/****************************************************************************** * * Copyright (C) 2015 The Android Open Source Project * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. * You may obtain a copy of the License at: * * http://www.apache.org/licenses/LICENSE-2.0 * * Unless required by applicable law or agreed to in writing, software * distributed under the License is distributed on an "AS IS" BASIS, * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. * See the License for the specific language governing permissions and * limitations under the License. * ***************************************************************************** * Originally developed and contributed by Ittiam Systems Pvt. Ltd, Bangalore */ /** ******************************************************************************* * @file * ih264_iquant_itrans_recon_ssse3.c * * @brief * Contains function definitions for inverse quantization, inverse * transform and reconstruction * * @author * Mohit [100664] * * @par List of Functions: * - ih264_iquant_itrans_recon_4x4_ssse3() * - ih264_iquant_itrans_recon_8x8_ssse3() * * @remarks * None * ******************************************************************************* */ /* User include files */ #include "ih264_typedefs.h" #include "ih264_defs.h" #include "ih264_trans_macros.h" #include "ih264_macros.h" #include "ih264_platform_macros.h" #include "ih264_trans_data.h" #include "ih264_size_defs.h" #include "ih264_structs.h" #include "ih264_trans_quant_itrans_iquant.h" #include <immintrin.h> /* ******************************************************************************** * * @brief This function reconstructs a 4x4 sub block from quantized resiude and * prediction buffer * * @par Description: * The quantized residue is first inverse quantized, then inverse transformed. * This inverse transformed content is added to the prediction buffer to recon- * struct the end output * * @param[in] pi2_src * quantized 4x4 block * * @param[in] pu1_pred * prediction 4x4 block * * @param[out] pu1_out * reconstructed 4x4 block * * @param[in] src_strd * quantization buffer stride * * @param[in] pred_strd, * Prediction buffer stride * * @param[in] out_strd * recon buffer Stride * * @param[in] pu2_scaling_list * pointer to scaling list * * @param[in] pu2_norm_adjust * pointer to inverse scale matrix * * @param[in] u4_qp_div_6 * Floor (qp/6) * * @param[in] pi4_tmp * temporary buffer of size 1*16 * * @returns none * * @remarks none * ******************************************************************************* */ void ih264_iquant_itrans_recon_4x4_ssse3(WORD16 *pi2_src, UWORD8 *pu1_pred, UWORD8 *pu1_out, WORD32 pred_strd, WORD32 out_strd, const UWORD16 *pu2_iscal_mat, const UWORD16 *pu2_weigh_mat, UWORD32 u4_qp_div_6, WORD16 *pi2_tmp, WORD32 iq_start_idx, WORD16 *pi2_dc_ld_addr) { UWORD32 *pu4_out = (UWORD32 *) pu1_out; __m128i src_r0_r1, src_r2_r3; __m128i src_r0, src_r1, src_r2, src_r3; __m128i scalemat_r0_r1, scalemat_r2_r3, predload_r; __m128i pred_r0, pred_r1, pred_r2, pred_r3; __m128i sign_reg, dequant_r0_r1, dequant_r2_r3; __m128i zero_8x16b = _mm_setzero_si128(); // all bits reset to zero __m128i temp0, temp1, temp2, temp3, temp4, temp5, temp6, temp7; __m128i resq_r0, resq_r1, resq_r2, resq_r3; __m128i add_rshift = _mm_set1_epi32((1 << (3 - u4_qp_div_6))); __m128i value_32 = _mm_set1_epi32(32); UNUSED (pi2_tmp); UNUSED (pi2_dc_ld_addr); /*************************************************************/ /* Dequantization of coefficients. Will be replaced by SIMD */ /* operations on platform */ /*************************************************************/ src_r0_r1 = _mm_loadu_si128((__m128i *) (pi2_src)); //a00 a01 a02 a03 a10 a11 a12 a13 -- the source matrix 0th,1st row src_r2_r3 = _mm_loadu_si128((__m128i *) (pi2_src + 8)); //a20 a21 a22 a23 a30 a31 a32 a33 -- the source matrix 2nd,3rd row scalemat_r0_r1 = _mm_loadu_si128((__m128i *) (pu2_iscal_mat)); //b00 b01 b02 b03 b10 b11 b12 b13 -- the scaling matrix 0th,1st row scalemat_r2_r3 = _mm_loadu_si128((__m128i *) (pu2_iscal_mat + 8)); //b20 b21 b22 b23 b30 b31 b32 b33 -- the scaling matrix 2nd,3rd row dequant_r0_r1 = _mm_loadu_si128((__m128i *) (pu2_weigh_mat)); //q00 q01 q02 q03 q10 q11 q12 q13 -- all 16 bits dequant_r2_r3 = _mm_loadu_si128((__m128i *) (pu2_weigh_mat + 8)); //q20 q21 q22 q23 q30 q31 q32 q33 -- all 16 bits temp0 = _mm_mullo_epi16(scalemat_r0_r1, dequant_r0_r1); //b00*q00 b01*q01 b02*q02 b03*q03 b10*q10 b11*q11 b12*q12 b13*q13 -- 16 bit result temp1 = _mm_mullo_epi16(scalemat_r2_r3, dequant_r2_r3); //b00*q00 b01*q01 b02*q02 b03*q03 b10*q10 b11*q11 b12*q12 b13*q13 -- 16 bit result temp4 = _mm_unpacklo_epi16(temp0, zero_8x16b); // b00*q00 0 b01*q01 0 b02*q02 0 b03*q03 0 -- 16 bit long temp5 = _mm_unpackhi_epi16(temp0, zero_8x16b); // b10*q10 0 b11*q11 0 b12*q12 0 b13*q13 0 -- 16 bit long temp6 = _mm_unpacklo_epi16(temp1, zero_8x16b); // b00*q00 0 b01*q01 0 b02*q02 0 b03*q03 0 -- 16 bit long temp7 = _mm_unpackhi_epi16(temp1, zero_8x16b); // b10*q10 0 b11*q11 0 b12*q12 0 b13*q13 0 -- 16 bit long src_r0 = _mm_unpacklo_epi16(src_r0_r1, zero_8x16b); // a00 0 a01 0 a02 0 a03 0 -- 16 bit long src_r1 = _mm_unpackhi_epi16(src_r0_r1, zero_8x16b); // a10 0 a11 0 a12 0 a13 0 -- 16 bit long src_r2 = _mm_unpacklo_epi16(src_r2_r3, zero_8x16b); // a20 0 a21 0 a22 0 a23 0 -- 16 bit long src_r3 = _mm_unpackhi_epi16(src_r2_r3, zero_8x16b); // a30 0 a31 0 a32 0 a33 0 -- 16 bit long temp4 = _mm_madd_epi16(src_r0, temp4); //a00*b00*q00 a10*b10*q10 a20*b20*q20 a30*b30 q30 -- 32 bits long temp5 = _mm_madd_epi16(src_r1, temp5); temp6 = _mm_madd_epi16(src_r2, temp6); temp7 = _mm_madd_epi16(src_r3, temp7); if (u4_qp_div_6 >= 4) { resq_r0 = _mm_slli_epi32(temp4, u4_qp_div_6 - 4); resq_r1 = _mm_slli_epi32(temp5, u4_qp_div_6 - 4); resq_r2 = _mm_slli_epi32(temp6, u4_qp_div_6 - 4); resq_r3 = _mm_slli_epi32(temp7, u4_qp_div_6 - 4); } else { temp4 = _mm_add_epi32(temp4, add_rshift); temp5 = _mm_add_epi32(temp5, add_rshift); temp6 = _mm_add_epi32(temp6, add_rshift); temp7 = _mm_add_epi32(temp7, add_rshift); resq_r0 = _mm_srai_epi32(temp4, 4 - u4_qp_div_6); resq_r1 = _mm_srai_epi32(temp5, 4 - u4_qp_div_6); resq_r2 = _mm_srai_epi32(temp6, 4 - u4_qp_div_6); resq_r3 = _mm_srai_epi32(temp7, 4 - u4_qp_div_6); } if (iq_start_idx == 1) { resq_r0 = _mm_insert_epi16(resq_r0,(WORD32)pi2_src[0],0); if (pi2_src[0] >= 0) resq_r0 = _mm_insert_epi16(resq_r0,0,1); else resq_r0 = _mm_insert_epi16(resq_r0,-1,1); } /* Perform Inverse transform */ /*-------------------------------------------------------------*/ /* IDCT [ Horizontal transformation ] */ /*-------------------------------------------------------------*/ // Matrix transpose /* * a0 a1 a2 a3 * b0 b1 b2 b3 * c0 c1 c2 c3 * d0 d1 d2 d3 */ temp1 = _mm_unpacklo_epi32(resq_r0, resq_r1); //a0 b0 a1 b1 temp3 = _mm_unpacklo_epi32(resq_r2, resq_r3); //c0 d0 c1 d1 temp2 = _mm_unpackhi_epi32(resq_r0, resq_r1); //a2 b2 a3 b3 temp4 = _mm_unpackhi_epi32(resq_r2, resq_r3); //c2 d2 c3 d3 resq_r0 = _mm_unpacklo_epi64(temp1, temp3); //a0 b0 c0 d0 resq_r1 = _mm_unpackhi_epi64(temp1, temp3); //a1 b1 c1 d1 resq_r2 = _mm_unpacklo_epi64(temp2, temp4); //a2 b2 c2 d2 resq_r3 = _mm_unpackhi_epi64(temp2, temp4); //a3 b3 c3 d3 //Transform starts -- horizontal transform /*------------------------------------------------------------------*/ /* z0 = w0 + w2 */ temp0 = _mm_add_epi32(resq_r0, resq_r2); /* z1 = w0 - w2 */ temp1 = _mm_sub_epi32(resq_r0, resq_r2); /* z2 = (w1 >> 1) - w3 */ temp2 = _mm_srai_epi32(resq_r1, 1); //(w1>>1) temp2 = _mm_sub_epi32(temp2, resq_r3); //(w1>>1) - w3 /* z3 = w1 + (w3 >> 1) */ temp3 = _mm_srai_epi32(resq_r3, 1); //(w3>>1) + w1 temp3 = _mm_add_epi32(temp3, resq_r1); /*----------------------------------------------------------*/ /* x0 = z0 + z3 */ resq_r0 = _mm_add_epi32(temp0, temp3); /* x1 = z1 + z2 */ resq_r1 = _mm_add_epi32(temp1, temp2); /* x2 = z1 - z2 */ resq_r2 = _mm_sub_epi32(temp1, temp2); /* x3 = z0 - z3 */ resq_r3 = _mm_sub_epi32(temp0, temp3); // Matrix transpose /* * a0 b0 c0 d0 * a1 b1 c1 d1 * a2 b2 c2 d2 * a3 b3 c3 d3 */ temp1 = _mm_unpacklo_epi32(resq_r0, resq_r1); //a0 a1 b0 b1 temp3 = _mm_unpacklo_epi32(resq_r2, resq_r3); //a2 a3 b2 b3 temp2 = _mm_unpackhi_epi32(resq_r0, resq_r1); //c0 c1 d0 d1 temp4 = _mm_unpackhi_epi32(resq_r2, resq_r3); //c2 c3 d2 d3 resq_r0 = _mm_unpacklo_epi64(temp1, temp3); //a0 a1 a2 a3 resq_r1 = _mm_unpackhi_epi64(temp1, temp3); //b0 b1 b2 b3 resq_r2 = _mm_unpacklo_epi64(temp2, temp4); //c0 c1 c2 c3 resq_r3 = _mm_unpackhi_epi64(temp2, temp4); //d0 d1 d2 d3 //Transform ends -- horizontal transform zero_8x16b = _mm_setzero_si128(); // all bits reset to zero //Load pred buffer predload_r = _mm_loadl_epi64((__m128i *) (&pu1_pred[0])); //p00 p01 p02 p03 0 0 0 0 0 0 0 0 -- all 8 bits pred_r0 = _mm_unpacklo_epi8(predload_r, zero_8x16b); //p00 p01 p02 p03 0 0 0 0 -- all 16 bits predload_r = _mm_loadl_epi64((__m128i *) (&pu1_pred[pred_strd])); //p10 p11 p12 p13 0 0 0 0 0 0 0 0 -- all 8 bits pred_r1 = _mm_unpacklo_epi8(predload_r, zero_8x16b); //p10 p11 p12 p13 0 0 0 0 -- all 16 bits predload_r = _mm_loadl_epi64((__m128i *) (&pu1_pred[2 * pred_strd])); //p20 p21 p22 p23 0 0 0 0 0 0 0 0 -- all 8 bits pred_r2 = _mm_unpacklo_epi8(predload_r, zero_8x16b); //p20 p21 p22 p23 0 0 0 0 -- all 16 bits predload_r = _mm_loadl_epi64((__m128i *) (&pu1_pred[3 * pred_strd])); //p30 p31 p32 p33 0 0 0 0 0 0 0 0 -- all 8 bits pred_r3 = _mm_unpacklo_epi8(predload_r, zero_8x16b); //p30 p31 p32 p33 0 0 0 0 -- all 16 bits pred_r0 = _mm_unpacklo_epi16(pred_r0, zero_8x16b); //p00 p01 p02 p03 -- 32 bits sign extended pred_r1 = _mm_unpacklo_epi16(pred_r1, zero_8x16b); //p10 p11 p12 p13 -- 32 bits sign extended pred_r2 = _mm_unpacklo_epi16(pred_r2, zero_8x16b); //p20 p21 p22 p23 -- 32 bits sign extended pred_r3 = _mm_unpacklo_epi16(pred_r3, zero_8x16b); //p30 p31 p32 p33 -- 32 bits sign extended /*--------------------------------------------------------------*/ /* IDCT [ Vertical transformation] and Xij = (xij + 32)>>6 */ /* */ /* Add the prediction and store it back to same buffer */ /*--------------------------------------------------------------*/ /* z0j = y0j + y2j */ temp0 = _mm_add_epi32(resq_r0, resq_r2); /* z1j = y0j - y2j */ temp1 = _mm_sub_epi32(resq_r0, resq_r2); /* z2j = (y1j>>1) - y3j */ temp2 = _mm_srai_epi32(resq_r1, 1); //(y1j>>1) temp2 = _mm_sub_epi32(temp2, resq_r3); /* z3j = y1j + (y3j>>1) */ temp3 = _mm_srai_epi32(resq_r3, 1); //(y3j>>1) temp3 = _mm_add_epi32(temp3, resq_r1); /* x0j = z0j + z3j */ temp4 = _mm_add_epi32(temp0, temp3); temp4 = _mm_add_epi32(temp4, value_32); temp4 = _mm_srai_epi32(temp4, 6); temp4 = _mm_add_epi32(temp4, pred_r0); /* x1j = z1j + z2j */ temp5 = _mm_add_epi32(temp1, temp2); temp5 = _mm_add_epi32(temp5, value_32); temp5 = _mm_srai_epi32(temp5, 6); temp5 = _mm_add_epi32(temp5, pred_r1); /* x2j = z1j - z2j */ temp6 = _mm_sub_epi32(temp1, temp2); temp6 = _mm_add_epi32(temp6, value_32); temp6 = _mm_srai_epi32(temp6, 6); temp6 = _mm_add_epi32(temp6, pred_r2); /* x3j = z0j - z3j */ temp7 = _mm_sub_epi32(temp0, temp3); temp7 = _mm_add_epi32(temp7, value_32); temp7 = _mm_srai_epi32(temp7, 6); temp7 = _mm_add_epi32(temp7, pred_r3); // 32-bit to 16-bit conversion temp0 = _mm_packs_epi32(temp4, temp5); temp1 = _mm_packs_epi32(temp6, temp7); /*------------------------------------------------------------------*/ //Clipping the results to 8 bits sign_reg = _mm_cmpgt_epi16(temp0, zero_8x16b); // sign check temp0 = _mm_and_si128(temp0, sign_reg); sign_reg = _mm_cmpgt_epi16(temp1, zero_8x16b); temp1 = _mm_and_si128(temp1, sign_reg); resq_r0 = _mm_packus_epi16(temp0, temp1); resq_r1 = _mm_srli_si128(resq_r0, 4); resq_r2 = _mm_srli_si128(resq_r1, 4); resq_r3 = _mm_srli_si128(resq_r2, 4); *pu4_out = _mm_cvtsi128_si32(resq_r0); pu1_out += out_strd; pu4_out = (UWORD32 *) (pu1_out); *(pu4_out) = _mm_cvtsi128_si32(resq_r1); pu1_out += out_strd; pu4_out = (UWORD32 *) (pu1_out); *(pu4_out) = _mm_cvtsi128_si32(resq_r2); pu1_out += out_strd; pu4_out = (UWORD32 *) (pu1_out); *(pu4_out) = _mm_cvtsi128_si32(resq_r3); } /** ******************************************************************************* * * @brief * This function performs inverse quant and Inverse transform type Ci4 for 8x8 block * * @par Description: * Performs inverse transform Ci8 and adds the residue to get the * reconstructed block * * @param[in] pi2_src * Input 8x8coefficients * * @param[in] pu1_pred * Prediction 8x8 block * * @param[out] pu1_recon * Output 8x8 block * * @param[in] q_div * QP/6 * * @param[in] q_rem * QP%6 * * @param[in] q_lev * Quantizer level * * @param[in] u4_src_stride * Input stride * * @param[in] u4_pred_stride, * Prediction stride * * @param[in] u4_out_stride * Output Stride * * @param[in] pi4_tmp * temporary buffer of size 1*64 * the tmp for each block * * @param[in] pu4_iquant_mat * Pointer to the inverse quantization matrix * * @returns Void * * @remarks * None * ******************************************************************************* */ void ih264_iquant_itrans_recon_8x8_ssse3(WORD16 *pi2_src, UWORD8 *pu1_pred, UWORD8 *pu1_out, WORD32 pred_strd, WORD32 out_strd, const UWORD16 *pu2_iscale_mat, const UWORD16 *pu2_weigh_mat, UWORD32 qp_div, WORD16 *pi2_tmp, WORD32 iq_start_idx, WORD16 *pi2_dc_ld_addr) { __m128i src_r0; __m128i scalemat_r0; __m128i zero_8x16b = _mm_setzero_si128(); // all bits reset to zero // __m128i one_8x16b = _mm_set1_epi8(255); // all bits set to 1 // __m128i one_zero_mask = _mm_unpacklo_epi16(one_8x16b, zero_8x16b); // 1 0 1 0 1 0 1 0 --- 16 bits size __m128i value_32 = _mm_set1_epi32(32); __m128i add_rshift = _mm_set1_epi32((1 << (5 - qp_div))); __m128i dequant_r0; __m128i predload_r; __m128i pred_r0_1, pred_r1_1, pred_r2_1, pred_r3_1, pred_r4_1, pred_r5_1, pred_r6_1, pred_r7_1; __m128i sign_reg; __m128i src_r0_1, src_r0_2; __m128i scalemat_r0_1, scalemat_r0_2; __m128i temp1, temp2, temp3, temp4, temp5, temp6, temp7, temp8; __m128i temp10, temp11, temp12, temp13, temp14, temp15, temp16, temp17, temp18, temp19, temp20; // To store dequantization results __m128i resq_r0_1, resq_r0_2, resq_r1_1, resq_r1_2, resq_r2_1, resq_r2_2, resq_r3_1, resq_r3_2, resq_r4_1, resq_r4_2, resq_r5_1, resq_r5_2, resq_r6_1, resq_r6_2, resq_r7_1, resq_r7_2; UNUSED (pi2_tmp); UNUSED (iq_start_idx); UNUSED (pi2_dc_ld_addr); /*************************************************************/ /* Dequantization of coefficients. Will be replaced by SIMD */ /* operations on platform. Note : DC coeff is not scaled */ /*************************************************************/ // Row 0 processing src_r0 = _mm_loadu_si128((__m128i *) (pi2_src)); //a00 a01 a02 a03 a04 a05 a06 a07 -- the source matrix 0th row scalemat_r0 = _mm_loadu_si128((__m128i *) (pu2_iscale_mat)); //b00 b01 b02 b03 b04 b05 b06 b07 -- the scaling matrix 0th row dequant_r0 = _mm_loadu_si128((__m128i *) (&pu2_weigh_mat[0])); //q0 q1 q2 q3 q4 q5 q6 q7 -- all 16 bits src_r0_1 = _mm_unpacklo_epi16(src_r0, zero_8x16b); //a00 0 a01 0 a02 0 a03 0 -- 16 bit long src_r0_2 = _mm_unpackhi_epi16(src_r0, zero_8x16b); // a04 0 a05 0 a06 0 a07 0 -- 16 bit long temp10 = _mm_mullo_epi16(scalemat_r0, dequant_r0); //b00*q0 b01*q1 b02*q2 b03*q3 b04*q4 b05*q5 b06*q6 b07*q7 -- 16 bit result scalemat_r0_1 = _mm_unpacklo_epi16(temp10, zero_8x16b); // b00*q0 0 b01*q1 0 b02*q2 0 b03*q3 0 -- 16 bit long scalemat_r0_2 = _mm_unpackhi_epi16(temp10, zero_8x16b); // b04*q4 0 b05*q5 0 b06*q6 0 b07*q7 0 -- 16 bit long temp5 = _mm_madd_epi16(src_r0_1, scalemat_r0_1); // a00*b00*q0 a01*b01*q1 a02*b02*q2 a03*b03*q3 -- 32 bits long temp7 = _mm_madd_epi16(src_r0_2, scalemat_r0_2); // a04*b04*q4 a05*b05*q5 a06*b06*q6 a07*b07*q7 -- 32 bits long if (qp_div >= 6) { resq_r0_1 = _mm_slli_epi32(temp5, qp_div - 6); resq_r0_2 = _mm_slli_epi32(temp7, qp_div - 6); } else { temp5 = _mm_add_epi32(temp5, add_rshift); temp7 = _mm_add_epi32(temp7, add_rshift); resq_r0_1 = _mm_srai_epi32(temp5, 6 - qp_div); resq_r0_2 = _mm_srai_epi32(temp7, 6 - qp_div); } resq_r0_1 = _mm_packs_epi32(resq_r0_1, resq_r0_2); //a00*b00*q0 a01*b01*q1 a02*b02*q2 a03*b03*q3 a04*b04*q4 a05*b05*q5 a06*b06*q6 a07*b07*q7 -- 16 bit long // Row 1 processing src_r0 = _mm_loadu_si128((__m128i *) (pi2_src + 8)); //a00 a01 a02 a03 a04 a05 a06 a07 a08 -- the source matrix 1st row scalemat_r0 = _mm_loadu_si128((__m128i *) (pu2_iscale_mat + 8)); //b00 b01 b02 b03 b04 b05 b06 b07 b08 -- the scaling matrix 1st row dequant_r0 = _mm_loadu_si128((__m128i *) (&pu2_weigh_mat[8])); //q0 q1 q2 q3 q4 q5 q6 q7 -- all 16 bits src_r0_1 = _mm_unpacklo_epi16(src_r0, zero_8x16b); //a00 0 a01 0 a02 0 a03 0 -- 16 bit long src_r0_2 = _mm_unpackhi_epi16(src_r0, zero_8x16b); // a04 0 a05 0 a06 0 a07 0 -- 16 bit long temp10 = _mm_mullo_epi16(scalemat_r0, dequant_r0); //b00*q0 b01*q1 b02*q2 b03*q3 b04*q4 b05*q5 b06*q6 b07*q7 -- 16 bit result scalemat_r0_1 = _mm_unpacklo_epi16(temp10, zero_8x16b); // b00*q0 0 b01*q1 0 b02*q2 0 b03*q3 0 -- 16 bit long scalemat_r0_2 = _mm_unpackhi_epi16(temp10, zero_8x16b); // b04*q4 0 b05*q5 0 b06*q6 0 b07*q7 0 -- 16 bit long temp5 = _mm_madd_epi16(src_r0_1, scalemat_r0_1); // a00*b00*q0 a01*b01*q1 a02*b02*q2 a03*b03*q3 -- 32 bits long temp7 = _mm_madd_epi16(src_r0_2, scalemat_r0_2); // a04*b04*q4 a05*b05*q5 a06*b06*q6 a07*b07*q7 -- 32 bits long if (qp_div >= 6) { resq_r1_1 = _mm_slli_epi32(temp5, qp_div - 6); resq_r1_2 = _mm_slli_epi32(temp7, qp_div - 6); } else { temp5 = _mm_add_epi32(temp5, add_rshift); temp7 = _mm_add_epi32(temp7, add_rshift); resq_r1_1 = _mm_srai_epi32(temp5, 6 - qp_div); resq_r1_2 = _mm_srai_epi32(temp7, 6 - qp_div); } resq_r1_1 = _mm_packs_epi32(resq_r1_1, resq_r1_2); //a00*b00*q0 a01*b01*q1 a02*b02*q2 a03*b03*q3 a04*b04*q4 a05*b05*q5 a06*b06*q6 a07*b07*q7 -- 16 bit long // Row 2 processing src_r0 = _mm_loadu_si128((__m128i *) (pi2_src + 16)); //a00 a01 a02 a03 a04 a05 a06 a07 a08 -- the source matrix 2nd row scalemat_r0 = _mm_loadu_si128((__m128i *) (pu2_iscale_mat + 16)); //b00 b01 b02 b03 b04 b05 b06 b07 b08 -- the scaling matrix 2nd row dequant_r0 = _mm_loadu_si128((__m128i *) (&pu2_weigh_mat[16])); //q0 q1 q2 q3 q4 q5 q6 q7 -- all 16 bits src_r0_1 = _mm_unpacklo_epi16(src_r0, zero_8x16b); //a00 0 a01 0 a02 0 a03 0 -- 16 bit long src_r0_2 = _mm_unpackhi_epi16(src_r0, zero_8x16b); // a04 0 a05 0 a06 0 a07 0 -- 16 bit long temp10 = _mm_mullo_epi16(scalemat_r0, dequant_r0); //b00*q0 b01*q1 b02*q2 b03*q3 b04*q4 b05*q5 b06*q6 b07*q7 -- 16 bit result scalemat_r0_1 = _mm_unpacklo_epi16(temp10, zero_8x16b); // b00*q0 0 b01*q1 0 b02*q2 0 b03*q3 0 -- 16 bit long scalemat_r0_2 = _mm_unpackhi_epi16(temp10, zero_8x16b); // b04*q4 0 b05*q5 0 b06*q6 0 b07*q7 0 -- 16 bit long temp5 = _mm_madd_epi16(src_r0_1, scalemat_r0_1); // a00*b00*q0 a01*b01*q1 a02*b02*q2 a03*b03*q3 -- 32 bits long temp7 = _mm_madd_epi16(src_r0_2, scalemat_r0_2); // a04*b04*q4 a05*b05*q5 a06*b06*q6 a07*b07*q7 -- 32 bits long if (qp_div >= 6) { resq_r2_1 = _mm_slli_epi32(temp5, qp_div - 6); resq_r2_2 = _mm_slli_epi32(temp7, qp_div - 6); } else { temp5 = _mm_add_epi32(temp5, add_rshift); temp7 = _mm_add_epi32(temp7, add_rshift); resq_r2_1 = _mm_srai_epi32(temp5, 6 - qp_div); resq_r2_2 = _mm_srai_epi32(temp7, 6 - qp_div); } resq_r2_1 = _mm_packs_epi32(resq_r2_1, resq_r2_2); //a00*b00*q0 a01*b01*q1 a02*b02*q2 a03*b03*q3 a04*b04*q4 a05*b05*q5 a06*b06*q6 a07*b07*q7 -- 16 bit long // Row 3 processing src_r0 = _mm_loadu_si128((__m128i *) (pi2_src + 24)); //a00 a01 a02 a03 a04 a05 a06 a07 a08 -- the source matrix 3rd row scalemat_r0 = _mm_loadu_si128((__m128i *) (pu2_iscale_mat + 24)); //b00 b01 b02 b03 b04 b05 b06 b07 b08 -- the scaling matrix 3rd row dequant_r0 = _mm_loadu_si128((__m128i *) (&pu2_weigh_mat[24])); //q0 q1 q2 q3 q4 q5 q6 q7 -- all 16 bits src_r0_1 = _mm_unpacklo_epi16(src_r0, zero_8x16b); //a00 0 a01 0 a02 0 a03 0 -- 16 bit long src_r0_2 = _mm_unpackhi_epi16(src_r0, zero_8x16b); // a04 0 a05 0 a06 0 a07 0 -- 16 bit long temp10 = _mm_mullo_epi16(scalemat_r0, dequant_r0); //b00*q0 b01*q1 b02*q2 b03*q3 b04*q4 b05*q5 b06*q6 b07*q7 -- 16 bit result scalemat_r0_1 = _mm_unpacklo_epi16(temp10, zero_8x16b); // b00*q0 0 b01*q1 0 b02*q2 0 b03*q3 0 -- 16 bit long scalemat_r0_2 = _mm_unpackhi_epi16(temp10, zero_8x16b); // b04*q4 0 b05*q5 0 b06*q6 0 b07*q7 0 -- 16 bit long temp5 = _mm_madd_epi16(src_r0_1, scalemat_r0_1); // a00*b00*q0 a01*b01*q1 a02*b02*q2 a03*b03*q3 - 32 bits long temp7 = _mm_madd_epi16(src_r0_2, scalemat_r0_2); // a04*b04*q4 a05*b05*q5 a06*b06*q6 a07*b07*q7 -- 32 bits long if (qp_div >= 6) { resq_r3_1 = _mm_slli_epi32(temp5, qp_div - 6); resq_r3_2 = _mm_slli_epi32(temp7, qp_div - 6); } else { temp5 = _mm_add_epi32(temp5, add_rshift); temp7 = _mm_add_epi32(temp7, add_rshift); resq_r3_1 = _mm_srai_epi32(temp5, 6 - qp_div); resq_r3_2 = _mm_srai_epi32(temp7, 6 - qp_div); } resq_r3_1 = _mm_packs_epi32(resq_r3_1, resq_r3_2); //a00*b00*q0 a01*b01*q1 a02*b02*q2 a03*b03*q3 a04*b04*q4 a05*b05*q5 a06*b06*q6 a07*b07*q7 -- 16 bit long // Row 4 processing src_r0 = _mm_loadu_si128((__m128i *) (pi2_src + 32)); //a00 a01 a02 a03 a04 a05 a06 a07 a08 -- the source matrix 4th row scalemat_r0 = _mm_loadu_si128((__m128i *) (pu2_iscale_mat + 32)); //b00 b01 b02 b03 b04 b05 b06 b07 b08 -- the scaling matrix 4th row dequant_r0 = _mm_loadu_si128((__m128i *) (&pu2_weigh_mat[32])); //q0 q1 q2 q3 q4 q5 q6 q7 -- all 16 bits src_r0_1 = _mm_unpacklo_epi16(src_r0, zero_8x16b); //a00 0 a01 0 a02 0 a03 0 -- 16 bit long src_r0_2 = _mm_unpackhi_epi16(src_r0, zero_8x16b); // a04 0 a05 0 a06 0 a07 0 -- 16 bit long temp10 = _mm_mullo_epi16(scalemat_r0, dequant_r0); //b00*q0 b01*q1 b02*q2 b03*q3 b04*q4 b05*q5 b06*q6 b07*q7 -- 16 bit result scalemat_r0_1 = _mm_unpacklo_epi16(temp10, zero_8x16b); // b00*q0 0 b01*q1 0 b02*q2 0 b03*q3 0 -- 16 bit long scalemat_r0_2 = _mm_unpackhi_epi16(temp10, zero_8x16b); // b04*q4 0 b05*q5 0 b06*q6 0 b07*q7 0 -- 16 bit long temp5 = _mm_madd_epi16(src_r0_1, scalemat_r0_1); // a00*b00*q0 a01*b01*q1 a02*b02*q2 a03*b03*q3 -- 32 bits long temp7 = _mm_madd_epi16(src_r0_2, scalemat_r0_2); // a04*b04*q4 a05*b05*q5 a06*b06*q6 a07*b07*q7 -- 32 bits long if (qp_div >= 6) { resq_r4_1 = _mm_slli_epi32(temp5, qp_div - 6); resq_r4_2 = _mm_slli_epi32(temp7, qp_div - 6); } else { temp5 = _mm_add_epi32(temp5, add_rshift); temp7 = _mm_add_epi32(temp7, add_rshift); resq_r4_1 = _mm_srai_epi32(temp5, 6 - qp_div); resq_r4_2 = _mm_srai_epi32(temp7, 6 - qp_div); } resq_r4_1 = _mm_packs_epi32(resq_r4_1, resq_r4_2); //a00*b00*q0 a01*b01*q1 a02*b02*q2 a03*b03*q3 a04*b04*q4 a05*b05*q5 a06*b06*q6 a07*b07*q7 -- 16 bit long // Row 5 processing src_r0 = _mm_loadu_si128((__m128i *) (pi2_src + 40)); //a00 a01 a02 a03 a04 a05 a06 a07 a08 -- the source matrix 5th row scalemat_r0 = _mm_loadu_si128((__m128i *) (pu2_iscale_mat + 40)); //b00 b01 b02 b03 b04 b05 b06 b07 b08 -- the scaling matrix 5th row dequant_r0 = _mm_loadu_si128((__m128i *) (&pu2_weigh_mat[40])); //q0 q1 q2 q3 q4 q5 q6 q7 -- all 16 bits src_r0_1 = _mm_unpacklo_epi16(src_r0, zero_8x16b); //a00 0 a01 0 a02 0 a03 0 -- 16 bit long src_r0_2 = _mm_unpackhi_epi16(src_r0, zero_8x16b); // a04 0 a05 0 a06 0 a07 0 -- 16 bit long temp10 = _mm_mullo_epi16(scalemat_r0, dequant_r0); //b00*q0 b01*q1 b02*q2 b03*q3 b04*q4 b05*q5 b06*q6 b07*q7 -- 16 bit result scalemat_r0_1 = _mm_unpacklo_epi16(temp10, zero_8x16b); // b00*q0 0 b01*q1 0 b02*q2 0 b03*q3 0 -- 16 bit long scalemat_r0_2 = _mm_unpackhi_epi16(temp10, zero_8x16b); // b04*q4 0 b05*q5 0 b06*q6 0 b07*q7 0 -- 16 bit long temp5 = _mm_madd_epi16(src_r0_1, scalemat_r0_1); // a00*b00*q0 a01*b01*q1 a02*b02*q2 a03*b03*q3 -- 32 bits long temp7 = _mm_madd_epi16(src_r0_2, scalemat_r0_2); // a04*b04*q4 a05*b05*q5 a06*b06*q6 a07*b07*q7 -- 32 bits long if (qp_div >= 6) { resq_r5_1 = _mm_slli_epi32(temp5, qp_div - 6); resq_r5_2 = _mm_slli_epi32(temp7, qp_div - 6); //resq_r5_1 = _mm_and_si128(resq_r5_1,one_zero_mask); //resq_r5_2 = _mm_and_si128(resq_r5_2,one_zero_mask); } else { temp5 = _mm_add_epi32(temp5, add_rshift); temp7 = _mm_add_epi32(temp7, add_rshift); resq_r5_1 = _mm_srai_epi32(temp5, 6 - qp_div); resq_r5_2 = _mm_srai_epi32(temp7, 6 - qp_div); } resq_r5_1 = _mm_packs_epi32(resq_r5_1, resq_r5_2); //a00*b00*q0 a01*b01*q1 a02*b02*q2 a03*b03*q3 a04*b04*q4 a05*b05*q5 a06*b06*q6 a07*b07*q7 -- 16 bit long // Row 6 processing src_r0 = _mm_loadu_si128((__m128i *) (pi2_src + 48)); //a00 a01 a02 a03 a04 a05 a06 a07 a08 -- the source matrix 6th row scalemat_r0 = _mm_loadu_si128((__m128i *) (pu2_iscale_mat + 48)); //b00 b01 b02 b03 b04 b05 b06 b07 b08 -- the scaling matrix 6th row dequant_r0 = _mm_loadu_si128((__m128i *) (&pu2_weigh_mat[48])); //q0 q1 q2 q3 q4 q5 q6 q7 -- all 16 bits src_r0_1 = _mm_unpacklo_epi16(src_r0, zero_8x16b); //a00 0 a01 0 a02 0 a03 0 -- 16 bit long src_r0_2 = _mm_unpackhi_epi16(src_r0, zero_8x16b); // a04 0 a05 0 a06 0 a07 0 -- 16 bit long temp10 = _mm_mullo_epi16(scalemat_r0, dequant_r0); //b00*q0 b01*q1 b02*q2 b03*q3 b04*q4 b05*q5 b06*q6 b07*q7 -- 16 bit result scalemat_r0_1 = _mm_unpacklo_epi16(temp10, zero_8x16b); // b00*q0 0 b01*q1 0 b02*q2 0 b03*q3 0 -- 16 bit long scalemat_r0_2 = _mm_unpackhi_epi16(temp10, zero_8x16b); // b04*q4 0 b05*q5 0 b06*q6 0 b07*q7 0 -- 16 bit long temp5 = _mm_madd_epi16(src_r0_1, scalemat_r0_1); // a00*b00*q0 a01*b01*q1 a02*b02*q2 a03*b03*q3 -- 32 bits long temp7 = _mm_madd_epi16(src_r0_2, scalemat_r0_2); // a04*b04*q4 a05*b05*q5 a06*b06*q6 a07*b07*q7 -- 32 bits long if (qp_div >= 6) { resq_r6_1 = _mm_slli_epi32(temp5, qp_div - 6); resq_r6_2 = _mm_slli_epi32(temp7, qp_div - 6); //resq_r6_1 = _mm_and_si128(resq_r6_1,one_zero_mask); //resq_r6_2 = _mm_and_si128(resq_r6_2,one_zero_mask); } else { temp5 = _mm_add_epi32(temp5, add_rshift); temp7 = _mm_add_epi32(temp7, add_rshift); resq_r6_1 = _mm_srai_epi32(temp5, 6 - qp_div); resq_r6_2 = _mm_srai_epi32(temp7, 6 - qp_div); //resq_r6_1 = _mm_and_si128(resq_r6_1,one_zero_mask); //resq_r6_2 = _mm_and_si128(resq_r6_2,one_zero_mask); } resq_r6_1 = _mm_packs_epi32(resq_r6_1, resq_r6_2); //a00*b00*q0 a01*b01*q1 a02*b02*q2 a03*b03*q3 a04*b04*q4 a05*b05*q5 a06*b06*q6 a07*b07*q7 -- 16 bit long // Row 7 processing src_r0 = _mm_loadu_si128((__m128i *) (pi2_src + 56)); //a00 a01 a02 a03 a04 a05 a06 a07 a08 -- the source matrix 7th row scalemat_r0 = _mm_loadu_si128((__m128i *) (pu2_iscale_mat + 56)); //b00 b01 b02 b03 b04 b05 b06 b07 b08 -- the scaling matrix 7th row dequant_r0 = _mm_loadu_si128((__m128i *) (&pu2_weigh_mat[56])); //q0 q1 q2 q3 q4 q5 q6 q7 -- all 16 bits src_r0_1 = _mm_unpacklo_epi16(src_r0, zero_8x16b); //a00 0 a01 0 a02 0 a03 0 -- 16 bit long src_r0_2 = _mm_unpackhi_epi16(src_r0, zero_8x16b); // a04 0 a05 0 a06 0 a07 0 -- 16 bit long temp10 = _mm_mullo_epi16(scalemat_r0, dequant_r0); //b00*q0 b01*q1 b02*q2 b03*q3 b04*q4 b05*q5 b06*q6 b07*q7 -- 16 bit result scalemat_r0_1 = _mm_unpacklo_epi16(temp10, zero_8x16b); // b00*q0 0 b01*q1 0 b02*q2 0 b03*q3 0 -- 16 bit long scalemat_r0_2 = _mm_unpackhi_epi16(temp10, zero_8x16b); // b04*q4 0 b05*q5 0 b06*q6 0 b07*q7 0 -- 16 bit long temp5 = _mm_madd_epi16(src_r0_1, scalemat_r0_1); // a00*b00*q0 a01*b01*q1 a02*b02*q2 a03*b03*q3 -- 32 bits long temp7 = _mm_madd_epi16(src_r0_2, scalemat_r0_2); // a04*b04*q4 a05*b05*q5 a06*b06*q6 a07*b07*q7 -- 32 bits long if (qp_div >= 6) { resq_r7_1 = _mm_slli_epi32(temp5, qp_div - 6); resq_r7_2 = _mm_slli_epi32(temp7, qp_div - 6); } else { temp5 = _mm_add_epi32(temp5, add_rshift); temp7 = _mm_add_epi32(temp7, add_rshift); resq_r7_1 = _mm_srai_epi32(temp5, 6 - qp_div); resq_r7_2 = _mm_srai_epi32(temp7, 6 - qp_div); } resq_r7_1 = _mm_packs_epi32(resq_r7_1, resq_r7_2); //a00*b00*q0 a01*b01*q1 a02*b02*q2 a03*b03*q3 a04*b04*q4 a05*b05*q5 a06*b06*q6 a07*b07*q7 -- 16 bit long /* Perform Inverse transform */ /*--------------------------------------------------------------------*/ /* IDCT [ Horizontal transformation ] */ /*--------------------------------------------------------------------*/ // Matrix transpose /* * a0 a1 a2 a3 a4 a5 a6 a7 * b0 b1 b2 b3 b4 b5 b6 b7 * c0 c1 c2 c3 c4 c5 c6 c7 * d0 d1 d2 d3 d4 d5 d6 d7 */ temp1 = _mm_unpacklo_epi16(resq_r0_1, resq_r1_1); //a0 b0 a1 b1 a2 b2 a3 b3 temp3 = _mm_unpacklo_epi16(resq_r2_1, resq_r3_1); //c0 d0 c1 d1 c2 d2 c3 d3 temp2 = _mm_unpackhi_epi16(resq_r0_1, resq_r1_1); //a4 b4 a5 b5 a6 b6 a7 b7 temp4 = _mm_unpackhi_epi16(resq_r2_1, resq_r3_1); //c4 d4 c5 d5 c6 d6 c7 d7 resq_r0_1 = _mm_unpacklo_epi32(temp1, temp3); //a0 b0 c0 d0 a1 b1 c1 d1 resq_r1_1 = _mm_unpackhi_epi32(temp1, temp3); //a2 b2 c2 d2 a3 b3 c3 d3 resq_r2_1 = _mm_unpacklo_epi32(temp2, temp4); //a4 b4 c4 d4 a5 b5 c5 d5 resq_r3_1 = _mm_unpackhi_epi32(temp2, temp4); //a6 b6 c6 d6 a7 b7 c7 d7 /* * e0 e1 e2 e3 e4 e5 e6 e7 * f0 f1 f2 f3 f4 f5 f6 f7 * g0 g1 g2 g3 g4 g5 g6 g7 * h0 h1 h2 h3 h4 h5 h6 h7 */ temp1 = _mm_unpacklo_epi16(resq_r4_1, resq_r5_1); //e0 f0 e1 f1 e2 f2 e2 f3 temp3 = _mm_unpacklo_epi16(resq_r6_1, resq_r7_1); //g0 h0 g1 h1 g2 h2 g3 h3 temp2 = _mm_unpackhi_epi16(resq_r4_1, resq_r5_1); //e4 f4 e5 f5 e6 f6 e7 f7 temp4 = _mm_unpackhi_epi16(resq_r6_1, resq_r7_1); //g4 h4 g5 h5 g6 h6 g7 h7 resq_r4_1 = _mm_unpacklo_epi32(temp1, temp3); //e0 f0 g0 h0 e1 f1 g1 h1 resq_r5_1 = _mm_unpackhi_epi32(temp1, temp3); //e2 f2 g2 h2 e3 f3 g3 h3 resq_r6_1 = _mm_unpacklo_epi32(temp2, temp4); //e4 f4 g4 h4 e5 f5 g5 h5 resq_r7_1 = _mm_unpackhi_epi32(temp2, temp4); //e6 f6 g6 h6 e7 f7 g7 h7 /* * a0 b0 c0 d0 a1 b1 c1 d1 * a2 b2 c2 d2 a3 b3 c3 d3 * a4 b4 c4 d4 a5 b5 c5 d5 * a6 b6 c6 d6 a7 b7 c7 d7 * e0 f0 g0 h0 e1 f1 g1 h1 * e2 f2 g2 h2 e3 f3 g3 h3 * e4 f4 g4 h4 e5 f5 g5 h5 * e6 f6 g6 h6 e7 f7 g7 h7 */ resq_r0_2 = _mm_unpacklo_epi64(resq_r0_1, resq_r4_1); //a0 b0 c0 d0 e0 f0 g0 h0 resq_r1_2 = _mm_unpackhi_epi64(resq_r0_1, resq_r4_1); //a1 b1 c1 d1 e1 f1 g1 h1 resq_r2_2 = _mm_unpacklo_epi64(resq_r1_1, resq_r5_1); //a2 b2 c2 d2 e2 f2 g2 h2 resq_r3_2 = _mm_unpackhi_epi64(resq_r1_1, resq_r5_1); //a3 b3 c3 d3 e3 f3 g3 h3 resq_r4_2 = _mm_unpacklo_epi64(resq_r2_1, resq_r6_1); //a4 b4 c4 d4 e4 f4 g4 h4 resq_r5_2 = _mm_unpackhi_epi64(resq_r2_1, resq_r6_1); //a5 b5 c5 d5 e5 f5 g5 h5 resq_r6_2 = _mm_unpacklo_epi64(resq_r3_1, resq_r7_1); //a6 b6 c6 d6 e6 f6 g6 h6 resq_r7_2 = _mm_unpackhi_epi64(resq_r3_1, resq_r7_1); //a7 b7 c7 d7 e7 f7 g7 h7 sign_reg = _mm_cmpgt_epi16(zero_8x16b, resq_r1_2); resq_r1_1 = _mm_unpacklo_epi16(resq_r1_2, sign_reg); //a1 b1 c1 d1 -- 32 bit resq_r1_2 = _mm_unpackhi_epi16(resq_r1_2, sign_reg); //e1 f1 g1 h1 -- 32 bit sign_reg = _mm_cmpgt_epi16(zero_8x16b, resq_r3_2); resq_r3_1 = _mm_unpacklo_epi16(resq_r3_2, sign_reg); //a3 b3 c3 d3 -- 32 bit resq_r3_2 = _mm_unpackhi_epi16(resq_r3_2, sign_reg); //e3 f3 g3 h3 -- 32 bit sign_reg = _mm_cmpgt_epi16(zero_8x16b, resq_r5_2); resq_r5_1 = _mm_unpacklo_epi16(resq_r5_2, sign_reg); //a5 b5 c5 d5 -- 32 bit resq_r5_2 = _mm_unpackhi_epi16(resq_r5_2, sign_reg); //e5 f5 g5 h5 -- 32 bit sign_reg = _mm_cmpgt_epi16(zero_8x16b, resq_r7_2); resq_r7_1 = _mm_unpacklo_epi16(resq_r7_2, sign_reg); //a7 b7 c7 d7 -- 32 bit resq_r7_2 = _mm_unpackhi_epi16(resq_r7_2, sign_reg); //e7 f7 g7 h7 -- 32 bit //Transform starts -- horizontal transform /*------------------------------------------------------------------*/ /* y0 = w0 + w4 */ temp1 = _mm_add_epi16(resq_r0_2, resq_r4_2); /* y2 = w0 - w4 */ temp3 = _mm_sub_epi16(resq_r0_2, resq_r4_2); /* y1 = -w3 + w5 - w7 - (w7 >> 1) */ temp2 = _mm_sub_epi32(resq_r5_1, resq_r3_1); //-w3+w5 temp10 = _mm_sub_epi32(resq_r5_2, resq_r3_2); temp4 = _mm_sub_epi32(temp2, resq_r7_1); //-w3+w5-w7 temp12 = _mm_sub_epi32(temp10, resq_r7_2); temp5 = _mm_srai_epi32(resq_r7_1, 1); //w7>>1 temp13 = _mm_srai_epi32(resq_r7_2, 1); temp2 = _mm_sub_epi32(temp4, temp5); //-w3+w5-w7 -(w7>>1) temp10 = _mm_sub_epi32(temp12, temp13); temp2 = _mm_packs_epi32(temp2, temp10); /* y3 = w1 + w7 - w3 - (w3 >> 1) */ temp4 = _mm_add_epi32(resq_r1_1, resq_r7_1); //w1+w7 temp12 = _mm_add_epi32(resq_r1_2, resq_r7_2); temp4 = _mm_sub_epi32(temp4, resq_r3_1); //w1+w7-w3 temp12 = _mm_sub_epi32(temp12, resq_r3_2); temp5 = _mm_srai_epi32(resq_r3_1, 1); //w3>>1 temp13 = _mm_srai_epi32(resq_r3_2, 1); temp4 = _mm_sub_epi32(temp4, temp5); //w1+w7-w3-(w3>>1) temp12 = _mm_sub_epi32(temp12, temp13); temp4 = _mm_packs_epi32(temp4, temp12); /* y4 = (w2 >> 1) - w6 */ temp5 = _mm_srai_epi16(resq_r2_2, 1); //w2>>1 temp5 = _mm_sub_epi16(temp5, resq_r6_2); //(w2>>1)-w6 /* y5 = -w1 + w7 + w5 + (w5 >> 1) */ temp6 = _mm_sub_epi32(resq_r7_1, resq_r1_1); //w7-w1 temp14 = _mm_sub_epi32(resq_r7_2, resq_r1_2); temp6 = _mm_add_epi32(temp6, resq_r5_1); //w7-w1+w5 temp14 = _mm_add_epi32(temp14, resq_r5_2); temp7 = _mm_srai_epi32(resq_r5_1, 1); //w5>>1 temp15 = _mm_srai_epi32(resq_r5_2, 1); temp6 = _mm_add_epi32(temp6, temp7); //w7-w1_w5+(w5>>1) temp14 = _mm_add_epi32(temp14, temp15); temp6 = _mm_packs_epi32(temp6, temp14); /* y6 = w2 + (w6 >> 1) */ temp7 = _mm_srai_epi16(resq_r6_2, 1); //w6>>1 temp7 = _mm_add_epi16(temp7, resq_r2_2); //(w6>>1)+w2 /* y7 = w3 + w5 + w1 + (w1 >> 1) */ temp8 = _mm_add_epi32(resq_r3_1, resq_r5_1); //w3+w5 temp16 = _mm_add_epi32(resq_r3_2, resq_r5_2); temp8 = _mm_add_epi32(temp8, resq_r1_1); //w3+w5+w1 temp16 = _mm_add_epi32(temp16, resq_r1_2); temp17 = _mm_srai_epi32(resq_r1_1, 1); //w1>>1 temp18 = _mm_srai_epi32(resq_r1_2, 1); temp8 = _mm_add_epi32(temp8, temp17); //w3+w5+w1+(w1>>1) temp16 = _mm_add_epi32(temp16, temp18); temp8 = _mm_packs_epi32(temp8, temp16); /*------------------------------------------------------------------*/ /*------------------------------------------------------------------*/ /* z0 = y0 + y6 */ resq_r0_1 = _mm_add_epi16(temp1, temp7); /* z1 = y1 + (y7 >> 2) */ resq_r1_1 = _mm_srai_epi16(temp8, 2); resq_r1_1 = _mm_add_epi16(resq_r1_1, temp2); /* z2 = y2 + y4 */ resq_r2_1 = _mm_add_epi16(temp3, temp5); /* z3 = y3 + (y5 >> 2) */ resq_r3_1 = _mm_srai_epi16(temp6, 2); resq_r3_1 = _mm_add_epi16(resq_r3_1, temp4); /* z4 = y2 - y4 */ resq_r4_1 = _mm_sub_epi16(temp3, temp5); /* z5 = (y3 >> 2) - y5 */ resq_r5_1 = _mm_srai_epi16(temp4, 2); resq_r5_1 = _mm_sub_epi16(resq_r5_1, temp6); /* z6 = y0 - y6 */ resq_r6_1 = _mm_sub_epi16(temp1, temp7); /* z7 = y7 - (y1 >> 2) */ resq_r7_1 = _mm_srai_epi16(temp2, 2); resq_r7_1 = _mm_sub_epi16(temp8, resq_r7_1); /*------------------------------------------------------------------*/ /*------------------------------------------------------------------*/ /* x0 = z0 + z7 */ temp1 = _mm_add_epi16(resq_r0_1, resq_r7_1); /* x1 = z2 + z5 */ temp2 = _mm_add_epi16(resq_r2_1, resq_r5_1); /* x2 = z4 + z3 */ temp3 = _mm_add_epi16(resq_r4_1, resq_r3_1); /* x3 = z6 + z1 */ temp4 = _mm_add_epi16(resq_r6_1, resq_r1_1); /* x4 = z6 - z1 */ temp5 = _mm_sub_epi16(resq_r6_1, resq_r1_1); /* x5 = z4 - z3 */ temp6 = _mm_sub_epi16(resq_r4_1, resq_r3_1); /* x6 = z2 - z5 */ temp7 = _mm_sub_epi16(resq_r2_1, resq_r5_1); /* x7 = z0 - z7 */ temp8 = _mm_sub_epi16(resq_r0_1, resq_r7_1); /*------------------------------------------------------------------*/ // Matrix transpose /* * a0 b0 c0 d0 e0 f0 g0 h0 * a1 b1 c1 d1 e1 f1 g1 h1 * a2 b2 c2 d2 e2 f2 g2 h2 * a3 b3 c3 d3 e3 f3 g3 h3 */ temp17 = _mm_unpacklo_epi16(temp1, temp2); //a0 a1 b0 b1 c0 c1 d0 d1 temp19 = _mm_unpacklo_epi16(temp3, temp4); //a2 a3 b2 b3 c2 c3 d2 d3 temp18 = _mm_unpackhi_epi16(temp1, temp2); //e0 e1 f0 f1 g0 g1 h0 h1 temp20 = _mm_unpackhi_epi16(temp3, temp4); //e2 e3 f2 f3 g2 g3 h2 h3 resq_r0_1 = _mm_unpacklo_epi32(temp17, temp19); //a0 a1 a2 a3 b0 b1 b2 b3 resq_r1_1 = _mm_unpackhi_epi32(temp17, temp19); //c0 c1 c2 c3 d0 d1 d2 d3 resq_r2_1 = _mm_unpacklo_epi32(temp18, temp20); //e0 e1 e2 e3 f0 f1 f2 f3 resq_r3_1 = _mm_unpackhi_epi32(temp18, temp20); //g0 g2 g2 g3 h0 h1 h2 h3 /* * a4 b4 c4 d4 e4 f4 g4 h4 * a5 b5 c5 d5 e5 f5 g5 h5 * a6 b6 c6 d6 e6 f6 g6 h6 * a7 b7 c7 d7 e7 f7 g7 h7 */ temp17 = _mm_unpacklo_epi16(temp5, temp6); //a4 a5 b4 b5 c4 c5 d4 d5 temp19 = _mm_unpacklo_epi16(temp7, temp8); //a6 a7 b6 b7 c6 c7 d6 d7 temp18 = _mm_unpackhi_epi16(temp5, temp6); //e4 e5 f4 f5 g4 g5 h4 h5 temp20 = _mm_unpackhi_epi16(temp7, temp8); //e6 e7 f6 f7 g6 g7 h6 h7 resq_r4_1 = _mm_unpacklo_epi32(temp17, temp19); //a4 a5 a6 a7 b4 b5 b6 b7 resq_r5_1 = _mm_unpackhi_epi32(temp17, temp19); //c4 c5 c6 c7 d4 d5 d6 d7 resq_r6_1 = _mm_unpacklo_epi32(temp18, temp20); //e4 e5 e6 e7 f4 f5 f6 f7 resq_r7_1 = _mm_unpackhi_epi32(temp18, temp20); //g4 g5 g6 g7 h4 h5 h6 h7 /* a0 a1 a2 a3 b0 b1 b2 b3 * c0 c1 c2 c3 d0 d1 d2 d3 * e0 e1 e2 e3 f0 f1 f2 f3 * g0 g2 g2 g3 h0 h1 h2 h3 * a4 a5 a6 a7 b4 b5 b6 b7 * c4 c5 c6 c7 d4 d5 d6 d7 * e4 e5 e6 e7 f4 f5 f6 f7 * g4 g5 g6 g7 h4 h5 h6 h7 */ resq_r0_2 = _mm_unpacklo_epi64(resq_r0_1, resq_r4_1); //a0 a1 a2 a3 a4 a5 a6 a7 resq_r1_2 = _mm_unpackhi_epi64(resq_r0_1, resq_r4_1); //b0 b1 b2 b3 b4 b5 b6 b7 resq_r2_2 = _mm_unpacklo_epi64(resq_r1_1, resq_r5_1); //c0 c1 c2 c3 c4 c5 c6 c7 resq_r3_2 = _mm_unpackhi_epi64(resq_r1_1, resq_r5_1); //d0 d1 d2 d3 d4 d5 d6 d7 resq_r4_2 = _mm_unpacklo_epi64(resq_r2_1, resq_r6_1); //e0 e1 e2 e3 e4 e5 e6 e7 resq_r5_2 = _mm_unpackhi_epi64(resq_r2_1, resq_r6_1); //f0 f1 f2 f3 f4 f5 f6 f7 resq_r6_2 = _mm_unpacklo_epi64(resq_r3_1, resq_r7_1); //g0 g1 g2 g3 g4 g5 g6 g7 resq_r7_2 = _mm_unpackhi_epi64(resq_r3_1, resq_r7_1); //h0 h1 h2 h3 h4 h5 h6 h7 sign_reg = _mm_cmpgt_epi16(zero_8x16b, resq_r1_2); resq_r1_1 = _mm_unpacklo_epi16(resq_r1_2, sign_reg); //a1 b1 c1 d1 -- 32 bit resq_r1_2 = _mm_unpackhi_epi16(resq_r1_2, sign_reg); //e1 f1 g1 h1 -- 32 bit sign_reg = _mm_cmpgt_epi16(zero_8x16b, resq_r3_2); resq_r3_1 = _mm_unpacklo_epi16(resq_r3_2, sign_reg); //a3 b3 c3 d3 -- 32 bit resq_r3_2 = _mm_unpackhi_epi16(resq_r3_2, sign_reg); //e3 f3 g3 h3 -- 32 bit sign_reg = _mm_cmpgt_epi16(zero_8x16b, resq_r5_2); resq_r5_1 = _mm_unpacklo_epi16(resq_r5_2, sign_reg); //a5 b5 c5 d5 -- 32 bit resq_r5_2 = _mm_unpackhi_epi16(resq_r5_2, sign_reg); //e5 f5 g5 h5 -- 32 bit sign_reg = _mm_cmpgt_epi16(zero_8x16b, resq_r7_2); resq_r7_1 = _mm_unpacklo_epi16(resq_r7_2, sign_reg); //a7 b7 c7 d7 -- 32 bit resq_r7_2 = _mm_unpackhi_epi16(resq_r7_2, sign_reg); //e7 f7 g7 h7 -- 32 bit zero_8x16b = _mm_setzero_si128(); // all bits reset to zero //Load pred buffer row 0 predload_r = _mm_loadl_epi64((__m128i *) (&pu1_pred[0])); //p0 p1 p2 p3 p4 p5 p6 p7 0 0 0 0 0 0 0 0 -- all 8 bits pred_r0_1 = _mm_unpacklo_epi8(predload_r, zero_8x16b); //p0 p1 p2 p3 p4 p5 p6 p7 -- all 16 bits //Load pred buffer row 1 predload_r = _mm_loadl_epi64((__m128i *) (&pu1_pred[pred_strd])); //p0 p1 p2 p3 p4 p5 p6 p7 0 0 0 0 0 0 0 0 -- all 8 bits pred_r1_1 = _mm_unpacklo_epi8(predload_r, zero_8x16b); //p0 p1 p2 p3 p4 p5 p6 p7 -- all 16 bits //Load pred buffer row 2 predload_r = _mm_loadl_epi64((__m128i *) (&pu1_pred[2 * pred_strd])); //p0 p1 p2 p3 p4 p5 p6 p7 0 0 0 0 0 0 0 0 -- all 8 bits pred_r2_1 = _mm_unpacklo_epi8(predload_r, zero_8x16b); //p0 p1 p2 p3 p4 p5 p6 p7 -- all 16 bits //Load pred buffer row 3 predload_r = _mm_loadl_epi64((__m128i *) (&pu1_pred[3 * pred_strd])); //p0 p1 p2 p3 p4 p5 p6 p7 0 0 0 0 0 0 0 0 -- all 8 bits pred_r3_1 = _mm_unpacklo_epi8(predload_r, zero_8x16b); //p0 p1 p2 p3 p4 p5 p6 p7 -- all 16 bits //Load pred buffer row 4 predload_r = _mm_loadl_epi64((__m128i *) (&pu1_pred[4 * pred_strd])); //p0 p1 p2 p3 p4 p5 p6 p7 0 0 0 0 0 0 0 0 -- all 8 bits pred_r4_1 = _mm_unpacklo_epi8(predload_r, zero_8x16b); //p0 p1 p2 p3 p4 p5 p6 p7 -- all 16 bits //Load pred buffer row 5 predload_r = _mm_loadl_epi64((__m128i *) (&pu1_pred[5 * pred_strd])); //p0 p1 p2 p3 p4 p5 p6 p7 0 0 0 0 0 0 0 0 -- all 8 bit pred_r5_1 = _mm_unpacklo_epi8(predload_r, zero_8x16b); //p0 p1 p2 p3 p4 p5 p6 p7 -- all 16 bits //Load pred buffer row 6 predload_r = _mm_loadl_epi64((__m128i *) (&pu1_pred[6 * pred_strd])); //p0 p1 p2 p3 p4 p5 p6 p7 0 0 0 0 0 0 0 0 -- all 8 bits pred_r6_1 = _mm_unpacklo_epi8(predload_r, zero_8x16b); //p0 p1 p2 p3 p4 p5 p6 p7 -- all 16 bits //Load pred buffer row 7 predload_r = _mm_loadl_epi64((__m128i *) (&pu1_pred[7 * pred_strd])); //p0 p1 p2 p3 p4 p5 p6 p7 0 0 0 0 0 0 0 0 -- all 8 bits pred_r7_1 = _mm_unpacklo_epi8(predload_r, zero_8x16b); //p0 p1 p2 p3 p4 p5 p6 p7 -- all 16 bits /*--------------------------------------------------------------------*/ /* IDCT [ Vertical transformation] and Xij = (xij + 32)>>6 */ /* */ /* Add the prediction and store it back to reconstructed frame buffer */ /* [Prediction buffer itself in this case] */ /*--------------------------------------------------------------------*/ /* y0j = w0j + w4j */ temp1 = _mm_add_epi16(resq_r0_2, resq_r4_2); /* y2j = w0j - w4j */ temp3 = _mm_sub_epi16(resq_r0_2, resq_r4_2); /* y1j = -w3j + w5j - w7j - (w7j >> 1) */ temp2 = _mm_sub_epi32(resq_r5_1, resq_r3_1); //-w3+w5 temp10 = _mm_sub_epi32(resq_r5_2, resq_r3_2); temp4 = _mm_sub_epi32(temp2, resq_r7_1); //-w3+w5-w7 temp12 = _mm_sub_epi32(temp10, resq_r7_2); temp5 = _mm_srai_epi32(resq_r7_1, 1); //w7>>1 temp13 = _mm_srai_epi32(resq_r7_2, 1); temp2 = _mm_sub_epi32(temp4, temp5); //-w3+w5-w7 -(w7>>1) temp10 = _mm_sub_epi32(temp12, temp13); temp2 = _mm_packs_epi32(temp2, temp10); /* y3j = w1j + w7j - w3j - (w3j >> 1) */ temp4 = _mm_add_epi32(resq_r1_1, resq_r7_1); //w1+w7 temp12 = _mm_add_epi32(resq_r1_2, resq_r7_2); temp4 = _mm_sub_epi32(temp4, resq_r3_1); //w1+w7-w3 temp12 = _mm_sub_epi32(temp12, resq_r3_2); temp5 = _mm_srai_epi32(resq_r3_1, 1); //w3>>1 temp13 = _mm_srai_epi32(resq_r3_2, 1); temp4 = _mm_sub_epi32(temp4, temp5); //w1+w7-w3-(w3>>1) temp12 = _mm_sub_epi32(temp12, temp13); temp4 = _mm_packs_epi32(temp4, temp12); /* y4j = (w2j >> 1) - w6j */ temp5 = _mm_srai_epi16(resq_r2_2, 1); //w2>>1 temp5 = _mm_sub_epi16(temp5, resq_r6_2); //(w2>>1)-w6 /* y5j = -w1j + w7j + w5j + (w5j >> 1) */ temp6 = _mm_sub_epi32(resq_r7_1, resq_r1_1); //w7-w1 temp14 = _mm_sub_epi32(resq_r7_2, resq_r1_2); temp6 = _mm_add_epi32(temp6, resq_r5_1); //w7-w1+w5 temp14 = _mm_add_epi32(temp14, resq_r5_2); temp7 = _mm_srai_epi32(resq_r5_1, 1); //w5>>1 temp15 = _mm_srai_epi32(resq_r5_2, 1); temp6 = _mm_add_epi32(temp6, temp7); //w7-w1_w5+(w5>>1) temp14 = _mm_add_epi32(temp14, temp15); temp6 = _mm_packs_epi32(temp6, temp14); /* y6j = w2j + (w6j >> 1) */ temp7 = _mm_srai_epi16(resq_r6_2, 1); //w6>>1 temp7 = _mm_add_epi16(temp7, resq_r2_2); //(w6>>1)+w2 /* y7j = w3j + w5j + w1j + (w1j >> 1) */ temp8 = _mm_add_epi32(resq_r3_1, resq_r5_1); //w3+w5 temp16 = _mm_add_epi32(resq_r3_2, resq_r5_2); temp8 = _mm_add_epi32(temp8, resq_r1_1); //w3+w5+w1 temp16 = _mm_add_epi32(temp16, resq_r1_2); temp17 = _mm_srai_epi32(resq_r1_1, 1); //w1>>1 temp18 = _mm_srai_epi32(resq_r1_2, 1); temp8 = _mm_add_epi32(temp8, temp17); //w3+w5+w1+(w1>>1) temp16 = _mm_add_epi32(temp16, temp18); temp8 = _mm_packs_epi32(temp8, temp16); /*------------------------------------------------------------------*/ /*------------------------------------------------------------------*/ /* z0j = y0j + y6j */ resq_r0_1 = _mm_add_epi16(temp1, temp7); /* z1j = y1j + (y7j >> 2) */ resq_r1_1 = _mm_srai_epi16(temp8, 2); resq_r1_1 = _mm_add_epi16(resq_r1_1, temp2); /* z2j = y2j + y4j */ resq_r2_1 = _mm_add_epi16(temp3, temp5); /* z3j = y3j + (y5j >> 2) */ resq_r3_1 = _mm_srai_epi16(temp6, 2); resq_r3_1 = _mm_add_epi16(resq_r3_1, temp4); /* z4j = y2j - y4j */ resq_r4_1 = _mm_sub_epi16(temp3, temp5); /* z5j = (y3j >> 2) - y5j */ resq_r5_1 = _mm_srai_epi16(temp4, 2); resq_r5_1 = _mm_sub_epi16(resq_r5_1, temp6); /* z6j = y0j - y6j */ resq_r6_1 = _mm_sub_epi16(temp1, temp7); /* z7j = y7j - (y1j >> 2) */ resq_r7_1 = _mm_srai_epi16(temp2, 2); resq_r7_1 = _mm_sub_epi16(temp8, resq_r7_1); /*------------------------------------------------------------------*/ /*------------------------------------------------------------------*/ /* x0j = z0j + z7j */ temp1 = _mm_add_epi16(resq_r0_1, resq_r7_1); sign_reg = _mm_cmpgt_epi16(zero_8x16b, temp1); temp10 = _mm_unpacklo_epi16(temp1, sign_reg); temp11 = _mm_unpackhi_epi16(temp1, sign_reg); temp10 = _mm_add_epi32(temp10, value_32); temp11 = _mm_add_epi32(temp11, value_32); temp10 = _mm_srai_epi32(temp10, 6); temp11 = _mm_srai_epi32(temp11, 6); temp10 = _mm_packs_epi32(temp10, temp11); temp1 = _mm_add_epi16(temp10, pred_r0_1); /* x1j = z2j + z5j */ temp2 = _mm_add_epi16(resq_r2_1, resq_r5_1); sign_reg = _mm_cmpgt_epi16(zero_8x16b, temp2); temp10 = _mm_unpacklo_epi16(temp2, sign_reg); temp11 = _mm_unpackhi_epi16(temp2, sign_reg); temp10 = _mm_add_epi32(temp10, value_32); temp11 = _mm_add_epi32(temp11, value_32); temp10 = _mm_srai_epi32(temp10, 6); temp11 = _mm_srai_epi32(temp11, 6); temp10 = _mm_packs_epi32(temp10, temp11); temp2 = _mm_add_epi16(temp10, pred_r1_1); /* x2j = z4j + z3j */ temp3 = _mm_add_epi16(resq_r4_1, resq_r3_1); sign_reg = _mm_cmpgt_epi16(zero_8x16b, temp3); temp10 = _mm_unpacklo_epi16(temp3, sign_reg); temp11 = _mm_unpackhi_epi16(temp3, sign_reg); temp10 = _mm_add_epi32(temp10, value_32); temp11 = _mm_add_epi32(temp11, value_32); temp10 = _mm_srai_epi32(temp10, 6); temp11 = _mm_srai_epi32(temp11, 6); temp10 = _mm_packs_epi32(temp10, temp11); temp3 = _mm_add_epi16(temp10, pred_r2_1); /* x3j = z6j + z1j */ temp4 = _mm_add_epi16(resq_r6_1, resq_r1_1); sign_reg = _mm_cmpgt_epi16(zero_8x16b, temp4); temp10 = _mm_unpacklo_epi16(temp4, sign_reg); temp11 = _mm_unpackhi_epi16(temp4, sign_reg); temp10 = _mm_add_epi32(temp10, value_32); temp11 = _mm_add_epi32(temp11, value_32); temp10 = _mm_srai_epi32(temp10, 6); temp11 = _mm_srai_epi32(temp11, 6); temp10 = _mm_packs_epi32(temp10, temp11); temp4 = _mm_add_epi16(temp10, pred_r3_1); /* x4j = z6j - z1j */ temp5 = _mm_sub_epi16(resq_r6_1, resq_r1_1); sign_reg = _mm_cmpgt_epi16(zero_8x16b, temp5); temp10 = _mm_unpacklo_epi16(temp5, sign_reg); temp11 = _mm_unpackhi_epi16(temp5, sign_reg); temp10 = _mm_add_epi32(temp10, value_32); temp11 = _mm_add_epi32(temp11, value_32); temp10 = _mm_srai_epi32(temp10, 6); temp11 = _mm_srai_epi32(temp11, 6); temp10 = _mm_packs_epi32(temp10, temp11); temp5 = _mm_add_epi16(temp10, pred_r4_1); /* x5j = z4j - z3j */ temp6 = _mm_sub_epi16(resq_r4_1, resq_r3_1); sign_reg = _mm_cmpgt_epi16(zero_8x16b, temp6); temp10 = _mm_unpacklo_epi16(temp6, sign_reg); temp11 = _mm_unpackhi_epi16(temp6, sign_reg); temp10 = _mm_add_epi32(temp10, value_32); temp11 = _mm_add_epi32(temp11, value_32); temp10 = _mm_srai_epi32(temp10, 6); temp11 = _mm_srai_epi32(temp11, 6); temp10 = _mm_packs_epi32(temp10, temp11); temp6 = _mm_add_epi16(temp10, pred_r5_1); /* x6j = z2j - z5j */ temp7 = _mm_sub_epi16(resq_r2_1, resq_r5_1); sign_reg = _mm_cmpgt_epi16(zero_8x16b, temp7); temp10 = _mm_unpacklo_epi16(temp7, sign_reg); temp11 = _mm_unpackhi_epi16(temp7, sign_reg); temp10 = _mm_add_epi32(temp10, value_32); temp11 = _mm_add_epi32(temp11, value_32); temp10 = _mm_srai_epi32(temp10, 6); temp11 = _mm_srai_epi32(temp11, 6); temp10 = _mm_packs_epi32(temp10, temp11); temp7 = _mm_add_epi16(temp10, pred_r6_1); /* x7j = z0j - z7j */ temp8 = _mm_sub_epi16(resq_r0_1, resq_r7_1); sign_reg = _mm_cmpgt_epi16(zero_8x16b, temp8); temp10 = _mm_unpacklo_epi16(temp8, sign_reg); temp11 = _mm_unpackhi_epi16(temp8, sign_reg); temp10 = _mm_add_epi32(temp10, value_32); temp11 = _mm_add_epi32(temp11, value_32); temp10 = _mm_srai_epi32(temp10, 6); temp11 = _mm_srai_epi32(temp11, 6); temp10 = _mm_packs_epi32(temp10, temp11); temp8 = _mm_add_epi16(temp10, pred_r7_1); /*------------------------------------------------------------------*/ //Clipping the results to 8 bits sign_reg = _mm_cmpgt_epi16(temp1, zero_8x16b); // sign check temp1 = _mm_and_si128(temp1, sign_reg); sign_reg = _mm_cmpgt_epi16(temp2, zero_8x16b); // sign check temp2 = _mm_and_si128(temp2, sign_reg); sign_reg = _mm_cmpgt_epi16(temp3, zero_8x16b); // sign check temp3 = _mm_and_si128(temp3, sign_reg); sign_reg = _mm_cmpgt_epi16(temp4, zero_8x16b); // sign check temp4 = _mm_and_si128(temp4, sign_reg); sign_reg = _mm_cmpgt_epi16(temp5, zero_8x16b); // sign check temp5 = _mm_and_si128(temp5, sign_reg); sign_reg = _mm_cmpgt_epi16(temp6, zero_8x16b); // sign check temp6 = _mm_and_si128(temp6, sign_reg); sign_reg = _mm_cmpgt_epi16(temp7, zero_8x16b); // sign check temp7 = _mm_and_si128(temp7, sign_reg); sign_reg = _mm_cmpgt_epi16(temp8, zero_8x16b); // sign check temp8 = _mm_and_si128(temp8, sign_reg); resq_r0_2 = _mm_packus_epi16(temp1, zero_8x16b); resq_r1_2 = _mm_packus_epi16(temp2, zero_8x16b); resq_r2_2 = _mm_packus_epi16(temp3, zero_8x16b); resq_r3_2 = _mm_packus_epi16(temp4, zero_8x16b); resq_r4_2 = _mm_packus_epi16(temp5, zero_8x16b); resq_r5_2 = _mm_packus_epi16(temp6, zero_8x16b); resq_r6_2 = _mm_packus_epi16(temp7, zero_8x16b); resq_r7_2 = _mm_packus_epi16(temp8, zero_8x16b); _mm_storel_epi64((__m128i *) (&pu1_out[0]), resq_r0_2); _mm_storel_epi64((__m128i *) (&pu1_out[out_strd]), resq_r1_2); _mm_storel_epi64((__m128i *) (&pu1_out[2 * out_strd]), resq_r2_2); _mm_storel_epi64((__m128i *) (&pu1_out[3 * out_strd]), resq_r3_2); _mm_storel_epi64((__m128i *) (&pu1_out[4 * out_strd]), resq_r4_2); _mm_storel_epi64((__m128i *) (&pu1_out[5 * out_strd]), resq_r5_2); _mm_storel_epi64((__m128i *) (&pu1_out[6 * out_strd]), resq_r6_2); _mm_storel_epi64((__m128i *) (&pu1_out[7 * out_strd]), resq_r7_2); }