that calls one intrinsic from each of the used groups. I
can see that being "more correct" but I can't really think of a situation
where just checking for the __AVX2__ define will fail to build wither.
- Chuck
On Tue, Jun 28, 2016 at 2:10 PM, Chuck Atkins <chuck.atkins at kitware.com>
wrote:
> So this seems to be different across versions as well. It looks like
> __AVX__ and __AVX2__ are the only ones we can really count on being there.
> I can drop the second check to just __AVX2__. I think it's redundant by
> chance though that all CPUs that supported AVX2 also seem to support the
> additional 2 instructions.
>
> - Chuck
>
> On Tue, Jun 28, 2016 at 1:52 PM, Rowley, Timothy O <
> timothy.o.rowley at intel.com> wrote:
>
>>
>> > On Jun 28, 2016, at 8:24 AM, Chuck Atkins <chuck.atkins at kitware.com>
>> wrote:
>> >
>> > Encapsulate the test for which flags are needed to get a compiler to
>> > support certain features. Along with this, give various options to try
>> > for AVX and AVX2 support. Ideally we want to use specific instruction
>> > set feature flags, like -mavx2 for instance instead of -march=haswell,
>> > but the flags required for certain compilers are different. This
>> > allows, for AVX2 for instance, GCC to use -mavx2 -mfma -mbmi2 -mf16c
>> > while the Intel compiler which doesn't support those flags can fall
>> > back to using -march=core-avx2.
>> >
>> > This addresses a bug where the Intel compiler will silently ignore the
>> > AVX2 instruction feature flags and then potentially fail to build.
>> >
>> > Cc: Tim Rowley <timothy.o.rowley at intel.com>
>> > Signed-off-by: Chuck Atkins <chuck.atkins at kitware.com>
>> > ---
>> > configure.ac | 86
>> +++++++++++++++++++++++++++++++++++++++++++-----------------
>> > 1 file changed, 62 insertions(+), 24 deletions(-)
>> >
>> > diff --git a/configure.ac b/configure.ac
>> > index cc9bc47..806850e 100644
>> > --- a/configure.ac
>> > +++ b/configure.ac
>> > @@ -2330,6 +2330,39 @@ swr_llvm_check() {
>> > fi
>> > }
>> >
>> > +swr_cxx_feature_flags_check() {
>> > + ifndef_test=$1
>> > + option_list="$2"
>> > + unset SWR_CXX_FEATURE_FLAGS
>> > + AC_LANG_PUSH([C++])
>> > + save_CXXFLAGS="$CXXFLAGS"
>> > + save_IFS="$IFS"
>> > + IFS=","
>> > + found=0
>> > + for opts in $option_list
>> > + do
>> > + unset IFS
>> > + CXXFLAGS="$opts $save_CXXFLAGS"
>> > + AC_COMPILE_IFELSE(
>> > + [AC_LANG_PROGRAM(
>> > + [ $ifndef_test
>> > + #error
>> > + #endif
>> > + ])],
>> > + [found=1; break],
>> > + [])
>> > + IFS=","
>> > + done
>> > + IFS="$save_IFS"
>> > + CXXFLAGS="$save_CXXFLAGS"
>> > + AC_LANG_POP([C++])
>> > + if test $found -eq 1; then
>> > + SWR_CXX_FEATURE_FLAGS="$opts"
>> > + return 0
>> > + fi
>> > + return 1
>> > +}
>> > +
>> > dnl Duplicates in GALLIUM_DRIVERS_DIRS are removed by sorting it after
>> this block
>> > if test -n "$with_gallium_drivers"; then
>> > gallium_drivers=`IFS=', '; echo $with_gallium_drivers`
>> > @@ -2399,31 +2432,36 @@ if test -n "$with_gallium_drivers"; then
>> > xswr)
>> > swr_llvm_check "swr"
>> >
>> > - AC_MSG_CHECKING([whether $CXX supports c++11/AVX/AVX2])
>> > - SWR_AVX_CXXFLAGS="-mavx"
>> > - SWR_AVX2_CXXFLAGS="-mavx2 -mfma -mbmi2 -mf16c"
>> > -
>> > - AC_LANG_PUSH([C++])
>> > - save_CXXFLAGS="$CXXFLAGS"
>> > - CXXFLAGS="-std=c++11 $CXXFLAGS"
>> > - AC_COMPILE_IFELSE([AC_LANG_PROGRAM()],[],
>> > - [AC_MSG_ERROR([c++11 compiler support
>> not detected])])
>> > - CXXFLAGS="$save_CXXFLAGS"
>> > -
>> > - save_CXXFLAGS="$CXXFLAGS"
>> > - CXXFLAGS="$SWR_AVX_CXXFLAGS $CXXFLAGS"
>> > - AC_COMPILE_IFELSE([AC_LANG_PROGRAM()],[],
>> > - [AC_MSG_ERROR([AVX compiler support not
>> detected])])
>> > - CXXFLAGS="$save_CXXFLAGS"
>> > -
>> > - save_CFLAGS="$CXXFLAGS"
>> > - CXXFLAGS="$SWR_AVX2_CXXFLAGS $CXXFLAGS"
>> > - AC_COMPILE_IFELSE([AC_LANG_PROGRAM()],[],
>> > - [AC_MSG_ERROR([AVX2 compiler support not
>> detected])])
>> > - CXXFLAGS="$save_CXXFLAGS"
>> > - AC_LANG_POP([C++])
>> > -
>> > + AC_MSG_CHECKING([whether $CXX supports c++11])
>> > + if ! swr_cxx_feature_flags_check \
>> > + "#if __cplusplus < 201103L" \
>> > + ",-std=c++11"; then
>> > + AC_MSG_RESULT([no])
>> > + AC_MSG_ERROR([swr requires C++11 support])
>> > + fi
>> > + AC_MSG_RESULT([$SWR_CXX_FEATURE_FLAGS])
>> > + CXXFLAGS="$SWR_CXX_FEATURE_FLAGS $CXXFLAGS”
>>
>> We don’t want to globally override CXXFLAGS; AC_SUBST on a SWR_CXXFLAGS
>> and using that in swr’s Makefile.am would be better.
>>
>> > +
>> > + AC_MSG_CHECKING([whether $CXX supports AVX])
>> > + if ! swr_cxx_feature_flags_check \
>> > + "#ifndef __AVX__" \
>> > + ",-mavx,-march=core-avx"; then
>> > + AC_MSG_RESULT([no])
>> > + AC_MSG_ERROR([swr requires AVX compiler support])
>> > + fi
>> > + AC_MSG_RESULT([$SWR_CXX_FEATURE_FLAGS])
>> > + SWR_AVX_CXXFLAGS="$SWR_CXX_FEATURE_FLAGS"
>> > AC_SUBST([SWR_AVX_CXXFLAGS])
>> > +
>> > + AC_MSG_CHECKING([whether $CXX supports AVX2])
>> > + if ! swr_cxx_feature_flags_check \
>> > + "#if
>> !(defined(__AVX2__)&&defined(__FMA__)&&defined(__BMI2__)&&defined(__F16C__))”
>> \
>>
>> Is there any standard that says these are defined if the compiler
>> supports them? With icc 16.0.3, the test falls into the #error path when
>> it tries the fallback test of -march=core-avx2.
>>
>> > + ",-mavx2 -mfma -mbmi2 -mf16c,-march=core-avx2"; then
>> > + AC_MSG_RESULT([no])
>> > + AC_MSG_ERROR([swr requires AVX2 compiler support])
>> > + fi
>> > + AC_MSG_RESULT([$SWR_CXX_FEATURE_FLAGS])
>> > + SWR_AVX2_CXXFLAGS="$SWR_CXX_FEATURE_FLAGS"
>> > AC_SUBST([SWR_AVX2_CXXFLAGS])
>> >
>> > HAVE_GALLIUM_SWR=yes
>> > --
>> > 2.5.5
>> >
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20160628/3db40d75/attachment.html>
More information about the mesa-dev
mailing list