libvpx: Pull from upstream

source/libvpx/vp9/common/x86/vp9_subpixel_8t_intrin_avx2.c

 /*
  *  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
  *
  *  Use of this source code is governed by a BSD-style license
  *  that can be found in the LICENSE file in the root of the source
  *  tree. An additional intellectual property rights grant can be found
  *  in the file PATENTS.  All contributing project authors may
  *  be found in the AUTHORS file in the root of the source tree.
  */
 
@@ skipping 14 matching lines @@
 DECLARE_ALIGNED(32, static const uint8_t, filt3_global_avx2[32]) = {
   4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9, 10, 10, 11, 11, 12,
   4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9, 10, 10, 11, 11, 12
 };
 
 DECLARE_ALIGNED(32, static const uint8_t, filt4_global_avx2[32]) = {
   6, 7, 7, 8, 8, 9, 9, 10, 10, 11, 11, 12, 12, 13, 13, 14,
   6, 7, 7, 8, 8, 9, 9, 10, 10, 11, 11, 12, 12, 13, 13, 14
 };
 
+#if defined(__clang__)
+# if __clang_major__ < 3 || (__clang_major__ == 3 && __clang_minor__ <= 3) || \
+      (defined(__APPLE__) && __clang_major__ == 5 && __clang_minor__ == 0)
+#  define MM256_BROADCASTSI128_SI256(x) \
+       _mm_broadcastsi128_si256((__m128i const *)&(x))
+# else  // clang > 3.3, and not 5.0 on macosx.
+#  define MM256_BROADCASTSI128_SI256(x) _mm256_broadcastsi128_si256(x)
+# endif  // clang <= 3.3
+#elif defined(__GNUC__)
+# if __GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ <= 6)
+#  define MM256_BROADCASTSI128_SI256(x) \
+       _mm_broadcastsi128_si256((__m128i const *)&(x))
+# elif __GNUC__ == 4 && __GNUC_MINOR__ == 7
+#  define MM256_BROADCASTSI128_SI256(x) _mm_broadcastsi128_si256(x)
+# else  // gcc > 4.7
+#  define MM256_BROADCASTSI128_SI256(x) _mm256_broadcastsi128_si256(x)
+# endif  // gcc <= 4.6
+#else  // !(gcc || clang)
+# define MM256_BROADCASTSI128_SI256(x) _mm256_broadcastsi128_si256(x)
+#endif  // __clang__
+
 void vp9_filter_block1d16_h8_avx2(unsigned char *src_ptr,
                                   unsigned int src_pixels_per_line,
                                   unsigned char *output_ptr,
                                   unsigned int  output_pitch,
                                   unsigned int  output_height,
                                   int16_t *filter) {
   __m128i filtersReg;
   __m256i addFilterReg64, filt1Reg, filt2Reg, filt3Reg, filt4Reg;
   __m256i firstFilters, secondFilters, thirdFilters, forthFilters;
   __m256i srcRegFilt32b1_1, srcRegFilt32b2_1, srcRegFilt32b2, srcRegFilt32b3;
   __m256i srcReg32b1, srcReg32b2, filtersReg32;
   unsigned int i;
   unsigned int src_stride, dst_stride;
 
   // create a register with 0,64,0,64,0,64,0,64,0,64,0,64,0,64,0,64
   addFilterReg64 = _mm256_set1_epi32((int)0x0400040u);
   filtersReg = _mm_loadu_si128((__m128i *)filter);
   // converting the 16 bit (short) to 8 bit (byte) and have the same data
   // in both lanes of 128 bit register.
   filtersReg =_mm_packs_epi16(filtersReg, filtersReg);
   // have the same data in both lanes of a 256 bit register
-#if defined (__GNUC__)
-#if ( __GNUC__ < 4 || (__GNUC__ == 4 && \
-(__GNUC_MINOR__ < 6 || (__GNUC_MINOR__ == 6 && __GNUC_PATCHLEVEL__ > 0))))
-  filtersReg32 = _mm_broadcastsi128_si256((__m128i const *)&filtersReg);
-#elif(__GNUC__ == 4 && (__GNUC_MINOR__ == 7 && __GNUC_PATCHLEVEL__ > 0))
-  filtersReg32 = _mm_broadcastsi128_si256(filtersReg);
-#else
-  filtersReg32 = _mm256_broadcastsi128_si256(filtersReg);
-#endif
-#else
-  filtersReg32 = _mm256_broadcastsi128_si256(filtersReg);
-#endif
+  filtersReg32 = MM256_BROADCASTSI128_SI256(filtersReg);
 
   // duplicate only the first 16 bits (first and second byte)
   // across 256 bit register
   firstFilters = _mm256_shuffle_epi8(filtersReg32,
                  _mm256_set1_epi16(0x100u));
   // duplicate only the second 16 bits (third and forth byte)
   // across 256 bit register
   secondFilters = _mm256_shuffle_epi8(filtersReg32,
                   _mm256_set1_epi16(0x302u));
   // duplicate only the third 16 bits (fifth and sixth byte)
@@ skipping 224 matching lines @@
   unsigned int i;
   unsigned int src_stride, dst_stride;
 
   // create a register with 0,64,0,64,0,64,0,64,0,64,0,64,0,64,0,64
   addFilterReg64 = _mm256_set1_epi32((int)0x0400040u);
   filtersReg = _mm_loadu_si128((__m128i *)filter);
   // converting the 16 bit (short) to  8 bit (byte) and have the
   // same data in both lanes of 128 bit register.
   filtersReg =_mm_packs_epi16(filtersReg, filtersReg);
   // have the same data in both lanes of a 256 bit register
-#if defined (__GNUC__)
-#if ( __GNUC__ < 4 || (__GNUC__ == 4 && \
-(__GNUC_MINOR__ < 6 || (__GNUC_MINOR__ == 6 && __GNUC_PATCHLEVEL__ > 0))))
-  filtersReg32 = _mm_broadcastsi128_si256((__m128i const *)&filtersReg);
-#elif(__GNUC__ == 4 && (__GNUC_MINOR__ == 7 && __GNUC_PATCHLEVEL__ > 0))
-  filtersReg32 = _mm_broadcastsi128_si256(filtersReg);
-#else
-  filtersReg32 = _mm256_broadcastsi128_si256(filtersReg);
-#endif
-#else
-  filtersReg32 = _mm256_broadcastsi128_si256(filtersReg);
-#endif
+  filtersReg32 = MM256_BROADCASTSI128_SI256(filtersReg);
 
   // duplicate only the first 16 bits (first and second byte)
   // across 256 bit register
   firstFilters = _mm256_shuffle_epi8(filtersReg32,
                  _mm256_set1_epi16(0x100u));
   // duplicate only the second 16 bits (third and forth byte)
   // across 256 bit register
   secondFilters = _mm256_shuffle_epi8(filtersReg32,
                   _mm256_set1_epi16(0x302u));
   // duplicate only the third 16 bits (fifth and sixth byte)
@@ skipping 202 matching lines @@
 
     // shrink to 8 bit each 16 bits, the first lane contain the first
     // convolve result and the second lane contain the second convolve
     // result
     srcRegFilt1 = _mm_packus_epi16(srcRegFilt1, srcRegFilt3);
 
     // save 16 bytes
     _mm_store_si128((__m128i*)output_ptr, srcRegFilt1);
   }
 }