prepare for adaptive color transform
Showing the changes to libavcodec (3 of 3 files). Show all
Unified · Split
| @@ -122,7 +122,6 @@ typedef struct VVCIntraDSPContext { · 121 unmodified lines | ||
| 122 | 122 | |
| 123 | 123 | typedef struct VVCItxDSPContext { |
| 124 | 124 | void (*add_residual)(uint8_t *dst, const int *res, int width, int height, ptrdiff_t stride); |
| 125 | − void (*add_residual_joint)(uint8_t *dst, const int *res, int width, int height, ptrdiff_t stride, int c_sign, int shift); | |
| 126 | 125 | void (*pred_residual_joint)(int *dst, const int *src, int width, int height, int c_sign, int shift); |
| 127 | 126 | |
| 128 | 127 | void (*itx[VVC_N_TX_TYPE][VVC_N_TX_SIZE])(int *coeffs, ptrdiff_t step, size_t nz); |
| @@ -45,23 +45,6 @@ static void FUNC(add_residual)(uint8_t *_dst, const int *res, · 44 unmodified lines | ||
| 45 | 45 | } |
| 46 | 46 | } |
| 47 | 47 | |
| 48 | −static void FUNC(add_residual_joint)(uint8_t *_dst, const int *res, | |
| 49 | − const int w, const int h, const ptrdiff_t _stride, const int c_sign, const int shift) | |
| 50 | −{ | |
| 51 | − pixel *dst = (pixel *)_dst; | |
| 52 | − | |
| 53 | − const int stride = _stride / sizeof(pixel); | |
| 54 | − | |
| 55 | − for (int y = 0; y < h; y++) { | |
| 56 | − for (int x = 0; x < w; x++) { | |
| 57 | − const int r = ((*res) * c_sign) >> shift; | |
| 58 | − dst[x] = av_clip_pixel(dst[x] + r); | |
| 59 | − res++; | |
| 60 | − } | |
| 61 | − dst += stride; | |
| 62 | − } | |
| 63 | −} | |
| 64 | − | |
| 65 | 48 | static void FUNC(pred_residual_joint)(int *dst, const int *src, const int w, const int h, |
| 66 | 49 | const int c_sign, const int shift) |
| 67 | 50 | { |
| @@ -121,7 +104,6 @@ static void FUNC(ff_vvc_itx_dsp_init)(VVCItxDSPContext *const itx) · 53 unmodified lines | ||
| 121 | 104 | VVC_ITX(TYPE, type, 32); |
| 122 | 105 | |
| 123 | 106 | itx->add_residual = FUNC(add_residual); |
| 124 | − itx->add_residual_joint = FUNC(add_residual_joint); | |
| 125 | 107 | itx->pred_residual_joint = FUNC(pred_residual_joint); |
| 126 | 108 | itx->transform_bdpcm = FUNC(transform_bdpcm); |
| 127 | 109 | VVC_ITX(DCT2, dct2, 2) |
| @@ -164,28 +164,6 @@ static void derive_transform_type(const VVCFrameContext *fc, const VVCLocalConte · 163 unmodified lines | ||
| 164 | 164 | *trv = mts_to_trv[cu->mts_idx]; |
| 165 | 165 | } |
| 166 | 166 | |
| 167 | −static void add_residual_for_joint_coding_chroma(VVCLocalContext *lc, | |
| 168 | − const TransformUnit *tu, TransformBlock *tb, const int chroma_scale) | |
| 169 | −{ | |
| 170 | − const VVCFrameContext *fc = lc->fc; | |
| 171 | − const CodingUnit *cu = lc->cu; | |
| 172 | − const int c_sign = 1 - 2 * fc->ps.ph.r->ph_joint_cbcr_sign_flag; | |
| 173 | − const int shift = tu->coded_flag[1] ^ tu->coded_flag[2]; | |
| 174 | − const int c_idx = 1 + tu->coded_flag[1]; | |
| 175 | − const ptrdiff_t stride = fc->frame->linesize[c_idx]; | |
| 176 | − const int hs = fc->ps.sps->hshift[c_idx]; | |
| 177 | − const int vs = fc->ps.sps->vshift[c_idx]; | |
| 178 | − uint8_t *dst = &fc->frame->data[c_idx][(tb->y0 >> vs) * stride + | |
| 179 | − ((tb->x0 >> hs) << fc->ps.sps->pixel_shift)]; | |
| 180 | − if (chroma_scale) { | |
| 181 | − fc->vvcdsp.itx.pred_residual_joint(tb->coeffs, tb->coeffs, tb->tb_width, tb->tb_height, c_sign, shift); | |
| 182 | − fc->vvcdsp.intra.lmcs_scale_chroma(lc, tb->coeffs, tb->coeffs, tb->tb_width, tb->tb_height, cu->x0, cu->y0); | |
| 183 | − fc->vvcdsp.itx.add_residual(dst, tb->coeffs, tb->tb_width, tb->tb_height, stride); | |
| 184 | − } else { | |
| 185 | − fc->vvcdsp.itx.add_residual_joint(dst, tb->coeffs, tb->tb_width, tb->tb_height, stride, c_sign, shift); | |
| 186 | − } | |
| 187 | −} | |
| 188 | − | |
| 189 | 167 | static int add_reconstructed_area(VVCLocalContext *lc, const int ch_type, const int x0, const int y0, const int w, const int h) |
| 190 | 168 | { |
| 191 | 169 | const VVCSPS *sps = lc->fc->ps.sps; |
| @@ -531,7 +509,7 @@ static void itransform(VVCLocalContext *lc, TransformUnit *tu, const int tu_idx, · 339 unmodified lines | ||
| 531 | 509 | const ptrdiff_t stride = fc->frame->linesize[c_idx]; |
| 532 | 510 | const int hs = sps->hshift[c_idx]; |
| 533 | 511 | const int vs = sps->vshift[c_idx]; |
| 534 | − uint8_t *dst = &fc->frame->data[c_idx][(tb->y0 >> vs) * stride + ((tb->x0 >> hs) << ps)]; | |
| 512 | + const int has_jcbcr = tu->joint_cbcr_residual_flag && c_idx; | |
| 535 | 513 | |
| 536 | 514 | if (cu->bdpcm_flag[tb->c_idx]) |
| 537 | 515 | transform_bdpcm(tb, lc, cu); |
| @@ -548,14 +526,25 @@ static void itransform(VVCLocalContext *lc, TransformUnit *tu, const int tu_idx, · 10 unmodified lines | ||
| 548 | 526 | itx_1d(fc, tb, trh, trv); |
| 549 | 527 | } |
| 550 | 528 | |
| 551 | − if (chroma_scale) | |
| 552 | − fc->vvcdsp.intra.lmcs_scale_chroma(lc, temp, tb->coeffs, w, h, cu->x0, cu->y0); | |
| 553 | − // TODO: Address performance issue here by combining transform, lmcs_scale_chroma, and add_residual into one function. | |
| 554 | − // Complete this task before implementing ASM code. | |
| 555 | − fc->vvcdsp.itx.add_residual(dst, chroma_scale ? temp : tb->coeffs, w, h, stride); | |
| 556 | − | |
| 557 | − if (tu->joint_cbcr_residual_flag && tb->c_idx) | |
| 558 | − add_residual_for_joint_coding_chroma(lc, tu, tb, chroma_scale); | |
| 529 | + for (int j = 0; j < 1 + has_jcbcr; j++) { | |
| 530 | + const bool is_jcbcr = j > 0; | |
| 531 | + const int jcbcr_idx = CB + tu->coded_flag[CB]; | |
| 532 | + TransformBlock *jcbcr = &tu->tbs[jcbcr_idx - tu->tbs[0].c_idx]; | |
| 533 | + const int c = is_jcbcr ? jcbcr_idx : tb->c_idx; | |
| 534 | + int *coeffs = is_jcbcr ? jcbcr->coeffs : tb->coeffs; | |
| 535 | + uint8_t *dst = &fc->frame->data[c][(tb->y0 >> vs) * stride + ((tb->x0 >> hs) << ps)]; | |
| 536 | + | |
| 537 | + if (!j && has_jcbcr) { | |
| 538 | + const int c_sign = 1 - 2 * fc->ps.ph.r->ph_joint_cbcr_sign_flag; | |
| 539 | + const int shift = tu->coded_flag[CB] ^ tu->coded_flag[CR]; | |
| 540 | + fc->vvcdsp.itx.pred_residual_joint(jcbcr->coeffs, tb->coeffs, tb->tb_width, tb->tb_height, c_sign, shift); | |
| 541 | + } | |
| 542 | + if (chroma_scale) | |
| 543 | + fc->vvcdsp.intra.lmcs_scale_chroma(lc, temp, coeffs, w, h, cu->x0, cu->y0); | |
| 544 | + // TODO: Address performance issue here by combining transform, lmcs_scale_chroma, and add_residual into one function. | |
| 545 | + // Complete this task before implementing ASM code. | |
| 546 | + fc->vvcdsp.itx.add_residual(dst, chroma_scale ? temp : coeffs, w, h, stride); | |
| 547 | + } | |
| 559 | 548 | } |
| 560 | 549 | } |
| 561 | 550 | } |