[Mesa-dev] [PATCH 1/2] mesa: Drop -fno-builtin-memcmp.

Roland Scheidegger sroland at vmware.com
Thu Jun 30 23:26:51 UTC 2016


Am 01.07.2016 um 00:59 schrieb Matt Turner:
> According to the referenced bug report, gcc-4.5 and newer do not inline
> memcmp(). I see no difference in performance of ipers with llvmpipe on a
> Sandybridge (which does not have "Enhanced REP MOVSB/STOSB") by removing
> this flag.
> 
> I attempted to confirm the problem with gcc-4.4, but it fails to compile
> for quite a few different reasons.
> ---
>  configure.ac | 8 --------
>  1 file changed, 8 deletions(-)
> 
> diff --git a/configure.ac b/configure.ac
> index bd8842d..95cfc36 100644
> --- a/configure.ac
> +++ b/configure.ac
> @@ -304,10 +304,6 @@ if test "x$GCC" = xyes; then
>      # We don't want floating-point math functions to set errno or trap
>      CFLAGS="$CFLAGS -fno-math-errno -fno-trapping-math"
>  
> -    # gcc's builtin memcmp is slower than glibc's
> -    # http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43052
> -    CFLAGS="$CFLAGS -fno-builtin-memcmp"
> -
>      # Flags to help ensure that certain portions of the code -- and only those
>      # portions -- can be built with MSVC:
>      # - src/util, src/gallium/auxiliary, rc/gallium/drivers/llvmpipe, and
> @@ -343,10 +339,6 @@ if test "x$GXX" = xyes; then
>  
>      # Restore CXXFLAGS; VISIBILITY_CXXFLAGS are added to it where needed.
>      CXXFLAGS=$save_CXXFLAGS
> -
> -    # gcc's builtin memcmp is slower than glibc's
> -    # http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43052
> -    CXXFLAGS="$CXXFLAGS -fno-builtin-memcmp"
>  fi
>  
>  AC_SUBST([MSVC2013_COMPAT_CFLAGS])
> 

FWIW we're using gcc 4.4.3...
Technically the configure script only requires gcc 4.2 (albeit some
parts have higher requirements) so if it doesn't compile that's a bug.

But we're using scons to build so this should be ok (albeit I think
different flags in configure.ac and scons are generally not a really
good idea).

Roland



More information about the mesa-dev mailing list