[FFmpeg-cvslog] x86: check for AVX support

[Mans Rullgard]

ffmpeg | branch: master | Mans Rullgard <mans at mansr.com> | Wed Feb 16 02:39:42 2011 +0000| [cdeba2de821df504ca799c103e48f3e9a92241f9] | committer: Michael Niedermayer

x86: check for AVX support

This adds configure and runtime checks for AVX support on x86 CPUs.

Signed-off-by: Mans Rullgard <mans at mansr.com>
(cherry picked from commit 87f1355f9b4fc11414d0e6a91404203c2745f89f)

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=cdeba2de821df504ca799c103e48f3e9a92241f9
---

 configure           |    4 ++++
 libavutil/cpu.c     |    3 ++-
 libavutil/cpu.h     |    1 +
 libavutil/x86/cpu.c |   12 ++++++++++++
 4 files changed, 19 insertions(+), 1 deletions(-)

diff --git a/configure b/configure
index 5adad70..9c8d005 100755
--- a/configure
+++ b/configure
@@ -222,6 +222,7 @@ Advanced options (experts only):
   --disable-mmx2           disable MMX2 optimizations
   --disable-sse            disable SSE optimizations
   --disable-ssse3          disable SSSE3 optimizations
+  --disable-avx            disable AVX optimizations
   --disable-armv5te        disable armv5te optimizations
   --disable-armv6          disable armv6 optimizations
   --disable-armv6t2        disable armv6t2 optimizations
@@ -975,6 +976,7 @@ ARCH_EXT_LIST='
     armv6
     armv6t2
     armvfp
+    avx
     iwmmxt
     mmi
     mmx
@@ -1183,6 +1185,7 @@ mmx_deps="x86"
 mmx2_deps="mmx"
 sse_deps="mmx"
 ssse3_deps="sse"
+avx_deps="ssse3"
 
 aligned_stack_if_any="ppc x86"
 fast_64bit_if_any="alpha ia64 mips64 parisc64 ppc64 sparc64 x86_64"
@@ -2677,6 +2680,7 @@ EOF
 
         check_yasm "pextrd [eax], xmm0, 1" && enable yasm ||
             die "yasm not found, use --disable-yasm for a crippled build"
+        check_yasm "vpaddw xmm0, xmm0, xmm0" || disable avx
     fi
 
     case "$cpu" in
diff --git a/libavutil/cpu.c b/libavutil/cpu.c
index 1e03491..3459ce4 100644
--- a/libavutil/cpu.c
+++ b/libavutil/cpu.c
@@ -44,7 +44,7 @@ int main(void)
     int cpu_flags = av_get_cpu_flags();
 
     printf("cpu_flags = 0x%08X\n", cpu_flags);
-    printf("cpu_flags = %s%s%s%s%s%s%s%s%s%s%s%s\n",
+    printf("cpu_flags = %s%s%s%s%s%s%s%s%s%s%s%s%s\n",
 #if   ARCH_ARM
            cpu_flags & AV_CPU_FLAG_IWMMXT   ? "IWMMXT "     : "",
 #elif ARCH_PPC
@@ -60,6 +60,7 @@ int main(void)
            cpu_flags & AV_CPU_FLAG_SSSE3    ? "SSSE3 "      : "",
            cpu_flags & AV_CPU_FLAG_SSE4     ? "SSE4.1 "     : "",
            cpu_flags & AV_CPU_FLAG_SSE42    ? "SSE4.2 "     : "",
+           cpu_flags & AV_CPU_FLAG_AVX      ? "AVX "        : "",
            cpu_flags & AV_CPU_FLAG_3DNOW    ? "3DNow "      : "",
            cpu_flags & AV_CPU_FLAG_3DNOWEXT ? "3DNowExt "   : "");
 #endif
diff --git a/libavutil/cpu.h b/libavutil/cpu.h
index 71cc265..d60e062 100644
--- a/libavutil/cpu.h
+++ b/libavutil/cpu.h
@@ -36,6 +36,7 @@
 #define AV_CPU_FLAG_SSSE3        0x0080 ///< Conroe SSSE3 functions
 #define AV_CPU_FLAG_SSE4         0x0100 ///< Penryn SSE4.1 functions
 #define AV_CPU_FLAG_SSE42        0x0200 ///< Nehalem SSE4.2 functions
+#define AV_CPU_FLAG_AVX          0x4000 ///< AVX functions: requires OS support even if YMM registers aren't used
 #define AV_CPU_FLAG_IWMMXT       0x0100 ///< XScale IWMMXT
 #define AV_CPU_FLAG_ALTIVEC      0x0001 ///< standard
 
diff --git a/libavutil/x86/cpu.c b/libavutil/x86/cpu.c
index 6fa4a46..2caac2f 100644
--- a/libavutil/x86/cpu.c
+++ b/libavutil/x86/cpu.c
@@ -35,6 +35,9 @@
            "=c" (ecx), "=d" (edx)\
          : "0" (index));
 
+#define xgetbv(index,eax,edx)                                   \
+    __asm__ ("xgetbv" : "=a"(eax), "=d"(edx) : "c" (index))
+
 /* Function to test if multimedia instructions are supported...  */
 int ff_get_cpu_flags_x86(void)
 {
@@ -93,6 +96,15 @@ int ff_get_cpu_flags_x86(void)
             rval |= AV_CPU_FLAG_SSE4;
         if (ecx & 0x00100000 )
             rval |= AV_CPU_FLAG_SSE42;
+#if HAVE_AVX
+        /* Check OXSAVE and AVX bits */
+        if ((ecx & 0x18000000) == 0x18000000) {
+            /* Check for OS support */
+            xgetbv(0, eax, edx);
+            if ((eax & 0x6) == 0x6)
+                rval |= AV_CPU_FLAG_AVX;
+        }
+#endif
 #endif
                   ;
     }