From 2d58761993f5d25fe3ed54fbcef3bc8f40d4fc89 Mon Sep 17 00:00:00 2001 From: James Zern Date: Fri, 13 Sep 2013 15:11:07 -0700 Subject: [PATCH] Revert "Improved 8t filters" This is incompatible with most toolchains other than gcc. Revert "Deleted #include " This reverts commit 4d018be950ef8b056a7c797a22ee58012443df26. This reverts commit d22a504d11a15dc3eab666859db0046b5a7d75c5. Change-Id: I1751dc6831f4395ee064e6748281418e967e1dcf --- vp9/common/x86/vp9_asm_stubs.c | 94 ++----- vp9/common/x86/vp9_subpixel_8t_intrin_ssse3.c | 375 -------------------------- vp9/vp9_common.mk | 1 - 3 files changed, 24 insertions(+), 446 deletions(-) delete mode 100644 vp9/common/x86/vp9_subpixel_8t_intrin_ssse3.c diff --git a/vp9/common/x86/vp9_asm_stubs.c b/vp9/common/x86/vp9_asm_stubs.c index f9f00b9..3f1c198 100644 --- a/vp9/common/x86/vp9_asm_stubs.c +++ b/vp9/common/x86/vp9_asm_stubs.c @@ -13,7 +13,6 @@ #include "./vpx_config.h" #include "./vp9_rtcd.h" #include "vpx_ports/mem.h" - /////////////////////////////////////////////////////////////////////////// // the mmx function that does the bilinear filtering and var calculation // // int one pass // @@ -37,8 +36,6 @@ DECLARE_ALIGNED(16, const short, vp9_bilinear_filters_mmx[16][8]) = { { 8, 8, 8, 8, 120, 120, 120, 120 } }; - -int num_func_entry = 0; #if HAVE_SSSE3 void vp9_filter_block1d16_v8_ssse3(const unsigned char *src_ptr, const unsigned int src_pitch, @@ -47,13 +44,6 @@ void vp9_filter_block1d16_v8_ssse3(const unsigned char *src_ptr, unsigned int output_height, const short *filter); -void vp9_filter_block1d16_v8_intrin_ssse3(const unsigned char *src_ptr, - const unsigned int src_pitch, - unsigned char *output_ptr, - unsigned int out_pitch, - unsigned int output_height, - const short *filter); - void vp9_filter_block1d16_h8_ssse3(const unsigned char *src_ptr, const unsigned int src_pitch, unsigned char *output_ptr, @@ -61,13 +51,6 @@ void vp9_filter_block1d16_h8_ssse3(const unsigned char *src_ptr, unsigned int output_height, const short *filter); -void vp9_filter_block1d16_h8_intrin_ssse3(const unsigned char *src_ptr, - const unsigned int src_pitch, - unsigned char *output_ptr, - unsigned int out_pitch, - unsigned int output_height, - const short *filter); - void vp9_filter_block1d8_v8_ssse3(const unsigned char *src_ptr, const unsigned int src_pitch, unsigned char *output_ptr, @@ -75,13 +58,6 @@ void vp9_filter_block1d8_v8_ssse3(const unsigned char *src_ptr, unsigned int output_height, const short *filter); -void vp9_filter_block1d8_v8_intrin_ssse3(const unsigned char *src_ptr, - const unsigned int src_pitch, - unsigned char *output_ptr, - unsigned int out_pitch, - unsigned int output_height, - const short *filter); - void vp9_filter_block1d8_h8_ssse3(const unsigned char *src_ptr, const unsigned int src_pitch, unsigned char *output_ptr, @@ -89,13 +65,6 @@ void vp9_filter_block1d8_h8_ssse3(const unsigned char *src_ptr, unsigned int output_height, const short *filter); -void vp9_filter_block1d8_h8_intrin_ssse3(const unsigned char *src_ptr, - const unsigned int src_pitch, - unsigned char *output_ptr, - unsigned int out_pitch, - unsigned int output_height, - const short *filter); - void vp9_filter_block1d4_v8_ssse3(const unsigned char *src_ptr, const unsigned int src_pitch, unsigned char *output_ptr, @@ -103,14 +72,6 @@ void vp9_filter_block1d4_v8_ssse3(const unsigned char *src_ptr, unsigned int output_height, const short *filter); -void vp9_filter_block1d4_v8_intrin_ssse3(const unsigned char *src_ptr, - const unsigned int src_pitch, - unsigned char *output_ptr, - unsigned int out_pitch, - unsigned int output_height, - const short *filter); - - void vp9_filter_block1d4_h8_ssse3(const unsigned char *src_ptr, const unsigned int src_pitch, unsigned char *output_ptr, @@ -118,13 +79,6 @@ void vp9_filter_block1d4_h8_ssse3(const unsigned char *src_ptr, unsigned int output_height, const short *filter); -void vp9_filter_block1d4_h8_intrin_ssse3(const unsigned char *src_ptr, - const unsigned int src_pitch, - unsigned char *output_ptr, - unsigned int out_pitch, - unsigned int output_height, - const short *filter); - void vp9_filter_block1d16_v8_avg_ssse3(const unsigned char *src_ptr, const unsigned int src_pitch, unsigned char *output_ptr, @@ -175,7 +129,7 @@ void vp9_convolve8_horiz_ssse3(const uint8_t *src, ptrdiff_t src_stride, /* Ensure the filter can be compressed to int16_t. */ if (x_step_q4 == 16 && filter_x[3] != 128) { while (w >= 16) { - vp9_filter_block1d16_h8_intrin_ssse3(src, src_stride, + vp9_filter_block1d16_h8_ssse3(src, src_stride, dst, dst_stride, h, filter_x); src += 16; @@ -183,7 +137,7 @@ void vp9_convolve8_horiz_ssse3(const uint8_t *src, ptrdiff_t src_stride, w -= 16; } while (w >= 8) { - vp9_filter_block1d8_h8_intrin_ssse3(src, src_stride, + vp9_filter_block1d8_h8_ssse3(src, src_stride, dst, dst_stride, h, filter_x); src += 8; @@ -191,7 +145,7 @@ void vp9_convolve8_horiz_ssse3(const uint8_t *src, ptrdiff_t src_stride, w -= 8; } while (w >= 4) { - vp9_filter_block1d4_h8_intrin_ssse3(src, src_stride, + vp9_filter_block1d4_h8_ssse3(src, src_stride, dst, dst_stride, h, filter_x); src += 4; @@ -213,25 +167,25 @@ void vp9_convolve8_vert_ssse3(const uint8_t *src, ptrdiff_t src_stride, int w, int h) { if (y_step_q4 == 16 && filter_y[3] != 128) { while (w >= 16) { - vp9_filter_block1d16_v8_intrin_ssse3(src - src_stride * 3, src_stride, - dst, dst_stride, - h, filter_y); + vp9_filter_block1d16_v8_ssse3(src - src_stride * 3, src_stride, + dst, dst_stride, + h, filter_y); src += 16; dst += 16; w -= 16; } while (w >= 8) { - vp9_filter_block1d8_v8_intrin_ssse3(src - src_stride * 3, src_stride, - dst, dst_stride, - h, filter_y); + vp9_filter_block1d8_v8_ssse3(src - src_stride * 3, src_stride, + dst, dst_stride, + h, filter_y); src += 8; dst += 8; w -= 8; } while (w >= 4) { - vp9_filter_block1d4_v8_intrin_ssse3(src - src_stride * 3, src_stride, - dst, dst_stride, - h, filter_y); + vp9_filter_block1d4_v8_ssse3(src - src_stride * 3, src_stride, + dst, dst_stride, + h, filter_y); src += 4; dst += 4; w -= 4; @@ -252,24 +206,24 @@ void vp9_convolve8_avg_horiz_ssse3(const uint8_t *src, ptrdiff_t src_stride, if (x_step_q4 == 16 && filter_x[3] != 128) { while (w >= 16) { vp9_filter_block1d16_h8_avg_ssse3(src, src_stride, - dst, dst_stride, - h, filter_x); + dst, dst_stride, + h, filter_x); src += 16; dst += 16; w -= 16; } while (w >= 8) { vp9_filter_block1d8_h8_avg_ssse3(src, src_stride, - dst, dst_stride, - h, filter_x); + dst, dst_stride, + h, filter_x); src += 8; dst += 8; w -= 8; } while (w >= 4) { vp9_filter_block1d4_h8_avg_ssse3(src, src_stride, - dst, dst_stride, - h, filter_x); + dst, dst_stride, + h, filter_x); src += 4; dst += 4; w -= 4; @@ -290,24 +244,24 @@ void vp9_convolve8_avg_vert_ssse3(const uint8_t *src, ptrdiff_t src_stride, if (y_step_q4 == 16 && filter_y[3] != 128) { while (w >= 16) { vp9_filter_block1d16_v8_avg_ssse3(src - src_stride * 3, src_stride, - dst, dst_stride, - h, filter_y); + dst, dst_stride, + h, filter_y); src += 16; dst += 16; w -= 16; } while (w >= 8) { vp9_filter_block1d8_v8_avg_ssse3(src - src_stride * 3, src_stride, - dst, dst_stride, - h, filter_y); + dst, dst_stride, + h, filter_y); src += 8; dst += 8; w -= 8; } while (w >= 4) { vp9_filter_block1d4_v8_avg_ssse3(src - src_stride * 3, src_stride, - dst, dst_stride, - h, filter_y); + dst, dst_stride, + h, filter_y); src += 4; dst += 4; w -= 4; diff --git a/vp9/common/x86/vp9_subpixel_8t_intrin_ssse3.c b/vp9/common/x86/vp9_subpixel_8t_intrin_ssse3.c deleted file mode 100644 index 79895c7..0000000 --- a/vp9/common/x86/vp9_subpixel_8t_intrin_ssse3.c +++ /dev/null @@ -1,375 +0,0 @@ -#include -#include -#include "vpx_ports/mem.h" -#include "vpx_ports/emmintrin_compat.h" -#pragma GCC push_options -#pragma GCC optimize("unroll-loops") - -DECLARE_ALIGNED(16, const unsigned char, filt1_4_h8[16]) = { - 0, 1, 1, 2, 2, 3, 3, 4, 2, 3, 3, 4, 4, 5, 5, 6 -}; -DECLARE_ALIGNED(16, const unsigned char, filt2_4_h8[16]) = { - 4, 5, 5, 6, 6, 7, 7, 8, 6, 7, 7, 8, 8, 9, 9, 10 -}; -DECLARE_ALIGNED(16, const unsigned char, filt1_global[16]) = { - 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8 -}; -DECLARE_ALIGNED(16, const unsigned char, filt2_global[16]) = { - 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9, 10 -}; -DECLARE_ALIGNED(16, const unsigned char, filt3_global[16]) = { - 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9, 10, 10, 11, 11, 12 -}; -DECLARE_ALIGNED(16, const unsigned char, filt4_global[16]) = { - 6, 7, 7, 8, 8, 9, 9, 10, 10, 11, 11, 12, 12, 13, 13, 14 -}; - -void vp9_filter_block1d4_h8_intrin_ssse3(unsigned char *src_ptr, - unsigned int src_pixels_per_line, - unsigned char *output_ptr, - unsigned int output_pitch, - unsigned int output_height, - short *filter) { - __m128i addFilterReg127, filtersReg, firstFilters, secondFilters, - thirdFilters, forthFilters, srcReg, srcRegFilt1, srcRegFilt2, - srcRegFilt3, srcRegFilt4; - unsigned int i; - - addFilterReg127 = _mm_shuffle_epi32(_mm_cvtsi32_si128((int)0x0400040u), 0); - filtersReg = _mm_loadu_si128((__m128i *)filter); - filtersReg = _mm_packs_epi16(filtersReg, filtersReg); - firstFilters = _mm_shufflelo_epi16(filtersReg, 0); - secondFilters = _mm_shufflelo_epi16(filtersReg, 0xAAu); - firstFilters = _mm_shufflehi_epi16(firstFilters, 0x55u); - secondFilters = _mm_shufflehi_epi16(secondFilters, 0xFFu); - thirdFilters =_mm_load_si128((__m128i const *)filt1_4_h8); - forthFilters = _mm_load_si128((__m128i const *)filt2_4_h8); - - for (i = 0; i < output_height; i++) { - srcReg = _mm_loadu_si128((__m128i *)(src_ptr-3)); - srcRegFilt1= _mm_shuffle_epi8(srcReg, thirdFilters); - srcRegFilt2= _mm_shuffle_epi8(srcReg, forthFilters); - srcRegFilt1 = _mm_maddubs_epi16(srcRegFilt1, firstFilters); - srcRegFilt2 = _mm_maddubs_epi16(srcRegFilt2, secondFilters); - srcRegFilt3 = _mm_srli_si128(srcRegFilt1, 8); - srcRegFilt4 = _mm_srli_si128(srcRegFilt2, 8); - srcRegFilt1 = _mm_adds_epi16(srcRegFilt1, srcRegFilt3); - srcRegFilt1 = _mm_adds_epi16(srcRegFilt1, srcRegFilt4); - srcRegFilt1 = _mm_adds_epi16(srcRegFilt1, srcRegFilt2); - srcRegFilt1 = _mm_adds_epi16(srcRegFilt1, addFilterReg127); - srcRegFilt1 = _mm_srai_epi16(srcRegFilt1, 7); - srcRegFilt1 = _mm_packus_epi16(srcRegFilt1, srcRegFilt1); - src_ptr += src_pixels_per_line; - *((int*)&output_ptr[0])= _mm_cvtsi128_si32(srcRegFilt1); - output_ptr += output_pitch; - } -} - -void vp9_filter_block1d8_h8_intrin_ssse3(unsigned char *src_ptr, - unsigned int src_pixels_per_line, - unsigned char *output_ptr, - unsigned int output_pitch, - unsigned int output_height, - short *filter) { - __m128i addFilterReg127, filtersReg, firstFilters, secondFilters, - thirdFilters, forthFilters, srcReg, srcRegFilt1, srcRegFilt2, - srcRegFilt3, srcRegFilt4; - __m128i filt1Reg, filt2Reg, filt3Reg, filt4Reg; - unsigned int i; - - addFilterReg127 = _mm_shuffle_epi32(_mm_cvtsi32_si128((int)0x0400040u), 0); - filtersReg = _mm_loadu_si128((__m128i *)filter); - filtersReg =_mm_packs_epi16(filtersReg, filtersReg); - - firstFilters = _mm_shuffle_epi8(filtersReg, _mm_set1_epi16(0x100u)); - secondFilters = _mm_shuffle_epi8(filtersReg, _mm_set1_epi16(0x302u)); - thirdFilters = _mm_shuffle_epi8(filtersReg, _mm_set1_epi16(0x504u)); - forthFilters = _mm_shuffle_epi8(filtersReg, _mm_set1_epi16(0x706u)); - - filt1Reg = _mm_load_si128((__m128i const *)filt1_global); - filt2Reg = _mm_load_si128((__m128i const *)filt2_global); - filt3Reg = _mm_load_si128((__m128i const *)filt3_global); - filt4Reg = _mm_load_si128((__m128i const *)filt4_global); - - for (i = 0 ; i < output_height ; i++) { - srcReg = _mm_loadu_si128((__m128i *)(src_ptr-3)); - srcRegFilt1= _mm_shuffle_epi8(srcReg, filt1Reg); - srcRegFilt2= _mm_shuffle_epi8(srcReg, filt2Reg); - srcRegFilt1 = _mm_maddubs_epi16(srcRegFilt1, firstFilters); - srcRegFilt2 = _mm_maddubs_epi16(srcRegFilt2, secondFilters); - srcRegFilt3= _mm_shuffle_epi8(srcReg, filt3Reg); - srcRegFilt4= _mm_shuffle_epi8(srcReg, filt4Reg); - srcRegFilt3 = _mm_maddubs_epi16(srcRegFilt3, thirdFilters); - srcRegFilt4 = _mm_maddubs_epi16(srcRegFilt4, forthFilters); - srcRegFilt1 = _mm_adds_epi16(srcRegFilt1, srcRegFilt2); - srcRegFilt1 = _mm_adds_epi16(srcRegFilt1, srcRegFilt4); - srcRegFilt1 = _mm_adds_epi16(srcRegFilt1, srcRegFilt3); - srcRegFilt1 = _mm_adds_epi16(srcRegFilt1, addFilterReg127); - srcRegFilt1 = _mm_srai_epi16(srcRegFilt1, 7); - srcRegFilt1 = _mm_packus_epi16(srcRegFilt1, srcRegFilt1); - src_ptr += src_pixels_per_line; - _mm_storel_epi64((__m128i *)&output_ptr[0], srcRegFilt1); - output_ptr += output_pitch; - } -} - -void vp9_filter_block1d16_h8_intrin_ssse3(unsigned char *src_ptr, - unsigned int src_pixels_per_line, - unsigned char *output_ptr, - unsigned int output_pitch, - unsigned int output_height, - short *filter) { - __m128i addFilterReg127, filtersReg, firstFilters, secondFilters, - thirdFilters, forthFilters, srcReg1, srcReg2, srcRegFilt1_1, - srcRegFilt2_1, srcRegFilt2, srcRegFilt3; - __m128i filt1Reg, filt2Reg, filt3Reg, filt4Reg; - unsigned int i; - - addFilterReg127 = _mm_shuffle_epi32(_mm_cvtsi32_si128((int)0x0400040u), 0); - filtersReg = _mm_loadu_si128((__m128i *)filter); - filtersReg =_mm_packs_epi16(filtersReg, filtersReg); - - firstFilters = _mm_shuffle_epi8(filtersReg, _mm_set1_epi16(0x100u)); - secondFilters = _mm_shuffle_epi8(filtersReg, _mm_set1_epi16(0x302u)); - thirdFilters = _mm_shuffle_epi8(filtersReg, _mm_set1_epi16(0x504u)); - forthFilters = _mm_shuffle_epi8(filtersReg, _mm_set1_epi16(0x706u)); - - filt1Reg = _mm_load_si128((__m128i const *)filt1_global); - filt2Reg = _mm_load_si128((__m128i const *)filt2_global); - filt3Reg = _mm_load_si128((__m128i const *)filt3_global); - filt4Reg = _mm_load_si128((__m128i const *)filt4_global); - - for (i = 0 ; i < output_height ; i++) { - srcReg1 = _mm_loadu_si128((__m128i *)(src_ptr-3)); - srcRegFilt1_1 = _mm_shuffle_epi8(srcReg1, filt1Reg); - srcRegFilt2 = _mm_shuffle_epi8(srcReg1, filt2Reg); - - srcRegFilt1_1 = _mm_maddubs_epi16(srcRegFilt1_1, firstFilters); - srcRegFilt2 = _mm_maddubs_epi16(srcRegFilt2, secondFilters); - - srcRegFilt1_1 = _mm_adds_epi16(srcRegFilt1_1, srcRegFilt2); - - srcRegFilt3 = _mm_shuffle_epi8(srcReg1, filt4Reg); - srcRegFilt2 = _mm_shuffle_epi8(srcReg1, filt3Reg); - - srcRegFilt3 = _mm_maddubs_epi16(srcRegFilt3, forthFilters); - srcRegFilt2 = _mm_maddubs_epi16(srcRegFilt2, thirdFilters); - - srcRegFilt1_1 = _mm_adds_epi16(srcRegFilt1_1, srcRegFilt3); - - srcReg2 = _mm_loadu_si128((__m128i *)(src_ptr+5)); - - srcRegFilt1_1 = _mm_adds_epi16(srcRegFilt1_1, srcRegFilt2); - - srcRegFilt2_1 = _mm_shuffle_epi8(srcReg2, filt1Reg); - srcRegFilt2 = _mm_shuffle_epi8(srcReg2, filt2Reg); - - srcRegFilt2_1 = _mm_maddubs_epi16(srcRegFilt2_1, firstFilters); - srcRegFilt2 = _mm_maddubs_epi16(srcRegFilt2, secondFilters); - - srcRegFilt2_1 = _mm_adds_epi16(srcRegFilt2_1, srcRegFilt2); - - srcRegFilt3 = _mm_shuffle_epi8(srcReg2, filt4Reg); - srcRegFilt2 = _mm_shuffle_epi8(srcReg2, filt3Reg); - - srcRegFilt3 = _mm_maddubs_epi16(srcRegFilt3, forthFilters); - srcRegFilt2 = _mm_maddubs_epi16(srcRegFilt2, thirdFilters); - - srcRegFilt2_1 = _mm_adds_epi16(srcRegFilt2_1, srcRegFilt3); - srcRegFilt2_1 = _mm_adds_epi16(srcRegFilt2_1, srcRegFilt2); - srcRegFilt1_1 = _mm_adds_epi16(srcRegFilt1_1, addFilterReg127); - srcRegFilt2_1 = _mm_adds_epi16(srcRegFilt2_1, addFilterReg127); - srcRegFilt1_1 = _mm_srai_epi16(srcRegFilt1_1, 7); - srcRegFilt2_1 = _mm_srai_epi16(srcRegFilt2_1, 7); - srcRegFilt1_1 = _mm_packus_epi16(srcRegFilt1_1, srcRegFilt2_1); - - src_ptr += src_pixels_per_line; - _mm_store_si128((__m128i*)output_ptr, srcRegFilt1_1); - output_ptr += output_pitch; - } -} - -void vp9_filter_block1d4_v8_intrin_ssse3(unsigned char *src_ptr, - unsigned int src_pitch, - unsigned char *output_ptr, - unsigned int out_pitch, - unsigned int output_height, - short *filter) { - __m128i addFilterReg127, filtersReg, firstFilters, secondFilters, srcRegFilt1, - srcRegFilt2, srcRegFilt3, srcRegFilt4; - unsigned int i; - - addFilterReg127 = _mm_shuffle_epi32(_mm_cvtsi32_si128((int)0x0400040u), 0); - filtersReg = _mm_loadu_si128((__m128i *)filter); - filtersReg =_mm_packs_epi16(filtersReg, filtersReg); - firstFilters = _mm_shufflelo_epi16(filtersReg, 0); - firstFilters = _mm_shufflehi_epi16(firstFilters, 0x55u); - secondFilters = _mm_shufflelo_epi16(filtersReg, 0xAAu); - secondFilters = _mm_shufflehi_epi16(secondFilters, 0xFFu); - - for (i = 0 ; i < output_height ; i++) { - srcRegFilt1 = _mm_cvtsi32_si128(*((int*)&src_ptr[0])); - srcRegFilt2 = _mm_cvtsi32_si128(*((int*)&(src_ptr+src_pitch)[0])); - - srcRegFilt1 = _mm_unpacklo_epi8(srcRegFilt1, srcRegFilt2); - - srcRegFilt2 = _mm_cvtsi32_si128(*((int*)&(src_ptr+src_pitch*2)[0])); - srcRegFilt3 = _mm_cvtsi32_si128(*((int*)&(src_ptr+src_pitch*3)[0])); - - srcRegFilt2 = _mm_unpacklo_epi8(srcRegFilt2, srcRegFilt3); - - srcRegFilt3 = _mm_cvtsi32_si128(*((int*)&(src_ptr+src_pitch*4)[0])); - srcRegFilt4 = _mm_cvtsi32_si128(*((int*)&(src_ptr+src_pitch*5)[0])); - - srcRegFilt3 = _mm_unpacklo_epi8(srcRegFilt3, srcRegFilt4); - srcRegFilt1 = _mm_unpacklo_epi64(srcRegFilt1, srcRegFilt2); - - srcRegFilt4 = _mm_cvtsi32_si128(*((int*)&(src_ptr+src_pitch*6)[0])); - srcRegFilt2 = _mm_cvtsi32_si128(*((int*)&(src_ptr+src_pitch*7)[0])); - - srcRegFilt4 = _mm_unpacklo_epi8(srcRegFilt4, srcRegFilt2); - srcRegFilt3 = _mm_unpacklo_epi64(srcRegFilt3, srcRegFilt4); - - srcRegFilt1 = _mm_maddubs_epi16(srcRegFilt1, firstFilters); - srcRegFilt3 = _mm_maddubs_epi16(srcRegFilt3, secondFilters); - srcRegFilt2 = _mm_srli_si128(srcRegFilt1, 8); - srcRegFilt1 = _mm_adds_epi16(srcRegFilt1, _mm_srli_si128(srcRegFilt3, 8)); - srcRegFilt1 = _mm_adds_epi16(srcRegFilt1, srcRegFilt2); - srcRegFilt1 = _mm_adds_epi16(srcRegFilt1, srcRegFilt3); - srcRegFilt1 = _mm_adds_epi16(srcRegFilt1, addFilterReg127); - srcRegFilt1 = _mm_srai_epi16(srcRegFilt1, 7); - srcRegFilt1 = _mm_packus_epi16(srcRegFilt1, srcRegFilt1); - - src_ptr += src_pitch; - *((int*)&output_ptr[0])= _mm_cvtsi128_si32(srcRegFilt1); - output_ptr += out_pitch; - } -} - -void vp9_filter_block1d8_v8_intrin_ssse3(unsigned char *src_ptr, - unsigned int src_pitch, - unsigned char *output_ptr, - unsigned int out_pitch, - unsigned int output_height, - short *filter) { - __m128i addFilterReg127, filtersReg, firstFilters, secondFilters, - thirdFilters, forthFilters, srcRegFilt1, srcRegFilt2, srcRegFilt3, - srcRegFilt4, srcRegFilt5, srcRegFilt6; - unsigned int i; - - addFilterReg127 = _mm_shuffle_epi32(_mm_cvtsi32_si128((int)0x0400040u), 0); - filtersReg = _mm_loadu_si128((__m128i *)filter); - filtersReg =_mm_packs_epi16(filtersReg, filtersReg); - - firstFilters = _mm_shuffle_epi8(filtersReg, _mm_set1_epi16(0x100u)); - secondFilters = _mm_shuffle_epi8(filtersReg, _mm_set1_epi16(0x302u)); - thirdFilters = _mm_shuffle_epi8(filtersReg, _mm_set1_epi16(0x504u)); - forthFilters = _mm_shuffle_epi8(filtersReg, _mm_set1_epi16(0x706u)); - - for (i = 0 ; i < output_height ; i++) { - srcRegFilt1 = _mm_loadl_epi64((__m128i*)&src_ptr[0]); - 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]); - - srcRegFilt1 = _mm_unpacklo_epi8(srcRegFilt1, srcRegFilt2); - - srcRegFilt3 = _mm_unpacklo_epi8(srcRegFilt3, srcRegFilt4); - - 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]); - - srcRegFilt2 = _mm_unpacklo_epi8(srcRegFilt2, srcRegFilt4); - srcRegFilt5 = _mm_unpacklo_epi8(srcRegFilt5, srcRegFilt6); - srcRegFilt1 = _mm_maddubs_epi16(srcRegFilt1, firstFilters); - srcRegFilt3 = _mm_maddubs_epi16(srcRegFilt3, secondFilters); - srcRegFilt2 = _mm_maddubs_epi16(srcRegFilt2, thirdFilters); - srcRegFilt5 = _mm_maddubs_epi16(srcRegFilt5, forthFilters); - srcRegFilt1 = _mm_adds_epi16(srcRegFilt1, srcRegFilt5); - srcRegFilt1 = _mm_adds_epi16(srcRegFilt1, srcRegFilt3); - srcRegFilt1 = _mm_adds_epi16(srcRegFilt1, srcRegFilt2); - srcRegFilt1 = _mm_adds_epi16(srcRegFilt1, addFilterReg127); - srcRegFilt1 = _mm_srai_epi16(srcRegFilt1, 7); - srcRegFilt1 = _mm_packus_epi16(srcRegFilt1, srcRegFilt1); - - src_ptr += src_pitch; - _mm_storel_epi64((__m128i *)&output_ptr[0], srcRegFilt1); - output_ptr += out_pitch; - } -} - -void vp9_filter_block1d16_v8_intrin_ssse3(unsigned char *src_ptr, - unsigned int src_pitch, - unsigned char *output_ptr, - unsigned int out_pitch, - unsigned int output_height, - short *filter) { - __m128i addFilterReg127, filtersReg, firstFilters, secondFilters, - thirdFilters, forthFilters, srcRegFilt1, srcRegFilt2, srcRegFilt3, - srcRegFilt4, srcRegFilt5, srcRegFilt6; - unsigned int i; - - addFilterReg127 = _mm_shuffle_epi32(_mm_cvtsi32_si128((int)0x0400040u), 0); - filtersReg = _mm_loadu_si128((__m128i *)filter); - filtersReg =_mm_packs_epi16(filtersReg, filtersReg); - - firstFilters = _mm_shuffle_epi8(filtersReg, _mm_set1_epi16(0x100u)); - secondFilters = _mm_shuffle_epi8(filtersReg, _mm_set1_epi16(0x302u)); - thirdFilters = _mm_shuffle_epi8(filtersReg, _mm_set1_epi16(0x504u)); - forthFilters = _mm_shuffle_epi8(filtersReg, _mm_set1_epi16(0x706u)); - - for (i = 0 ; i < output_height ; i++) { - srcRegFilt1 = _mm_loadu_si128((__m128i *)(src_ptr)); - srcRegFilt2 = _mm_loadu_si128((__m128i *)(src_ptr+src_pitch)); - - srcRegFilt3 = _mm_loadu_si128((__m128i *)(src_ptr+src_pitch*6)); - srcRegFilt4 = _mm_loadu_si128((__m128i *)(src_ptr+src_pitch*7)); - - srcRegFilt5 = _mm_unpacklo_epi8(srcRegFilt1, srcRegFilt2); - srcRegFilt6 = _mm_unpacklo_epi8(srcRegFilt3, srcRegFilt4); - srcRegFilt1 = _mm_unpackhi_epi8(srcRegFilt1, srcRegFilt2); - srcRegFilt3 = _mm_unpackhi_epi8(srcRegFilt3, srcRegFilt4); - - srcRegFilt5 = _mm_maddubs_epi16(srcRegFilt5, firstFilters); - srcRegFilt6 = _mm_maddubs_epi16(srcRegFilt6, forthFilters); - srcRegFilt1 = _mm_maddubs_epi16(srcRegFilt1, firstFilters); - srcRegFilt3 = _mm_maddubs_epi16(srcRegFilt3, forthFilters); - - srcRegFilt5 = _mm_adds_epi16(srcRegFilt5, srcRegFilt6); - srcRegFilt1 = _mm_adds_epi16(srcRegFilt1, srcRegFilt3); - - srcRegFilt2 = _mm_loadu_si128((__m128i *)(src_ptr+src_pitch*2)); - srcRegFilt3 = _mm_loadu_si128((__m128i *)(src_ptr+src_pitch*3)); - - srcRegFilt4 = _mm_unpacklo_epi8(srcRegFilt2, srcRegFilt3); - srcRegFilt6 = _mm_unpackhi_epi8(srcRegFilt2, srcRegFilt3); - srcRegFilt4 = _mm_maddubs_epi16(srcRegFilt4, secondFilters); - srcRegFilt6 = _mm_maddubs_epi16(srcRegFilt6, secondFilters); - - srcRegFilt5 = _mm_adds_epi16(srcRegFilt5, srcRegFilt4); - srcRegFilt1 = _mm_adds_epi16(srcRegFilt1, srcRegFilt6); - - srcRegFilt2 = _mm_loadu_si128((__m128i *)(src_ptr+src_pitch*4)); - srcRegFilt3 = _mm_loadu_si128((__m128i *)(src_ptr+src_pitch*5)); - - srcRegFilt4 = _mm_unpacklo_epi8(srcRegFilt2, srcRegFilt3); - srcRegFilt6 = _mm_unpackhi_epi8(srcRegFilt2, srcRegFilt3); - srcRegFilt4 = _mm_maddubs_epi16(srcRegFilt4, thirdFilters); - srcRegFilt6 = _mm_maddubs_epi16(srcRegFilt6, thirdFilters); - srcRegFilt5 = _mm_adds_epi16(srcRegFilt5, srcRegFilt4); - srcRegFilt1 = _mm_adds_epi16(srcRegFilt1, srcRegFilt6); - srcRegFilt5 = _mm_adds_epi16(srcRegFilt5, addFilterReg127); - srcRegFilt1 = _mm_adds_epi16(srcRegFilt1, addFilterReg127); - srcRegFilt5 = _mm_srai_epi16(srcRegFilt5, 7); - srcRegFilt1 = _mm_srai_epi16(srcRegFilt1, 7); - srcRegFilt1 = _mm_packus_epi16(srcRegFilt5, srcRegFilt1); - - src_ptr += src_pitch; - _mm_store_si128((__m128i*)output_ptr, srcRegFilt1); - output_ptr += out_pitch; - } -} diff --git a/vp9/vp9_common.mk b/vp9/vp9_common.mk index 7f2523f..687fb48 100644 --- a/vp9/vp9_common.mk +++ b/vp9/vp9_common.mk @@ -77,7 +77,6 @@ VP9_COMMON_SRCS-$(CONFIG_VP9_POSTPROC) += common/vp9_postproc.h VP9_COMMON_SRCS-$(CONFIG_VP9_POSTPROC) += common/vp9_postproc.c VP9_COMMON_SRCS-$(HAVE_MMX) += common/x86/vp9_loopfilter_mmx.asm VP9_COMMON_SRCS-$(HAVE_SSSE3) += common/x86/vp9_subpixel_8t_ssse3.asm -VP9_COMMON_SRCS-$(HAVE_SSSE3) += common/x86/vp9_subpixel_8t_intrin_ssse3.c ifeq ($(CONFIG_VP9_POSTPROC),yes) VP9_COMMON_SRCS-$(HAVE_MMX) += common/x86/vp9_postproc_mmx.asm VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_postproc_sse2.asm -- 2.7.4