thr3ads.net - opus - [opus] [PATCH 1/2] Modify autoconf tests for intrinsics to stop clang from optimizing them away. [May 2016]

If this information is useful, please help other people find it:
Share via:

Jonathan Lennox

2016-May-31 19:52 UTC

[opus] [PATCH 1/2] Modify autoconf tests for intrinsics to stop clang from optimizing them away.

---
 configure.ac | 30 +++++++++++++++++++++---------
 1 file changed, 21 insertions(+), 9 deletions(-)

diff --git a/configure.ac b/configure.ac
index a67aa37..c722556 100644
--- a/configure.ac
+++ b/configure.ac
@@ -472,6 +472,7 @@ AS_IF([test x"$enable_intrinsics" =
x"yes"],[
          [[
             static float32x4_t A0, A1, SUMM;
             SUMM = vmlaq_f32(SUMM, A0, A1);
+            return (int)vgetq_lane_f32(SUMM, 0);
          ]]
       )
       AS_IF([test x"$OPUS_ARM_MAY_HAVE_NEON_INTR" = x"1"
&& test x"$OPUS_ARM_PRESUME_NEON_INTR" != x"1"],
@@ -521,10 +522,13 @@ AS_IF([test x"$enable_intrinsics" =
x"yes"],[
          [OPUS_X86_MAY_HAVE_SSE],
          [OPUS_X86_PRESUME_SSE],
          [[#include <xmmintrin.h>
+           #include <time.h>
          ]],
          [[
-             static __m128 mtest;
-             mtest = _mm_setzero_ps();
+             __m128 mtest;
+             mtest = _mm_set1_ps((float)time(NULL));
+             mtest = _mm_mul_ps(mtest, mtest);
+             return _mm_cvtss_si32(mtest);
          ]]
       )
       AS_IF([test x"$OPUS_X86_MAY_HAVE_SSE" = x"1"
&& test x"$OPUS_X86_PRESUME_SSE" != x"1"],
@@ -539,10 +543,13 @@ AS_IF([test x"$enable_intrinsics" =
x"yes"],[
          [OPUS_X86_MAY_HAVE_SSE2],
          [OPUS_X86_PRESUME_SSE2],
          [[#include <emmintrin.h>
+           #include <time.h>
          ]],
          [[
-             static __m128i mtest;
-             mtest = _mm_setzero_si128();
+            __m128i mtest;
+            mtest = _mm_set1_epi32((int)time(NULL));
+            mtest = _mm_mul_epu32(mtest, mtest);
+            return _mm_cvtsi128_si32(mtest);
          ]]
       )
       AS_IF([test x"$OPUS_X86_MAY_HAVE_SSE2" = x"1"
&& test x"$OPUS_X86_PRESUME_SSE2" != x"1"],
@@ -557,11 +564,13 @@ AS_IF([test x"$enable_intrinsics" =
x"yes"],[
          [OPUS_X86_MAY_HAVE_SSE4_1],
          [OPUS_X86_PRESUME_SSE4_1],
          [[#include <smmintrin.h>
+           #include <time.h>
          ]],
          [[
-            static __m128i mtest;
-            mtest = _mm_setzero_si128();
-            mtest = _mm_cmpeq_epi64(mtest, mtest);
+            __m128i mtest;
+            mtest = _mm_set1_epi32((int)time(NULL));
+            mtest = _mm_mul_epi32(mtest, mtest);
+            return _mm_cvtsi128_si32(mtest);
          ]]
       )
       AS_IF([test x"$OPUS_X86_MAY_HAVE_SSE4_1" = x"1"
&& test x"$OPUS_X86_PRESUME_SSE4_1" != x"1"],
@@ -576,10 +585,13 @@ AS_IF([test x"$enable_intrinsics" =
x"yes"],[
          [OPUS_X86_MAY_HAVE_AVX],
          [OPUS_X86_PRESUME_AVX],
          [[#include <immintrin.h>
+           #include <time.h>
          ]],
          [[
-            static __m256 mtest;
-            mtest = _mm256_setzero_ps();
+             __m256 mtest;
+             mtest = _mm256_set1_ps((float)time(NULL));
+             mtest = _mm256_addsub_ps(mtest, mtest);
+             return _mm_cvtss_si32(_mm256_extractf128_ps(mtest, 0));
          ]]
       )
       AS_IF([test x"$OPUS_X86_MAY_HAVE_AVX" = x"1"
&& test x"$OPUS_X86_PRESUME_AVX" != x"1"],
-- 
2.7.4 (Apple Git-66)

Jonathan Lennox

2016-May-31 19:52 UTC

head link

[opus] [PATCH 2/2] List intrinsics flags being tested after existing cflags, so they override other options.

---
 m4/opus-intrinsics.m4 | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/m4/opus-intrinsics.m4 b/m4/opus-intrinsics.m4
index b93ddd3..a262ca1 100644
--- a/m4/opus-intrinsics.m4
+++ b/m4/opus-intrinsics.m4
@@ -15,7 +15,7 @@ AC_DEFUN([OPUS_CHECK_INTRINSICS],
         $4=0
         AC_MSG_RESULT([no])
         AC_MSG_CHECKING([if compiler supports $1 intrinsics with $2])
-        save_CFLAGS="$CFLAGS"; CFLAGS="$2 $CFLAGS"
+        save_CFLAGS="$CFLAGS"; CFLAGS="$CFLAGS $2"
         AC_LINK_IFELSE([AC_LANG_PROGRAM($5, $6)],
         [
            AC_MSG_RESULT([yes])
-- 
2.7.4 (Apple Git-66)

Jean-Marc Valin

2016-May-31 21:59 UTC

head link

[opus] [PATCH 1/2] Modify autoconf tests for intrinsics to stop clang from optimizing them away.

Hi Jonathan,

Thanks for the patch. It's now in master.

Cheers,

	Jean-Marc

On 05/31/2016 03:52 PM, Jonathan Lennox wrote:> ---
>  configure.ac | 30 +++++++++++++++++++++---------
>  1 file changed, 21 insertions(+), 9 deletions(-)
> 
> diff --git a/configure.ac b/configure.ac
> index a67aa37..c722556 100644
> --- a/configure.ac
> +++ b/configure.ac
> @@ -472,6 +472,7 @@ AS_IF([test x"$enable_intrinsics" =
x"yes"],[
>           [[
>              static float32x4_t A0, A1, SUMM;
>              SUMM = vmlaq_f32(SUMM, A0, A1);
> +            return (int)vgetq_lane_f32(SUMM, 0);
>           ]]
>        )
>        AS_IF([test x"$OPUS_ARM_MAY_HAVE_NEON_INTR" =
x"1" && test x"$OPUS_ARM_PRESUME_NEON_INTR" !=
x"1"],
> @@ -521,10 +522,13 @@ AS_IF([test x"$enable_intrinsics" =
x"yes"],[
>           [OPUS_X86_MAY_HAVE_SSE],
>           [OPUS_X86_PRESUME_SSE],
>           [[#include <xmmintrin.h>
> +           #include <time.h>
>           ]],
>           [[
> -             static __m128 mtest;
> -             mtest = _mm_setzero_ps();
> +             __m128 mtest;
> +             mtest = _mm_set1_ps((float)time(NULL));
> +             mtest = _mm_mul_ps(mtest, mtest);
> +             return _mm_cvtss_si32(mtest);
>           ]]
>        )
>        AS_IF([test x"$OPUS_X86_MAY_HAVE_SSE" = x"1"
&& test x"$OPUS_X86_PRESUME_SSE" != x"1"],
> @@ -539,10 +543,13 @@ AS_IF([test x"$enable_intrinsics" =
x"yes"],[
>           [OPUS_X86_MAY_HAVE_SSE2],
>           [OPUS_X86_PRESUME_SSE2],
>           [[#include <emmintrin.h>
> +           #include <time.h>
>           ]],
>           [[
> -             static __m128i mtest;
> -             mtest = _mm_setzero_si128();
> +            __m128i mtest;
> +            mtest = _mm_set1_epi32((int)time(NULL));
> +            mtest = _mm_mul_epu32(mtest, mtest);
> +            return _mm_cvtsi128_si32(mtest);
>           ]]
>        )
>        AS_IF([test x"$OPUS_X86_MAY_HAVE_SSE2" = x"1"
&& test x"$OPUS_X86_PRESUME_SSE2" != x"1"],
> @@ -557,11 +564,13 @@ AS_IF([test x"$enable_intrinsics" =
x"yes"],[
>           [OPUS_X86_MAY_HAVE_SSE4_1],
>           [OPUS_X86_PRESUME_SSE4_1],
>           [[#include <smmintrin.h>
> +           #include <time.h>
>           ]],
>           [[
> -            static __m128i mtest;
> -            mtest = _mm_setzero_si128();
> -            mtest = _mm_cmpeq_epi64(mtest, mtest);
> +            __m128i mtest;
> +            mtest = _mm_set1_epi32((int)time(NULL));
> +            mtest = _mm_mul_epi32(mtest, mtest);
> +            return _mm_cvtsi128_si32(mtest);
>           ]]
>        )
>        AS_IF([test x"$OPUS_X86_MAY_HAVE_SSE4_1" = x"1"
&& test x"$OPUS_X86_PRESUME_SSE4_1" != x"1"],
> @@ -576,10 +585,13 @@ AS_IF([test x"$enable_intrinsics" =
x"yes"],[
>           [OPUS_X86_MAY_HAVE_AVX],
>           [OPUS_X86_PRESUME_AVX],
>           [[#include <immintrin.h>
> +           #include <time.h>
>           ]],
>           [[
> -            static __m256 mtest;
> -            mtest = _mm256_setzero_ps();
> +             __m256 mtest;
> +             mtest = _mm256_set1_ps((float)time(NULL));
> +             mtest = _mm256_addsub_ps(mtest, mtest);
> +             return _mm_cvtss_si32(_mm256_extractf128_ps(mtest, 0));
>           ]]
>        )
>        AS_IF([test x"$OPUS_X86_MAY_HAVE_AVX" = x"1"
&& test x"$OPUS_X86_PRESUME_AVX" != x"1"],
>

Reasonably Related Threads

Search for more seemingly similar threads

opus - May 2016 - [PATCH 1/2] Modify autoconf tests for intrinsics to stop clang from optimizing them away.

[opus] [PATCH 1/2] Modify autoconf tests for intrinsics to stop clang from optimizing them away.

[opus] [PATCH 2/2] List intrinsics flags being tested after existing cflags, so they override other options.

[opus] [PATCH 1/2] Modify autoconf tests for intrinsics to stop clang from optimizing them away.

Reasonably Related Threads