shithub: opus

Info  •  Files  •  Log  •  Branches

Adding support for SSE2 and SSSE3

--- a/dnn/vec.h

+++ b/dnn/vec.h

@@ -35,7 +35,7 @@

 #include "arch.h"

-#if defined(__AVX__) || defined(__SSE4_1__)

+#if defined(__AVX__) || defined(__SSE2__)

 #include "vec_avx.h"

 #elif defined(__ARM_NEON__) || defined(__ARM_NEON)

 #include "vec_neon.h"

--- a/dnn/vec_avx.h

+++ b/dnn/vec_avx.h

@@ -34,11 +34,28 @@

 #include <immintrin.h>

-#ifndef DISABLE_DOT_PROD

+/* Use 8-bit dot products unless disabled or if stuck with SSE2. */

+#if (defined(__AVX2__) || defined(__SSSE3__)) && !defined(DISABLE_DOT_PROD)

 #define DOT_PROD

 #define USE_SU_BIAS

+

+#else

+

+#warning "Only SSE and SSE2 are available. On newer machines, enable SSSE3/AVX/AVX2 using -march= to get better performance"

+

 #endif

+

+#ifndef __SSE_4_1__

+static inline __m128 mm_floor_ps(__m128 x) {

+  __m128 half = _mm_set1_ps(0.5);

+  return _mm_cvtepi32_ps(_mm_cvtps_epi32(_mm_sub_ps(x, half)));

+}

+#undef _mm_floor_ps

+#define _mm_floor_ps(x) mm_floor_ps(x)

+#endif

+

+

 /* If we don't have AVX available, emulate what we need with SSE up to 4.1. */

 #ifndef __AVX__

@@ -135,6 +152,7 @@

 static inline __m128 mm256_extractf128_ps(mm256_emu x, int i) {

     return (i==0) ? x.lo : x.hi;

 }

+#undef _mm256_extractf128_ps

 #define _mm256_extractf128_ps(x,i) mm256_extractf128_ps(x,i)

 static inline mm256_emu mm256_insertf128_ps(mm256_emu dst, __m128 src, int i) {

@@ -142,6 +160,7 @@

     else dst.hi = src;

     return dst;

 }

+#undef _mm256_insertf128_ps

 #define _mm256_insertf128_ps(dst,src,i) mm256_insertf128_ps(dst,src,i)

 #endif /* __AVX__ */

-- 

⑨