/* *  Copyright 2016 The LibYuv Project Authors. All rights reserved. * *  Use of this source code is governed by a BSD-style license *  that can be found in the LICENSE file in the root of the source *  tree. An additional intellectual property rights grant can be found *  in the file PATENTS. All contributing project authors may *  be found in the AUTHORS file in the root of the source tree. */ #include <string.h> #include "libyuv/row.h" // This module is for GCC MSA#if !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa)#include "libyuv/macros_msa.h" #ifdef __cplusplusnamespace libyuv {extern "C" {#endif #define ALPHA_VAL (-1) // Fill YUV -> RGB conversion constants into vectors#define YUVTORGB_SETUP(yuvconst, ub, vr, ug, vg, bb, bg, br, yg) \  {                                                              \    ub = __msa_fill_w(yuvconst->kUVToB[0]);                      \    vr = __msa_fill_w(yuvconst->kUVToR[1]);                      \    ug = __msa_fill_w(yuvconst->kUVToG[0]);                      \    vg = __msa_fill_w(yuvconst->kUVToG[1]);                      \    bb = __msa_fill_w(yuvconst->kUVBiasB[0]);                    \    bg = __msa_fill_w(yuvconst->kUVBiasG[0]);                    \    br = __msa_fill_w(yuvconst->kUVBiasR[0]);                    \    yg = __msa_fill_w(yuvconst->kYToRgb[0]);                     \  } // Load YUV 422 pixel data#define READYUV422(psrc_y, psrc_u, psrc_v, out_y, out_u, out_v)    \  {                                                                \    uint64_t y_m;                                                  \    uint32_t u_m, v_m;                                             \    v4i32 zero_m = {0};                                            \    y_m = LD(psrc_y);                                              \    u_m = LW(psrc_u);                                              \    v_m = LW(psrc_v);                                              \    out_y = (v16u8)__msa_insert_d((v2i64)zero_m, 0, (int64_t)y_m); \    out_u = (v16u8)__msa_insert_w(zero_m, 0, (int32_t)u_m);        \    out_v = (v16u8)__msa_insert_w(zero_m, 0, (int32_t)v_m);        \  } // Clip input vector elements between 0 to 255#define CLIP_0TO255(in0, in1, in2, in3, in4, in5) \  {                                               \    v4i32 max_m = __msa_ldi_w(0xFF);              \                                                  \    in0 = __msa_maxi_s_w(in0, 0);                 \    in1 = __msa_maxi_s_w(in1, 0);                 \    in2 = __msa_maxi_s_w(in2, 0);                 \    in3 = __msa_maxi_s_w(in3, 0);                 \    in4 = __msa_maxi_s_w(in4, 0);                 \    in5 = __msa_maxi_s_w(in5, 0);                 \    in0 = __msa_min_s_w(max_m, in0);              \    in1 = __msa_min_s_w(max_m, in1);              \    in2 = __msa_min_s_w(max_m, in2);              \    in3 = __msa_min_s_w(max_m, in3);              \    in4 = __msa_min_s_w(max_m, in4);              \    in5 = __msa_min_s_w(max_m, in5);              \  } // Convert 8 pixels of YUV 420 to RGB.#define YUVTORGB(in_y, in_uv, ubvr, ugvg, bb, bg, br, yg, out_b, out_g, out_r) \  {                                                                            \    v8i16 vec0_m, vec1_m;                                                      \    v4i32 reg0_m, reg1_m, reg2_m, reg3_m, reg4_m;                              \    v4i32 reg5_m, reg6_m, reg7_m;                                              \    v16i8 zero_m = {0};                                                        \                                                                               \    vec0_m = (v8i16)__msa_ilvr_b((v16i8)in_y, (v16i8)in_y);                    \    vec1_m = (v8i16)__msa_ilvr_b((v16i8)zero_m, (v16i8)in_uv);                 \    reg0_m = (v4i32)__msa_ilvr_h((v8i16)zero_m, (v8i16)vec0_m);                \    reg1_m = (v4i32)__msa_ilvl_h((v8i16)zero_m, (v8i16)vec0_m);                \    reg2_m = (v4i32)__msa_ilvr_h((v8i16)zero_m, (v8i16)vec1_m);                \    reg3_m = (v4i32)__msa_ilvl_h((v8i16)zero_m, (v8i16)vec1_m);                \    reg0_m *= yg;                                                              \    reg1_m *= yg;                                                              \    reg2_m *= ubvr;                                                            \    reg3_m *= ubvr;                                                            \    reg0_m = __msa_srai_w(reg0_m, 16);                                         \    reg1_m = __msa_srai_w(reg1_m, 16);                                         \    reg4_m = __msa_dotp_s_w((v8i16)vec1_m, (v8i16)ugvg);                       \    reg5_m = __msa_ilvev_w(reg2_m, reg2_m);                                    \    reg6_m = __msa_ilvev_w(reg3_m, reg3_m);                                    \    reg7_m = __msa_ilvr_w(reg4_m, reg4_m);                                     \    reg2_m = __msa_ilvod_w(reg2_m, reg2_m);                                    \    reg3_m = __msa_ilvod_w(reg3_m, reg3_m);                                    \    reg4_m = __msa_ilvl_w(reg4_m, reg4_m);                                     \    reg5_m = reg0_m - reg5_m;                                                  \    reg6_m = reg1_m - reg6_m;                                                  \    reg2_m = reg0_m - reg2_m;                                                  \    reg3_m = reg1_m - reg3_m;                                                  \    reg7_m = reg0_m - reg7_m;                                                  \    reg4_m = reg1_m - reg4_m;                                                  \    reg5_m += bb;                                                              \    reg6_m += bb;                                                              \    reg7_m += bg;                                                              \    reg4_m += bg;                                                              \    reg2_m += br;                                                              \    reg3_m += br;                                                              \    reg5_m = __msa_srai_w(reg5_m, 6);                                          \    reg6_m = __msa_srai_w(reg6_m, 6);                                          \    reg7_m = __msa_srai_w(reg7_m, 6);                                          \    reg4_m = __msa_srai_w(reg4_m, 6);                                          \    reg2_m = __msa_srai_w(reg2_m, 6);                                          \    reg3_m = __msa_srai_w(reg3_m, 6);                                          \    CLIP_0TO255(reg5_m, reg6_m, reg7_m, reg4_m, reg2_m, reg3_m);               \    out_b = __msa_pckev_h((v8i16)reg6_m, (v8i16)reg5_m);                       \    out_g = __msa_pckev_h((v8i16)reg4_m, (v8i16)reg7_m);                       \    out_r = __msa_pckev_h((v8i16)reg3_m, (v8i16)reg2_m);                       \  } // Pack and Store 8 ARGB values.#define STOREARGB(in0, in1, in2, in3, pdst_argb)           \  {                                                        \    v8i16 vec0_m, vec1_m;                                  \    v16u8 dst0_m, dst1_m;                                  \    vec0_m = (v8i16)__msa_ilvev_b((v16i8)in1, (v16i8)in0); \    vec1_m = (v8i16)__msa_ilvev_b((v16i8)in3, (v16i8)in2); \    dst0_m = (v16u8)__msa_ilvr_h(vec1_m, vec0_m);          \    dst1_m = (v16u8)__msa_ilvl_h(vec1_m, vec0_m);          \    ST_UB2(dst0_m, dst1_m, pdst_argb, 16);                 \  } // Takes ARGB input and calculates Y.#define ARGBTOY(argb0, argb1, argb2, argb3, const0, const1, const2, shift, \                y_out)                                                     \  {                                                                        \    v16u8 vec0_m, vec1_m, vec2_m, vec3_m;                                  \    v8u16 reg0_m, reg1_m;                                                  \                                                                           \    vec0_m = (v16u8)__msa_pckev_h((v8i16)argb1, (v8i16)argb0);             \    vec1_m = (v16u8)__msa_pckev_h((v8i16)argb3, (v8i16)argb2);             \    vec2_m = (v16u8)__msa_pckod_h((v8i16)argb1, (v8i16)argb0);             \    vec3_m = (v16u8)__msa_pckod_h((v8i16)argb3, (v8i16)argb2);             \    reg0_m = __msa_dotp_u_h(vec0_m, const0);                               \    reg1_m = __msa_dotp_u_h(vec1_m, const0);                               \    reg0_m = __msa_dpadd_u_h(reg0_m, vec2_m, const1);                      \    reg1_m = __msa_dpadd_u_h(reg1_m, vec3_m, const1);                      \    reg0_m += const2;                                                      \    reg1_m += const2;                                                      \    reg0_m = (v8u16)__msa_srai_h((v8i16)reg0_m, shift);                    \    reg1_m = (v8u16)__msa_srai_h((v8i16)reg1_m, shift);                    \    y_out = (v16u8)__msa_pckev_b((v16i8)reg1_m, (v16i8)reg0_m);            \  } // Loads current and next row of ARGB input and averages it to calculate U and V#define READ_ARGB(s_ptr, t_ptr, argb0, argb1, argb2, argb3)               \  {                                                                       \    v16u8 src0_m, src1_m, src2_m, src3_m, src4_m, src5_m, src6_m, src7_m; \    v16u8 vec0_m, vec1_m, vec2_m, vec3_m, vec4_m, vec5_m, vec6_m, vec7_m; \    v16u8 vec8_m, vec9_m;                                                 \    v8u16 reg0_m, reg1_m, reg2_m, reg3_m, reg4_m, reg5_m, reg6_m, reg7_m; \    v8u16 reg8_m, reg9_m;                                                 \                                                                          \    src0_m = (v16u8)__msa_ld_b((v16i8*)s, 0);                             \    src1_m = (v16u8)__msa_ld_b((v16i8*)s, 16);                            \    src2_m = (v16u8)__msa_ld_b((v16i8*)s, 32);                            \    src3_m = (v16u8)__msa_ld_b((v16i8*)s, 48);                            \    src4_m = (v16u8)__msa_ld_b((v16i8*)t, 0);                             \    src5_m = (v16u8)__msa_ld_b((v16i8*)t, 16);                            \    src6_m = (v16u8)__msa_ld_b((v16i8*)t, 32);                            \    src7_m = (v16u8)__msa_ld_b((v16i8*)t, 48);                            \    vec0_m = (v16u8)__msa_ilvr_b((v16i8)src0_m, (v16i8)src4_m);           \    vec1_m = (v16u8)__msa_ilvr_b((v16i8)src1_m, (v16i8)src5_m);           \    vec2_m = (v16u8)__msa_ilvr_b((v16i8)src2_m, (v16i8)src6_m);           \    vec3_m = (v16u8)__msa_ilvr_b((v16i8)src3_m, (v16i8)src7_m);           \    vec4_m = (v16u8)__msa_ilvl_b((v16i8)src0_m, (v16i8)src4_m);           \    vec5_m = (v16u8)__msa_ilvl_b((v16i8)src1_m, (v16i8)src5_m);           \    vec6_m = (v16u8)__msa_ilvl_b((v16i8)src2_m, (v16i8)src6_m);           \    vec7_m = (v16u8)__msa_ilvl_b((v16i8)src3_m, (v16i8)src7_m);           \    reg0_m = __msa_hadd_u_h(vec0_m, vec0_m);                              \    reg1_m = __msa_hadd_u_h(vec1_m, vec1_m);                              \    reg2_m = __msa_hadd_u_h(vec2_m, vec2_m);                              \    reg3_m = __msa_hadd_u_h(vec3_m, vec3_m);                              \    reg4_m = __msa_hadd_u_h(vec4_m, vec4_m);                              \    reg5_m = __msa_hadd_u_h(vec5_m, vec5_m);                              \    reg6_m = __msa_hadd_u_h(vec6_m, vec6_m);                              \    reg7_m = __msa_hadd_u_h(vec7_m, vec7_m);                              \    reg8_m = (v8u16)__msa_pckev_d((v2i64)reg4_m, (v2i64)reg0_m);          \    reg9_m = (v8u16)__msa_pckev_d((v2i64)reg5_m, (v2i64)reg1_m);          \    reg8_m += (v8u16)__msa_pckod_d((v2i64)reg4_m, (v2i64)reg0_m);         \    reg9_m += (v8u16)__msa_pckod_d((v2i64)reg5_m, (v2i64)reg1_m);         \    reg0_m = (v8u16)__msa_pckev_d((v2i64)reg6_m, (v2i64)reg2_m);          \    reg1_m = (v8u16)__msa_pckev_d((v2i64)reg7_m, (v2i64)reg3_m);          \    reg0_m += (v8u16)__msa_pckod_d((v2i64)reg6_m, (v2i64)reg2_m);         \    reg1_m += (v8u16)__msa_pckod_d((v2i64)reg7_m, (v2i64)reg3_m);         \    reg8_m = (v8u16)__msa_srai_h((v8i16)reg8_m, 2);                       \    reg9_m = (v8u16)__msa_srai_h((v8i16)reg9_m, 2);                       \    reg0_m = (v8u16)__msa_srai_h((v8i16)reg0_m, 2);                       \    reg1_m = (v8u16)__msa_srai_h((v8i16)reg1_m, 2);                       \    argb0 = (v16u8)__msa_pckev_b((v16i8)reg9_m, (v16i8)reg8_m);           \    argb1 = (v16u8)__msa_pckev_b((v16i8)reg1_m, (v16i8)reg0_m);           \    src0_m = (v16u8)__msa_ld_b((v16i8*)s, 64);                            \    src1_m = (v16u8)__msa_ld_b((v16i8*)s, 80);                            \    src2_m = (v16u8)__msa_ld_b((v16i8*)s, 96);                            \    src3_m = (v16u8)__msa_ld_b((v16i8*)s, 112);                           \    src4_m = (v16u8)__msa_ld_b((v16i8*)t, 64);                            \    src5_m = (v16u8)__msa_ld_b((v16i8*)t, 80);                            \    src6_m = (v16u8)__msa_ld_b((v16i8*)t, 96);                            \    src7_m = (v16u8)__msa_ld_b((v16i8*)t, 112);                           \    vec2_m = (v16u8)__msa_ilvr_b((v16i8)src0_m, (v16i8)src4_m);           \    vec3_m = (v16u8)__msa_ilvr_b((v16i8)src1_m, (v16i8)src5_m);           \    vec4_m = (v16u8)__msa_ilvr_b((v16i8)src2_m, (v16i8)src6_m);           \    vec5_m = (v16u8)__msa_ilvr_b((v16i8)src3_m, (v16i8)src7_m);           \    vec6_m = (v16u8)__msa_ilvl_b((v16i8)src0_m, (v16i8)src4_m);           \    vec7_m = (v16u8)__msa_ilvl_b((v16i8)src1_m, (v16i8)src5_m);           \    vec8_m = (v16u8)__msa_ilvl_b((v16i8)src2_m, (v16i8)src6_m);           \    vec9_m = (v16u8)__msa_ilvl_b((v16i8)src3_m, (v16i8)src7_m);           \    reg0_m = __msa_hadd_u_h(vec2_m, vec2_m);                              \    reg1_m = __msa_hadd_u_h(vec3_m, vec3_m);                              \    reg2_m = __msa_hadd_u_h(vec4_m, vec4_m);                              \    reg3_m = __msa_hadd_u_h(vec5_m, vec5_m);                              \    reg4_m = __msa_hadd_u_h(vec6_m, vec6_m);                              \    reg5_m = __msa_hadd_u_h(vec7_m, vec7_m);                              \    reg6_m = __msa_hadd_u_h(vec8_m, vec8_m);                              \    reg7_m = __msa_hadd_u_h(vec9_m, vec9_m);                              \    reg8_m = (v8u16)__msa_pckev_d((v2i64)reg4_m, (v2i64)reg0_m);          \    reg9_m = (v8u16)__msa_pckev_d((v2i64)reg5_m, (v2i64)reg1_m);          \    reg8_m += (v8u16)__msa_pckod_d((v2i64)reg4_m, (v2i64)reg0_m);         \    reg9_m += (v8u16)__msa_pckod_d((v2i64)reg5_m, (v2i64)reg1_m);         \    reg0_m = (v8u16)__msa_pckev_d((v2i64)reg6_m, (v2i64)reg2_m);          \    reg1_m = (v8u16)__msa_pckev_d((v2i64)reg7_m, (v2i64)reg3_m);          \    reg0_m += (v8u16)__msa_pckod_d((v2i64)reg6_m, (v2i64)reg2_m);         \    reg1_m += (v8u16)__msa_pckod_d((v2i64)reg7_m, (v2i64)reg3_m);         \    reg8_m = (v8u16)__msa_srai_h((v8i16)reg8_m, 2);                       \    reg9_m = (v8u16)__msa_srai_h((v8i16)reg9_m, 2);                       \    reg0_m = (v8u16)__msa_srai_h((v8i16)reg0_m, 2);                       \    reg1_m = (v8u16)__msa_srai_h((v8i16)reg1_m, 2);                       \    argb2 = (v16u8)__msa_pckev_b((v16i8)reg9_m, (v16i8)reg8_m);           \    argb3 = (v16u8)__msa_pckev_b((v16i8)reg1_m, (v16i8)reg0_m);           \  } // Takes ARGB input and calculates U and V.#define ARGBTOUV(argb0, argb1, argb2, argb3, const0, const1, const2, const3, \                 shf0, shf1, shf2, shf3, v_out, u_out)                       \  {                                                                          \    v16u8 vec0_m, vec1_m, vec2_m, vec3_m, vec4_m, vec5_m, vec6_m, vec7_m;    \    v8u16 reg0_m, reg1_m, reg2_m, reg3_m;                                    \                                                                             \    vec0_m = (v16u8)__msa_vshf_b(shf0, (v16i8)argb1, (v16i8)argb0);          \    vec1_m = (v16u8)__msa_vshf_b(shf0, (v16i8)argb3, (v16i8)argb2);          \    vec2_m = (v16u8)__msa_vshf_b(shf1, (v16i8)argb1, (v16i8)argb0);          \    vec3_m = (v16u8)__msa_vshf_b(shf1, (v16i8)argb3, (v16i8)argb2);          \    vec4_m = (v16u8)__msa_vshf_b(shf2, (v16i8)argb1, (v16i8)argb0);          \    vec5_m = (v16u8)__msa_vshf_b(shf2, (v16i8)argb3, (v16i8)argb2);          \    vec6_m = (v16u8)__msa_vshf_b(shf3, (v16i8)argb1, (v16i8)argb0);          \    vec7_m = (v16u8)__msa_vshf_b(shf3, (v16i8)argb3, (v16i8)argb2);          \    reg0_m = __msa_dotp_u_h(vec0_m, const1);                                 \    reg1_m = __msa_dotp_u_h(vec1_m, const1);                                 \    reg2_m = __msa_dotp_u_h(vec4_m, const1);                                 \    reg3_m = __msa_dotp_u_h(vec5_m, const1);                                 \    reg0_m += const3;                                                        \    reg1_m += const3;                                                        \    reg2_m += const3;                                                        \    reg3_m += const3;                                                        \    reg0_m -= __msa_dotp_u_h(vec2_m, const0);                                \    reg1_m -= __msa_dotp_u_h(vec3_m, const0);                                \    reg2_m -= __msa_dotp_u_h(vec6_m, const2);                                \    reg3_m -= __msa_dotp_u_h(vec7_m, const2);                                \    v_out = (v16u8)__msa_pckod_b((v16i8)reg1_m, (v16i8)reg0_m);              \    u_out = (v16u8)__msa_pckod_b((v16i8)reg3_m, (v16i8)reg2_m);              \  } // Load I444 pixel data#define READI444(psrc_y, psrc_u, psrc_v, out_y, out_u, out_v) \  {                                                           \    uint64_t y_m, u_m, v_m;                                   \    v2i64 zero_m = {0};                                       \    y_m = LD(psrc_y);                                         \    u_m = LD(psrc_u);                                         \    v_m = LD(psrc_v);                                         \    out_y = (v16u8)__msa_insert_d(zero_m, 0, (int64_t)y_m);   \    out_u = (v16u8)__msa_insert_d(zero_m, 0, (int64_t)u_m);   \    out_v = (v16u8)__msa_insert_d(zero_m, 0, (int64_t)v_m);   \  } void MirrorRow_MSA(const uint8_t* src, uint8_t* dst, int width) {  int x;  v16u8 src0, src1, src2, src3;  v16u8 dst0, dst1, dst2, dst3;  v16i8 shuffler = {15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0};  src += width - 64;   for (x = 0; x < width; x += 64) {    LD_UB4(src, 16, src3, src2, src1, src0);    VSHF_B2_UB(src3, src3, src2, src2, shuffler, shuffler, dst3, dst2);    VSHF_B2_UB(src1, src1, src0, src0, shuffler, shuffler, dst1, dst0);    ST_UB4(dst0, dst1, dst2, dst3, dst, 16);    dst += 64;    src -= 64;  }} void ARGBMirrorRow_MSA(const uint8_t* src, uint8_t* dst, int width) {  int x;  v16u8 src0, src1, src2, src3;  v16u8 dst0, dst1, dst2, dst3;  v16i8 shuffler = {12, 13, 14, 15, 8, 9, 10, 11, 4, 5, 6, 7, 0, 1, 2, 3};  src += width * 4 - 64;   for (x = 0; x < width; x += 16) {    LD_UB4(src, 16, src3, src2, src1, src0);    VSHF_B2_UB(src3, src3, src2, src2, shuffler, shuffler, dst3, dst2);    VSHF_B2_UB(src1, src1, src0, src0, shuffler, shuffler, dst1, dst0);    ST_UB4(dst0, dst1, dst2, dst3, dst, 16);    dst += 64;    src -= 64;  }} void I422ToYUY2Row_MSA(const uint8_t* src_y,                       const uint8_t* src_u,                       const uint8_t* src_v,                       uint8_t* dst_yuy2,                       int width) {  int x;  v16u8 src_u0, src_v0, src_y0, src_y1, vec_uv0, vec_uv1;  v16u8 dst_yuy2_0, dst_yuy2_1, dst_yuy2_2, dst_yuy2_3;   for (x = 0; x < width; x += 32) {    src_u0 = LD_UB(src_u);    src_v0 = LD_UB(src_v);    LD_UB2(src_y, 16, src_y0, src_y1);    ILVRL_B2_UB(src_v0, src_u0, vec_uv0, vec_uv1);    ILVRL_B2_UB(vec_uv0, src_y0, dst_yuy2_0, dst_yuy2_1);    ILVRL_B2_UB(vec_uv1, src_y1, dst_yuy2_2, dst_yuy2_3);    ST_UB4(dst_yuy2_0, dst_yuy2_1, dst_yuy2_2, dst_yuy2_3, dst_yuy2, 16);    src_u += 16;    src_v += 16;    src_y += 32;    dst_yuy2 += 64;  }} void I422ToUYVYRow_MSA(const uint8_t* src_y,                       const uint8_t* src_u,                       const uint8_t* src_v,                       uint8_t* dst_uyvy,                       int width) {  int x;  v16u8 src_u0, src_v0, src_y0, src_y1, vec_uv0, vec_uv1;  v16u8 dst_uyvy0, dst_uyvy1, dst_uyvy2, dst_uyvy3;   for (x = 0; x < width; x += 32) {    src_u0 = LD_UB(src_u);    src_v0 = LD_UB(src_v);    LD_UB2(src_y, 16, src_y0, src_y1);    ILVRL_B2_UB(src_v0, src_u0, vec_uv0, vec_uv1);    ILVRL_B2_UB(src_y0, vec_uv0, dst_uyvy0, dst_uyvy1);    ILVRL_B2_UB(src_y1, vec_uv1, dst_uyvy2, dst_uyvy3);    ST_UB4(dst_uyvy0, dst_uyvy1, dst_uyvy2, dst_uyvy3, dst_uyvy, 16);    src_u += 16;    src_v += 16;    src_y += 32;    dst_uyvy += 64;  }} void I422ToARGBRow_MSA(const uint8_t* src_y,                       const uint8_t* src_u,                       const uint8_t* src_v,                       uint8_t* dst_argb,                       const struct YuvConstants* yuvconstants,                       int width) {  int x;  v16u8 src0, src1, src2;  v8i16 vec0, vec1, vec2;  v4i32 vec_ub, vec_vr, vec_ug, vec_vg, vec_bb, vec_bg, vec_br, vec_yg;  v4i32 vec_ubvr, vec_ugvg;  v16u8 alpha = (v16u8)__msa_ldi_b(ALPHA_VAL);   YUVTORGB_SETUP(yuvconstants, vec_ub, vec_vr, vec_ug, vec_vg, vec_bb, vec_bg,                 vec_br, vec_yg);  vec_ubvr = __msa_ilvr_w(vec_vr, vec_ub);  vec_ugvg = (v4i32)__msa_ilvev_h((v8i16)vec_vg, (v8i16)vec_ug);   for (x = 0; x < width; x += 8) {    READYUV422(src_y, src_u, src_v, src0, src1, src2);    src1 = (v16u8)__msa_ilvr_b((v16i8)src2, (v16i8)src1);    YUVTORGB(src0, src1, vec_ubvr, vec_ugvg, vec_bb, vec_bg, vec_br, vec_yg,             vec0, vec1, vec2);    STOREARGB(vec0, vec1, vec2, alpha, dst_argb);    src_y += 8;    src_u += 4;    src_v += 4;    dst_argb += 32;  }} void I422ToRGBARow_MSA(const uint8_t* src_y,                       const uint8_t* src_u,                       const uint8_t* src_v,                       uint8_t* dst_argb,                       const struct YuvConstants* yuvconstants,                       int width) {  int x;  v16u8 src0, src1, src2;  v8i16 vec0, vec1, vec2;  v4i32 vec_ub, vec_vr, vec_ug, vec_vg, vec_bb, vec_bg, vec_br, vec_yg;  v4i32 vec_ubvr, vec_ugvg;  v16u8 alpha = (v16u8)__msa_ldi_b(ALPHA_VAL);   YUVTORGB_SETUP(yuvconstants, vec_ub, vec_vr, vec_ug, vec_vg, vec_bb, vec_bg,                 vec_br, vec_yg);  vec_ubvr = __msa_ilvr_w(vec_vr, vec_ub);  vec_ugvg = (v4i32)__msa_ilvev_h((v8i16)vec_vg, (v8i16)vec_ug);   for (x = 0; x < width; x += 8) {    READYUV422(src_y, src_u, src_v, src0, src1, src2);    src1 = (v16u8)__msa_ilvr_b((v16i8)src2, (v16i8)src1);    YUVTORGB(src0, src1, vec_ubvr, vec_ugvg, vec_bb, vec_bg, vec_br, vec_yg,             vec0, vec1, vec2);    STOREARGB(alpha, vec0, vec1, vec2, dst_argb);    src_y += 8;    src_u += 4;    src_v += 4;    dst_argb += 32;  }} void I422AlphaToARGBRow_MSA(const uint8_t* src_y,                            const uint8_t* src_u,                            const uint8_t* src_v,                            const uint8_t* src_a,                            uint8_t* dst_argb,                            const struct YuvConstants* yuvconstants,                            int width) {  int x;  int64_t data_a;  v16u8 src0, src1, src2, src3;  v8i16 vec0, vec1, vec2;  v4i32 vec_ub, vec_vr, vec_ug, vec_vg, vec_bb, vec_bg, vec_br, vec_yg;  v4i32 vec_ubvr, vec_ugvg;  v4i32 zero = {0};   YUVTORGB_SETUP(yuvconstants, vec_ub, vec_vr, vec_ug, vec_vg, vec_bb, vec_bg,                 vec_br, vec_yg);  vec_ubvr = __msa_ilvr_w(vec_vr, vec_ub);  vec_ugvg = (v4i32)__msa_ilvev_h((v8i16)vec_vg, (v8i16)vec_ug);   for (x = 0; x < width; x += 8) {    data_a = LD(src_a);    READYUV422(src_y, src_u, src_v, src0, src1, src2);    src1 = (v16u8)__msa_ilvr_b((v16i8)src2, (v16i8)src1);    src3 = (v16u8)__msa_insert_d((v2i64)zero, 0, data_a);    YUVTORGB(src0, src1, vec_ubvr, vec_ugvg, vec_bb, vec_bg, vec_br, vec_yg,             vec0, vec1, vec2);    src3 = (v16u8)__msa_ilvr_b((v16i8)src3, (v16i8)src3);    STOREARGB(vec0, vec1, vec2, src3, dst_argb);    src_y += 8;    src_u += 4;    src_v += 4;    src_a += 8;    dst_argb += 32;  }} void I422ToRGB24Row_MSA(const uint8_t* src_y,                        const uint8_t* src_u,                        const uint8_t* src_v,                        uint8_t* dst_argb,                        const struct YuvConstants* yuvconstants,                        int32_t width) {  int x;  int64_t data_u, data_v;  v16u8 src0, src1, src2, src3, src4, dst0, dst1, dst2;  v8i16 vec0, vec1, vec2, vec3, vec4, vec5;  v4i32 vec_ub, vec_vr, vec_ug, vec_vg, vec_bb, vec_bg, vec_br, vec_yg;  v4i32 vec_ubvr, vec_ugvg;  v16u8 reg0, reg1, reg2, reg3;  v2i64 zero = {0};  v16i8 shuffler0 = {0, 1, 16, 2, 3, 17, 4, 5, 18, 6, 7, 19, 8, 9, 20, 10};  v16i8 shuffler1 = {0, 21, 1, 2, 22, 3, 4, 23, 5, 6, 24, 7, 8, 25, 9, 10};  v16i8 shuffler2 = {26, 6,  7,  27, 8,  9,  28, 10,                     11, 29, 12, 13, 30, 14, 15, 31};   YUVTORGB_SETUP(yuvconstants, vec_ub, vec_vr, vec_ug, vec_vg, vec_bb, vec_bg,                 vec_br, vec_yg);  vec_ubvr = __msa_ilvr_w(vec_vr, vec_ub);  vec_ugvg = (v4i32)__msa_ilvev_h((v8i16)vec_vg, (v8i16)vec_ug);   for (x = 0; x < width; x += 16) {    src0 = (v16u8)__msa_ld_b((v16u8*)src_y, 0);    data_u = LD(src_u);    data_v = LD(src_v);    src1 = (v16u8)__msa_insert_d(zero, 0, data_u);    src2 = (v16u8)__msa_insert_d(zero, 0, data_v);    src1 = (v16u8)__msa_ilvr_b((v16i8)src2, (v16i8)src1);    src3 = (v16u8)__msa_sldi_b((v16i8)src0, (v16i8)src0, 8);

Lines 1–500 of 3512. Load more