[vlc-devel] [PATCH] startcode_helper: ensure dependencies between asm blocks

[Francois Cartegnie]

uses attribute checking and vector declarations.
verified with clang and gcc

---
 modules/packetizer/startcode_helper.h | 29 +++++++++++++++++++--------
 1 file changed, 21 insertions(+), 8 deletions(-)

diff --git a/modules/packetizer/startcode_helper.h b/modules/packetizer/startcode_helper.h
index 2b61e5cf98..32296f33e8 100644
--- a/modules/packetizer/startcode_helper.h
+++ b/modules/packetizer/startcode_helper.h
@@ -23,7 +23,11 @@
 #include <vlc_cpu.h>
 
 #if !defined(CAN_COMPILE_SSE2) && defined(HAVE_SSE2_INTRINSICS)
-   #include <emmintrin.h>
+#  include <emmintrin.h>
+#endif
+
+#if __has_attribute(__vector_size__)
+typedef unsigned char v16qu __attribute__((__vector_size__(16)));
 #endif
 
 /* Looks up efficiently for an AnnexB startcode 0x00 0x00 0x01
@@ -63,26 +67,35 @@ static inline const uint8_t * startcode_FindAnnexB_SSE2( const uint8_t *p, const
     alignedend = end - ((intptr_t) end & 15);
     if( alignedend > p )
     {
-#ifdef CAN_COMPILE_SSE2
-        asm volatile(
-            "pxor   %%xmm1, %%xmm1\n"
-            ::: "xmm1"
-        );
+#if defined(CAN_COMPILE_SSE2) && __has_attribute(__vector_size__)
+        const v16qu zeros = { 0 };
 #else
-        __m128i zeros = _mm_set1_epi8( 0x00 );
+        const __m128i zeros = _mm_set1_epi8( 0x00 );
 #endif
         for( ; p < alignedend; p += 16)
         {
             uint32_t match;
 #ifdef CAN_COMPILE_SSE2
+#  if __has_attribute(__vector_size__)
+            asm volatile(
+                "movdqa   0(%[v]),   %%xmm0\n"
+                "pcmpeqb %[czero],   %%xmm0\n"
+                "pmovmskb  %%xmm0,   %[match]\n"
+                : [match]"=r"(match)
+                : [v]"r"(p), [czero]"x"(zeros)
+                : "xmm0"
+            );
+#  else
             asm volatile(
                 "movdqa   0(%[v]),   %%xmm0\n"
+                "pxor      %%xmm1,   %%xmm1\n"
                 "pcmpeqb   %%xmm1,   %%xmm0\n"
                 "pmovmskb  %%xmm0,   %[match]\n"
                 : [match]"=r"(match)
                 : [v]"r"(p)
-                : "xmm0"
+                : "xmm0", "xmm1"
             );
+#  endif
 #else
             __m128i v = _mm_load_si128((__m128i*)p);
             __m128i res = _mm_cmpeq_epi8( zeros, v );
-- 
2.25.4