/* * Copyright (c) 2022 Samsung Electronics Co., Ltd. * All Rights Reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * - Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. * * - Redistributions in binary form must reproduce the above copyright notice, * this list of conditions and the following disclaimer in the documentation * and/or other materials provided with the distribution. * * - Neither the name of the copyright owner, nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED.IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE * POSSIBILITY OF SUCH DAMAGE. */ #include "oapv_sad_sse.h" #if X86_SSE /* SSD ***********************************************************************/ #define SSE_SSD_16B_8PEL(src1, src2, shift, s00, s01, s02, s00a) \ s00 = _mm_loadu_si128((__m128i*)(src1)); \ s01 = _mm_loadu_si128((__m128i*)(src2)); \ s02 = _mm_sub_epi16(s00, s01); \ \ s00 = _mm_cvtepi16_epi32(s02); \ s00 = _mm_mullo_epi32(s00, s00); \ \ s01 = _mm_srli_si128(s02, 8); \ s01 = _mm_cvtepi16_epi32(s01); \ s01 = _mm_mullo_epi32(s01, s01); \ \ s00 = _mm_srli_epi32(s00, shift); \ s01 = _mm_srli_epi32(s01, shift); \ s00a = _mm_add_epi32(s00a, s00); \ s00a = _mm_add_epi32(s00a, s01); static s64 ssd_16b_sse_8x8(int w, int h, void * src1, void * src2, int s_src1, int s_src2) { s64 ssd; s16 * s1; s16 * s2; const int shift = 0; __m128i s00, s01, s02, s00a; s1 = (s16 *)src1; s2 = (s16 *)src2; s00a = _mm_setzero_si128(); SSE_SSD_16B_8PEL(s1, s2, shift, s00, s01, s02, s00a); SSE_SSD_16B_8PEL(s1 + s_src1, s2 + s_src2, shift, s00, s01, s02, s00a); SSE_SSD_16B_8PEL(s1 + s_src1*2, s2 + s_src2*2, shift, s00, s01, s02, s00a); SSE_SSD_16B_8PEL(s1 + s_src1*3, s2 + s_src2*3, shift, s00, s01, s02, s00a); SSE_SSD_16B_8PEL(s1 + s_src1*4, s2 + s_src2*4, shift, s00, s01, s02, s00a); SSE_SSD_16B_8PEL(s1 + s_src1*5, s2 + s_src2*5, shift, s00, s01, s02, s00a); SSE_SSD_16B_8PEL(s1 + s_src1*6, s2 + s_src2*6, shift, s00, s01, s02, s00a); SSE_SSD_16B_8PEL(s1 + s_src1*7, s2 + s_src2*7, shift, s00, s01, s02, s00a); ssd = _mm_extract_epi32(s00a, 0); ssd += _mm_extract_epi32(s00a, 1); ssd += _mm_extract_epi32(s00a, 2); ssd += _mm_extract_epi32(s00a, 3); return ssd; } const oapv_fn_ssd_t oapv_tbl_fn_ssd_16b_sse[2] = { ssd_16b_sse_8x8, NULL }; int oapv_dc_removed_had8x8_sse(pel* org, int s_org) { int sad = 0; /* all 128 bit registers are named with a suffix mxnb, where m is the */ /* number of n bits packed in the register */ __m128i src0_8x16b, src1_8x16b, src2_8x16b, src3_8x16b; __m128i src4_8x16b, src5_8x16b, src6_8x16b, src7_8x16b; __m128i pred0_8x16b, pred1_8x16b, pred2_8x16b, pred3_8x16b; __m128i pred4_8x16b, pred5_8x16b, pred6_8x16b, pred7_8x16b; __m128i out0_8x16b, out1_8x16b, out2_8x16b, out3_8x16b; __m128i out4_8x16b, out5_8x16b, out6_8x16b, out7_8x16b; src0_8x16b = _mm_loadu_si128((__m128i*) org); org = org + s_org; src1_8x16b = _mm_loadu_si128((__m128i*) org); org = org + s_org; src2_8x16b = _mm_loadu_si128((__m128i*) org); org = org + s_org; src3_8x16b = _mm_loadu_si128((__m128i*) org); org = org + s_org; src4_8x16b = _mm_loadu_si128((__m128i*) org); org = org + s_org; src5_8x16b = _mm_loadu_si128((__m128i*) org); org = org + s_org; src6_8x16b = _mm_loadu_si128((__m128i*) org); org = org + s_org; src7_8x16b = _mm_loadu_si128((__m128i*) org); org = org + s_org; /**************** 8x8 horizontal transform *******************************/ /*********************** 8x8 16 bit Transpose ************************/ out3_8x16b = _mm_unpacklo_epi16(src0_8x16b, src1_8x16b); pred0_8x16b = _mm_unpacklo_epi16(src2_8x16b, src3_8x16b); out2_8x16b = _mm_unpacklo_epi16(src4_8x16b, src5_8x16b); pred3_8x16b = _mm_unpacklo_epi16(src6_8x16b, src7_8x16b); out7_8x16b = _mm_unpackhi_epi16(src0_8x16b, src1_8x16b); src2_8x16b = _mm_unpackhi_epi16(src2_8x16b, src3_8x16b); pred7_8x16b = _mm_unpackhi_epi16(src4_8x16b, src5_8x16b); src6_8x16b = _mm_unpackhi_epi16(src6_8x16b, src7_8x16b); out1_8x16b = _mm_unpacklo_epi32(out3_8x16b, pred0_8x16b); out3_8x16b = _mm_unpackhi_epi32(out3_8x16b, pred0_8x16b); pred1_8x16b = _mm_unpacklo_epi32(out2_8x16b, pred3_8x16b); pred3_8x16b = _mm_unpackhi_epi32(out2_8x16b, pred3_8x16b); out5_8x16b = _mm_unpacklo_epi32(out7_8x16b, src2_8x16b); out7_8x16b = _mm_unpackhi_epi32(out7_8x16b, src2_8x16b); pred5_8x16b = _mm_unpacklo_epi32(pred7_8x16b, src6_8x16b); pred7_8x16b = _mm_unpackhi_epi32(pred7_8x16b, src6_8x16b); out0_8x16b = _mm_unpacklo_epi64(out1_8x16b, pred1_8x16b); out1_8x16b = _mm_unpackhi_epi64(out1_8x16b, pred1_8x16b); out2_8x16b = _mm_unpacklo_epi64(out3_8x16b, pred3_8x16b); out3_8x16b = _mm_unpackhi_epi64(out3_8x16b, pred3_8x16b); out4_8x16b = _mm_unpacklo_epi64(out5_8x16b, pred5_8x16b); out5_8x16b = _mm_unpackhi_epi64(out5_8x16b, pred5_8x16b); out6_8x16b = _mm_unpacklo_epi64(out7_8x16b, pred7_8x16b); out7_8x16b = _mm_unpackhi_epi64(out7_8x16b, pred7_8x16b); /********************** 8x8 16 bit Transpose End *********************/ /* r0 + r1 */ pred0_8x16b = _mm_add_epi16(out0_8x16b, out1_8x16b); /* r2 + r3 */ pred2_8x16b = _mm_add_epi16(out2_8x16b, out3_8x16b); /* r4 + r5 */ pred4_8x16b = _mm_add_epi16(out4_8x16b, out5_8x16b); /* r6 + r7 */ pred6_8x16b = _mm_add_epi16(out6_8x16b, out7_8x16b); /* r0 + r1 + r2 + r3 */ pred1_8x16b = _mm_add_epi16(pred0_8x16b, pred2_8x16b); /* r4 + r5 + r6 + r7 */ pred5_8x16b = _mm_add_epi16(pred4_8x16b, pred6_8x16b); /* r0 + r1 + r2 + r3 + r4 + r5 + r6 + r7 */ src0_8x16b = _mm_add_epi16(pred1_8x16b, pred5_8x16b); /* r0 + r1 + r2 + r3 - r4 - r5 - r6 - r7 */ src4_8x16b = _mm_sub_epi16(pred1_8x16b, pred5_8x16b); /* r0 + r1 - r2 - r3 */ pred1_8x16b = _mm_sub_epi16(pred0_8x16b, pred2_8x16b); /* r4 + r5 - r6 - r7 */ pred5_8x16b = _mm_sub_epi16(pred4_8x16b, pred6_8x16b); /* r0 + r1 - r2 - r3 + r4 + r5 - r6 - r7 */ src2_8x16b = _mm_add_epi16(pred1_8x16b, pred5_8x16b); /* r0 + r1 - r2 - r3 - r4 - r5 + r6 + r7 */ src6_8x16b = _mm_sub_epi16(pred1_8x16b, pred5_8x16b); /* r0 - r1 */ pred0_8x16b = _mm_sub_epi16(out0_8x16b, out1_8x16b); /* r2 - r3 */ pred2_8x16b = _mm_sub_epi16(out2_8x16b, out3_8x16b); /* r4 - r5 */ pred4_8x16b = _mm_sub_epi16(out4_8x16b, out5_8x16b); /* r6 - r7 */ pred6_8x16b = _mm_sub_epi16(out6_8x16b, out7_8x16b); /* r0 - r1 + r2 - r3 */ pred1_8x16b = _mm_add_epi16(pred0_8x16b, pred2_8x16b); /* r4 - r5 + r6 - r7 */ pred5_8x16b = _mm_add_epi16(pred4_8x16b, pred6_8x16b); /* r0 - r1 + r2 - r3 + r4 - r5 + r6 - r7 */ src1_8x16b = _mm_add_epi16(pred1_8x16b, pred5_8x16b); /* r0 - r1 + r2 - r3 - r4 + r5 - r6 + r7 */ src5_8x16b = _mm_sub_epi16(pred1_8x16b, pred5_8x16b); /* r0 - r1 - r2 + r3 */ pred1_8x16b = _mm_sub_epi16(pred0_8x16b, pred2_8x16b); /* r4 - r5 - r6 + r7 */ pred5_8x16b = _mm_sub_epi16(pred4_8x16b, pred6_8x16b); /* r0 - r1 - r2 + r3 + r4 - r5 - r6 + r7 */ src3_8x16b = _mm_add_epi16(pred1_8x16b, pred5_8x16b); /* r0 - r1 - r2 + r3 - r4 + r5 + r6 - r7 */ src7_8x16b = _mm_sub_epi16(pred1_8x16b, pred5_8x16b); /*********************** 8x8 16 bit Transpose ************************/ out3_8x16b = _mm_unpacklo_epi16(src0_8x16b, src1_8x16b); pred0_8x16b = _mm_unpacklo_epi16(src2_8x16b, src3_8x16b); out2_8x16b = _mm_unpacklo_epi16(src4_8x16b, src5_8x16b); pred3_8x16b = _mm_unpacklo_epi16(src6_8x16b, src7_8x16b); out7_8x16b = _mm_unpackhi_epi16(src0_8x16b, src1_8x16b); src2_8x16b = _mm_unpackhi_epi16(src2_8x16b, src3_8x16b); pred7_8x16b = _mm_unpackhi_epi16(src4_8x16b, src5_8x16b); src6_8x16b = _mm_unpackhi_epi16(src6_8x16b, src7_8x16b); out1_8x16b = _mm_unpacklo_epi32(out3_8x16b, pred0_8x16b); out3_8x16b = _mm_unpackhi_epi32(out3_8x16b, pred0_8x16b); pred1_8x16b = _mm_unpacklo_epi32(out2_8x16b, pred3_8x16b); pred3_8x16b = _mm_unpackhi_epi32(out2_8x16b, pred3_8x16b); out5_8x16b = _mm_unpacklo_epi32(out7_8x16b, src2_8x16b); out7_8x16b = _mm_unpackhi_epi32(out7_8x16b, src2_8x16b); pred5_8x16b = _mm_unpacklo_epi32(pred7_8x16b, src6_8x16b); pred7_8x16b = _mm_unpackhi_epi32(pred7_8x16b, src6_8x16b); src0_8x16b = _mm_unpacklo_epi64(out1_8x16b, pred1_8x16b); src1_8x16b = _mm_unpackhi_epi64(out1_8x16b, pred1_8x16b); src2_8x16b = _mm_unpacklo_epi64(out3_8x16b, pred3_8x16b); src3_8x16b = _mm_unpackhi_epi64(out3_8x16b, pred3_8x16b); src4_8x16b = _mm_unpacklo_epi64(out5_8x16b, pred5_8x16b); src5_8x16b = _mm_unpackhi_epi64(out5_8x16b, pred5_8x16b); src6_8x16b = _mm_unpacklo_epi64(out7_8x16b, pred7_8x16b); src7_8x16b = _mm_unpackhi_epi64(out7_8x16b, pred7_8x16b); /********************** 8x8 16 bit Transpose End *********************/ /**************** 8x8 horizontal transform *******************************/ { __m128i out0a_8x16b, out1a_8x16b, out2a_8x16b, out3a_8x16b; __m128i out4a_8x16b, out5a_8x16b, out6a_8x16b, out7a_8x16b; __m128i tmp0_8x16b, tmp1_8x16b, tmp2_8x16b, tmp3_8x16b; __m128i tmp4_8x16b, tmp5_8x16b, tmp6_8x16b, tmp7_8x16b; /************************* 8x8 Vertical Transform*************************/ tmp0_8x16b = _mm_srli_si128(src0_8x16b, 8); tmp1_8x16b = _mm_srli_si128(src1_8x16b, 8); tmp2_8x16b = _mm_srli_si128(src2_8x16b, 8); tmp3_8x16b = _mm_srli_si128(src3_8x16b, 8); tmp4_8x16b = _mm_srli_si128(src4_8x16b, 8); tmp5_8x16b = _mm_srli_si128(src5_8x16b, 8); tmp6_8x16b = _mm_srli_si128(src6_8x16b, 8); tmp7_8x16b = _mm_srli_si128(src7_8x16b, 8); /*************************First 4 pixels ********************************/ src0_8x16b = _mm_cvtepi16_epi32(src0_8x16b); src1_8x16b = _mm_cvtepi16_epi32(src1_8x16b); src2_8x16b = _mm_cvtepi16_epi32(src2_8x16b); src3_8x16b = _mm_cvtepi16_epi32(src3_8x16b); src4_8x16b = _mm_cvtepi16_epi32(src4_8x16b); src5_8x16b = _mm_cvtepi16_epi32(src5_8x16b); src6_8x16b = _mm_cvtepi16_epi32(src6_8x16b); src7_8x16b = _mm_cvtepi16_epi32(src7_8x16b); /* r0 + r1 */ pred0_8x16b = _mm_add_epi32(src0_8x16b, src1_8x16b); /* r2 + r3 */ pred2_8x16b = _mm_add_epi32(src2_8x16b, src3_8x16b); /* r4 + r5 */ pred4_8x16b = _mm_add_epi32(src4_8x16b, src5_8x16b); /* r6 + r7 */ pred6_8x16b = _mm_add_epi32(src6_8x16b, src7_8x16b); /* r0 + r1 + r2 + r3 */ pred1_8x16b = _mm_add_epi32(pred0_8x16b, pred2_8x16b); /* r4 + r5 + r6 + r7 */ pred5_8x16b = _mm_add_epi32(pred4_8x16b, pred6_8x16b); /* r0 + r1 + r2 + r3 + r4 + r5 + r6 + r7 */ out0_8x16b = _mm_add_epi32(pred1_8x16b, pred5_8x16b); /* r0 + r1 + r2 + r3 - r4 - r5 - r6 - r7 */ out4_8x16b = _mm_sub_epi32(pred1_8x16b, pred5_8x16b); /* r0 + r1 - r2 - r3 */ pred1_8x16b = _mm_sub_epi32(pred0_8x16b, pred2_8x16b); /* r4 + r5 - r6 - r7 */ pred5_8x16b = _mm_sub_epi32(pred4_8x16b, pred6_8x16b); /* r0 + r1 - r2 - r3 + r4 + r5 - r6 - r7 */ out2_8x16b = _mm_add_epi32(pred1_8x16b, pred5_8x16b); /* r0 + r1 - r2 - r3 - r4 - r5 + r6 + r7 */ out6_8x16b = _mm_sub_epi32(pred1_8x16b, pred5_8x16b); /* r0 - r1 */ pred0_8x16b = _mm_sub_epi32(src0_8x16b, src1_8x16b); /* r2 - r3 */ pred2_8x16b = _mm_sub_epi32(src2_8x16b, src3_8x16b); /* r4 - r5 */ pred4_8x16b = _mm_sub_epi32(src4_8x16b, src5_8x16b); /* r6 - r7 */ pred6_8x16b = _mm_sub_epi32(src6_8x16b, src7_8x16b); /* r0 - r1 + r2 - r3 */ pred1_8x16b = _mm_add_epi32(pred0_8x16b, pred2_8x16b); /* r4 - r5 + r6 - r7 */ pred5_8x16b = _mm_add_epi32(pred4_8x16b, pred6_8x16b); /* r0 - r1 + r2 - r3 + r4 - r5 + r6 - r7 */ out1_8x16b = _mm_add_epi32(pred1_8x16b, pred5_8x16b); /* r0 - r1 + r2 - r3 - r4 + r5 - r6 + r7 */ out5_8x16b = _mm_sub_epi32(pred1_8x16b, pred5_8x16b); /* r0 - r1 - r2 + r3 */ pred1_8x16b = _mm_sub_epi32(pred0_8x16b, pred2_8x16b); /* r4 - r5 - r6 + r7 */ pred5_8x16b = _mm_sub_epi32(pred4_8x16b, pred6_8x16b); /* r0 - r1 - r2 + r3 + r4 - r5 - r6 + r7 */ out3_8x16b = _mm_add_epi32(pred1_8x16b, pred5_8x16b); /* r0 - r1 - r2 + r3 - r4 + r5 + r6 - r7 */ out7_8x16b = _mm_sub_epi32(pred1_8x16b, pred5_8x16b); /*************************First 4 pixels ********************************/ /**************************Next 4 pixels *******************************/ src0_8x16b = _mm_cvtepi16_epi32(tmp0_8x16b); src1_8x16b = _mm_cvtepi16_epi32(tmp1_8x16b); src2_8x16b = _mm_cvtepi16_epi32(tmp2_8x16b); src3_8x16b = _mm_cvtepi16_epi32(tmp3_8x16b); src4_8x16b = _mm_cvtepi16_epi32(tmp4_8x16b); src5_8x16b = _mm_cvtepi16_epi32(tmp5_8x16b); src6_8x16b = _mm_cvtepi16_epi32(tmp6_8x16b); src7_8x16b = _mm_cvtepi16_epi32(tmp7_8x16b); /* r0 + r1 */ pred0_8x16b = _mm_add_epi32(src0_8x16b, src1_8x16b); /* r2 + r3 */ pred2_8x16b = _mm_add_epi32(src2_8x16b, src3_8x16b); /* r4 + r5 */ pred4_8x16b = _mm_add_epi32(src4_8x16b, src5_8x16b); /* r6 + r7 */ pred6_8x16b = _mm_add_epi32(src6_8x16b, src7_8x16b); /* r0 + r1 + r2 + r3 */ pred1_8x16b = _mm_add_epi32(pred0_8x16b, pred2_8x16b); /* r4 + r5 + r6 + r7 */ pred5_8x16b = _mm_add_epi32(pred4_8x16b, pred6_8x16b); /* r0 + r1 + r2 + r3 + r4 + r5 + r6 + r7 */ out0a_8x16b = _mm_add_epi32(pred1_8x16b, pred5_8x16b); /* r0 + r1 + r2 + r3 - r4 - r5 - r6 - r7 */ out4a_8x16b = _mm_sub_epi32(pred1_8x16b, pred5_8x16b); /* r0 + r1 - r2 - r3 */ pred1_8x16b = _mm_sub_epi32(pred0_8x16b, pred2_8x16b); /* r4 + r5 - r6 - r7 */ pred5_8x16b = _mm_sub_epi32(pred4_8x16b, pred6_8x16b); /* r0 + r1 - r2 - r3 + r4 + r5 - r6 - r7 */ out2a_8x16b = _mm_add_epi32(pred1_8x16b, pred5_8x16b); /* r0 + r1 - r2 - r3 - r4 - r5 + r6 + r7 */ out6a_8x16b = _mm_sub_epi32(pred1_8x16b, pred5_8x16b); /* r0 - r1 */ pred0_8x16b = _mm_sub_epi32(src0_8x16b, src1_8x16b); /* r2 - r3 */ pred2_8x16b = _mm_sub_epi32(src2_8x16b, src3_8x16b); /* r4 - r5 */ pred4_8x16b = _mm_sub_epi32(src4_8x16b, src5_8x16b); /* r6 - r7 */ pred6_8x16b = _mm_sub_epi32(src6_8x16b, src7_8x16b); /* r0 - r1 + r2 - r3 */ pred1_8x16b = _mm_add_epi32(pred0_8x16b, pred2_8x16b); /* r4 - r5 + r6 - r7 */ pred5_8x16b = _mm_add_epi32(pred4_8x16b, pred6_8x16b); /* r0 - r1 + r2 - r3 + r4 - r5 + r6 - r7 */ out1a_8x16b = _mm_add_epi32(pred1_8x16b, pred5_8x16b); /* r0 - r1 + r2 - r3 - r4 + r5 - r6 + r7 */ out5a_8x16b = _mm_sub_epi32(pred1_8x16b, pred5_8x16b); /* r0 - r1 - r2 + r3 */ pred1_8x16b = _mm_sub_epi32(pred0_8x16b, pred2_8x16b); /* r4 - r5 - r6 + r7 */ pred5_8x16b = _mm_sub_epi32(pred4_8x16b, pred6_8x16b); /* r0 - r1 - r2 + r3 + r4 - r5 - r6 + r7 */ out3a_8x16b = _mm_add_epi32(pred1_8x16b, pred5_8x16b); /* r0 - r1 - r2 + r3 - r4 + r5 + r6 - r7 */ out7a_8x16b = _mm_sub_epi32(pred1_8x16b, pred5_8x16b); /**************************Next 4 pixels *******************************/ /************************* 8x8 Vertical Transform*************************/ /****************************SATD calculation ****************************/ src0_8x16b = _mm_abs_epi32(out0_8x16b); src1_8x16b = _mm_abs_epi32(out1_8x16b); src2_8x16b = _mm_abs_epi32(out2_8x16b); src3_8x16b = _mm_abs_epi32(out3_8x16b); src4_8x16b = _mm_abs_epi32(out4_8x16b); src5_8x16b = _mm_abs_epi32(out5_8x16b); src6_8x16b = _mm_abs_epi32(out6_8x16b); src7_8x16b = _mm_abs_epi32(out7_8x16b); s32* p = (s32*)&src0_8x16b; p[0] = 0; src0_8x16b = _mm_add_epi32(src0_8x16b, src1_8x16b); src2_8x16b = _mm_add_epi32(src2_8x16b, src3_8x16b); src4_8x16b = _mm_add_epi32(src4_8x16b, src5_8x16b); src6_8x16b = _mm_add_epi32(src6_8x16b, src7_8x16b); src0_8x16b = _mm_add_epi32(src0_8x16b, src2_8x16b); src4_8x16b = _mm_add_epi32(src4_8x16b, src6_8x16b); src0_8x16b = _mm_add_epi32(src0_8x16b, src4_8x16b); src0_8x16b = _mm_hadd_epi32(src0_8x16b, src0_8x16b); src0_8x16b = _mm_hadd_epi32(src0_8x16b, src0_8x16b); sad += _mm_cvtsi128_si32(src0_8x16b); src0_8x16b = _mm_abs_epi32(out0a_8x16b); src1_8x16b = _mm_abs_epi32(out1a_8x16b); src2_8x16b = _mm_abs_epi32(out2a_8x16b); src3_8x16b = _mm_abs_epi32(out3a_8x16b); src4_8x16b = _mm_abs_epi32(out4a_8x16b); src5_8x16b = _mm_abs_epi32(out5a_8x16b); src6_8x16b = _mm_abs_epi32(out6a_8x16b); src7_8x16b = _mm_abs_epi32(out7a_8x16b); src0_8x16b = _mm_add_epi32(src0_8x16b, src1_8x16b); src2_8x16b = _mm_add_epi32(src2_8x16b, src3_8x16b); src4_8x16b = _mm_add_epi32(src4_8x16b, src5_8x16b); src6_8x16b = _mm_add_epi32(src6_8x16b, src7_8x16b); src0_8x16b = _mm_add_epi32(src0_8x16b, src2_8x16b); src4_8x16b = _mm_add_epi32(src4_8x16b, src6_8x16b); src0_8x16b = _mm_add_epi32(src0_8x16b, src4_8x16b); src0_8x16b = _mm_hadd_epi32(src0_8x16b, src0_8x16b); src0_8x16b = _mm_hadd_epi32(src0_8x16b, src0_8x16b); sad += _mm_cvtsi128_si32(src0_8x16b); sad = (sad + 2) >> 2; return sad; } } #endif /* X86_SSE */