shithub: libvpx

Info  •  Files  •  Log  •  Branches

Remove redundant loads on 1d8_v8 filter.

This CL showed a modest gain in performance on some systems.

Change-Id: Iad636a89a1a9804ab7a0dea302bf2c6a4d1653a4

--- a/vp9/common/x86/vp9_subpixel_8t_intrin_ssse3.c

+++ b/vp9/common/x86/vp9_subpixel_8t_intrin_ssse3.c

@@ -312,9 +312,11 @@

                                          unsigned int out_pitch,

                                          unsigned int output_height,

                                          int16_t *filter) {

-  __m128i addFilterReg64, filtersReg, minReg, srcRegFilt6;

+  __m128i addFilterReg64, filtersReg, minReg;

   __m128i firstFilters, secondFilters, thirdFilters, forthFilters;

-  __m128i srcRegFilt1, srcRegFilt2, srcRegFilt3, srcRegFilt4, srcRegFilt5;

+  __m128i srcRegFilt1, srcRegFilt2, srcRegFilt3, srcRegFilt5;

+  __m128i srcReg1, srcReg2, srcReg3, srcReg4, srcReg5, srcReg6, srcReg7;

+  __m128i srcReg8;

   unsigned int i;

   // create a register with 0,64,0,64,0,64,0,64,0,64,0,64,0,64,0,64

@@ -333,27 +335,26 @@

   // duplicate only the forth 16 bits in the filter

   forthFilters = _mm_shuffle_epi8(filtersReg, _mm_set1_epi16(0x706u));

+  // load the first 7 rows of 8 bytes

+  srcReg1 = _mm_loadl_epi64((__m128i *)&src_ptr[0]);

+  srcReg2 = _mm_loadl_epi64((__m128i *)&(src_ptr + src_pitch)[0]);

+  srcReg3 = _mm_loadl_epi64((__m128i *)&(src_ptr + src_pitch * 2)[0]);

+  srcReg4 = _mm_loadl_epi64((__m128i *)&(src_ptr + src_pitch * 3)[0]);

+  srcReg5 = _mm_loadl_epi64((__m128i *)&(src_ptr + src_pitch * 4)[0]);

+  srcReg6 = _mm_loadl_epi64((__m128i *)&(src_ptr + src_pitch * 5)[0]);

+  srcReg7 = _mm_loadl_epi64((__m128i *)&(src_ptr + src_pitch * 6)[0]);

+

   for (i = 0; i < output_height; i++) {

-    // load the first 8 bytes

-    srcRegFilt1 = _mm_loadl_epi64((__m128i *)&src_ptr[0]);

-    // load the next 8 bytes in stride of src_pitch

-    srcRegFilt2 = _mm_loadl_epi64((__m128i *)&(src_ptr+src_pitch)[0]);

-    srcRegFilt3 = _mm_loadl_epi64((__m128i *)&(src_ptr+src_pitch*2)[0]);

-    srcRegFilt4 = _mm_loadl_epi64((__m128i *)&(src_ptr+src_pitch*3)[0]);

+    // load the last 8 bytes

+    srcReg8 = _mm_loadl_epi64((__m128i *)&(src_ptr + src_pitch * 7)[0]);

     // merge the result together

-    srcRegFilt1 = _mm_unpacklo_epi8(srcRegFilt1, srcRegFilt2);

-    srcRegFilt3 = _mm_unpacklo_epi8(srcRegFilt3, srcRegFilt4);

+    srcRegFilt1 = _mm_unpacklo_epi8(srcReg1, srcReg2);

+    srcRegFilt3 = _mm_unpacklo_epi8(srcReg3, srcReg4);

-    // load the next 8 bytes in stride of src_pitch

-    srcRegFilt2 = _mm_loadl_epi64((__m128i *)&(src_ptr+src_pitch*4)[0]);

-    srcRegFilt4 = _mm_loadl_epi64((__m128i *)&(src_ptr+src_pitch*5)[0]);

-    srcRegFilt5 = _mm_loadl_epi64((__m128i *)&(src_ptr+src_pitch*6)[0]);

-    srcRegFilt6 = _mm_loadl_epi64((__m128i *)&(src_ptr+src_pitch*7)[0]);

-

     // merge the result together

-    srcRegFilt2 = _mm_unpacklo_epi8(srcRegFilt2, srcRegFilt4);

-    srcRegFilt5 = _mm_unpacklo_epi8(srcRegFilt5, srcRegFilt6);

+    srcRegFilt2 = _mm_unpacklo_epi8(srcReg5, srcReg6);

+    srcRegFilt5 = _mm_unpacklo_epi8(srcReg7, srcReg8);

     // multiply 2 adjacent elements with the filter and add the result

     srcRegFilt1 = _mm_maddubs_epi16(srcRegFilt1, firstFilters);

@@ -376,6 +377,15 @@

     srcRegFilt1 = _mm_packus_epi16(srcRegFilt1, srcRegFilt1);

     src_ptr+=src_pitch;

+

+    // shift down a row

+    srcReg1 = srcReg2;

+    srcReg2 = srcReg3;

+    srcReg3 = srcReg4;

+    srcReg4 = srcReg5;

+    srcReg5 = srcReg6;

+    srcReg6 = srcReg7;

+    srcReg7 = srcReg8;

     // save only 8 bytes convolve result

     _mm_storel_epi64((__m128i*)&output_ptr[0], srcRegFilt1);