2 * Copyright (C) 2015 Pedro Arthur <bygrandao@gmail.com>
4 * This file is part of FFmpeg.
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 #include "swscale_internal.h"
22 typedef struct VScalerContext
29 yuv2packedX_fn yuv2packedX;
33 static int lum_planar_vscale(SwsContext *c, SwsFilterDescriptor *desc, int sliceY, int sliceH)
35 VScalerContext *inst = desc->instance;
36 int dstW = desc->dst->width;
38 int first = FFMAX(1-inst->filter_size, inst->filter_pos[sliceY]);
39 int sp = first - desc->src->plane[0].sliceY;
40 int dp = sliceY - desc->dst->plane[0].sliceY;
41 uint8_t **src = desc->src->plane[0].line + sp;
42 uint8_t **dst = desc->dst->plane[0].line + dp;
43 uint16_t *filter = inst->filter[0] + (inst->isMMX ? 0 : sliceY * inst->filter_size);
45 if (inst->filter_size == 1)
46 ((yuv2planar1_fn)inst->pfn)((const int16_t*)src[0], dst[0], dstW, c->lumDither8, 0);
48 ((yuv2planarX_fn)inst->pfn)(filter, inst->filter_size, (const int16_t**)src, dst[0], dstW, c->lumDither8, 0);
51 int sp = first - desc->src->plane[3].sliceY;
52 int dp = sliceY - desc->dst->plane[3].sliceY;
53 uint8_t **src = desc->src->plane[3].line + sp;
54 uint8_t **dst = desc->dst->plane[3].line + dp;
55 uint16_t *filter = inst->filter[1] + (inst->isMMX ? 0 : sliceY * inst->filter_size);
57 if (inst->filter_size == 1)
58 ((yuv2planar1_fn)inst->pfn)((const int16_t*)src[0], dst[0], dstW, c->lumDither8, 0);
60 ((yuv2planarX_fn)inst->pfn)(filter, inst->filter_size, (const int16_t**)src, dst[0], dstW, c->lumDither8, 0);
66 static int chr_planar_vscale(SwsContext *c, SwsFilterDescriptor *desc, int sliceY, int sliceH)
68 const int chrSkipMask = (1 << desc->dst->v_chr_sub_sample) - 1;
69 if (sliceY & chrSkipMask)
72 VScalerContext *inst = desc->instance;
73 int dstW = AV_CEIL_RSHIFT(desc->dst->width, desc->dst->h_chr_sub_sample);
74 int chrSliceY = sliceY >> desc->dst->v_chr_sub_sample;
76 int first = FFMAX(1-inst->filter_size, inst->filter_pos[chrSliceY]);
77 int sp1 = first - desc->src->plane[1].sliceY;
78 int sp2 = first - desc->src->plane[2].sliceY;
79 int dp1 = chrSliceY - desc->dst->plane[1].sliceY;
80 int dp2 = chrSliceY - desc->dst->plane[2].sliceY;
81 uint8_t **src1 = desc->src->plane[1].line + sp1;
82 uint8_t **src2 = desc->src->plane[2].line + sp2;
83 uint8_t **dst1 = desc->dst->plane[1].line + dp1;
84 uint8_t **dst2 = desc->dst->plane[2].line + dp2;
85 uint16_t *filter = inst->filter[0] + (inst->isMMX ? 0 : chrSliceY * inst->filter_size);
88 ((yuv2interleavedX_fn)inst->pfn)(c, filter, inst->filter_size, (const int16_t**)src1, (const int16_t**)src2, dst1[0], dstW);
89 } else if (inst->filter_size == 1) {
90 ((yuv2planar1_fn)inst->pfn)((const int16_t*)src1[0], dst1[0], dstW, c->chrDither8, 0);
91 ((yuv2planar1_fn)inst->pfn)((const int16_t*)src2[0], dst2[0], dstW, c->chrDither8, 3);
93 ((yuv2planarX_fn)inst->pfn)(filter, inst->filter_size, (const int16_t**)src1, dst1[0], dstW, c->chrDither8, 0);
94 ((yuv2planarX_fn)inst->pfn)(filter, inst->filter_size, (const int16_t**)src2, dst2[0], dstW, c->chrDither8, inst->isMMX ? (c->uv_offx2 >> 1) : 3);
101 static int packed_vscale(SwsContext *c, SwsFilterDescriptor *desc, int sliceY, int sliceH)
103 VScalerContext *inst = desc->instance;
104 int dstW = desc->dst->width;
105 int chrSliceY = sliceY >> desc->dst->v_chr_sub_sample;
107 int lum_fsize = inst[0].filter_size;
108 int chr_fsize = inst[1].filter_size;
109 uint16_t *lum_filter = inst[0].filter[0];
110 uint16_t *chr_filter = inst[1].filter[0];
112 int firstLum = FFMAX(1-lum_fsize, inst[0].filter_pos[ sliceY]);
113 int firstChr = FFMAX(1-chr_fsize, inst[1].filter_pos[chrSliceY]);
115 int sp0 = firstLum - desc->src->plane[0].sliceY;
116 int sp1 = firstChr - desc->src->plane[1].sliceY;
117 int sp2 = firstChr - desc->src->plane[2].sliceY;
118 int sp3 = firstLum - desc->src->plane[3].sliceY;
119 int dp = sliceY - desc->dst->plane[0].sliceY;
120 uint8_t **src0 = desc->src->plane[0].line + sp0;
121 uint8_t **src1 = desc->src->plane[1].line + sp1;
122 uint8_t **src2 = desc->src->plane[2].line + sp2;
123 uint8_t **src3 = desc->alpha ? desc->src->plane[3].line + sp3 : NULL;
124 uint8_t **dst = desc->dst->plane[0].line + dp;
127 if (c->yuv2packed1 && lum_fsize == 1 && chr_fsize == 1) { // unscaled RGB
128 ((yuv2packed1_fn)inst->pfn)(c, (const int16_t*)*src0, (const int16_t**)src1, (const int16_t**)src2,
129 (const int16_t*)(desc->alpha ? *src3 : NULL), *dst, dstW, 0, sliceY);
130 } else if (c->yuv2packed1 && lum_fsize == 1 && chr_fsize == 2 &&
131 chr_filter[2 * chrSliceY + 1] + chr_filter[2 * chrSliceY] == 4096 &&
132 chr_filter[2 * chrSliceY + 1] <= 4096U) { // unscaled RGB
133 int chrAlpha = chr_filter[2 * chrSliceY + 1];
134 ((yuv2packed1_fn)inst->pfn)(c, (const int16_t*)*src0, (const int16_t**)src1, (const int16_t**)src2,
135 (const int16_t*)(desc->alpha ? *src3 : NULL), *dst, dstW, chrAlpha, sliceY);
136 } else if (c->yuv2packed2 && lum_fsize == 2 && chr_fsize == 2 &&
137 lum_filter[2 * sliceY + 1] + lum_filter[2 * sliceY] == 4096 &&
138 lum_filter[2 * sliceY + 1] <= 4096U &&
139 chr_filter[2 * chrSliceY + 1] + chr_filter[2 * chrSliceY] == 4096 &&
140 chr_filter[2 * chrSliceY + 1] <= 4096U
141 ) { // bilinear upscale RGB
142 int lumAlpha = lum_filter[2 * sliceY + 1];
143 int chrAlpha = chr_filter[2 * chrSliceY + 1];
145 c->lumMmxFilter[3] = lum_filter[2 * sliceY] * 0x10001;
147 c->chrMmxFilter[3] = chr_filter[2 * chrSliceY] * 0x10001;
148 ((yuv2packed2_fn)inst->pfn)(c, (const int16_t**)src0, (const int16_t**)src1, (const int16_t**)src2, (const int16_t**)src3,
149 *dst, dstW, lumAlpha, chrAlpha, sliceY);
150 } else { // general RGB
151 if ((c->yuv2packed1 && lum_fsize == 1 && chr_fsize == 2) ||
152 (c->yuv2packed2 && lum_fsize == 2 && chr_fsize == 2)) {
153 if (!c->warned_unuseable_bilinear)
154 av_log(c, AV_LOG_INFO, "Optimized 2 tap filter code cannot be used\n");
155 c->warned_unuseable_bilinear = 1;
158 inst->yuv2packedX(c, lum_filter + sliceY * lum_fsize,
159 (const int16_t**)src0, lum_fsize, chr_filter + chrSliceY * chr_fsize,
160 (const int16_t**)src1, (const int16_t**)src2, chr_fsize, (const int16_t**)src3, *dst, dstW, sliceY);
165 static int any_vscale(SwsContext *c, SwsFilterDescriptor *desc, int sliceY, int sliceH)
167 VScalerContext *inst = desc->instance;
168 int dstW = desc->dst->width;
169 int chrSliceY = sliceY >> desc->dst->v_chr_sub_sample;
171 int lum_fsize = inst[0].filter_size;
172 int chr_fsize = inst[1].filter_size;
173 uint16_t *lum_filter = inst[0].filter[0];
174 uint16_t *chr_filter = inst[1].filter[0];
176 int firstLum = FFMAX(1-lum_fsize, inst[0].filter_pos[ sliceY]);
177 int firstChr = FFMAX(1-chr_fsize, inst[1].filter_pos[chrSliceY]);
179 int sp0 = firstLum - desc->src->plane[0].sliceY;
180 int sp1 = firstChr - desc->src->plane[1].sliceY;
181 int sp2 = firstChr - desc->src->plane[2].sliceY;
182 int sp3 = firstLum - desc->src->plane[3].sliceY;
183 int dp0 = sliceY - desc->dst->plane[0].sliceY;
184 int dp1 = chrSliceY - desc->dst->plane[1].sliceY;
185 int dp2 = chrSliceY - desc->dst->plane[2].sliceY;
186 int dp3 = sliceY - desc->dst->plane[3].sliceY;
188 uint8_t **src0 = desc->src->plane[0].line + sp0;
189 uint8_t **src1 = desc->src->plane[1].line + sp1;
190 uint8_t **src2 = desc->src->plane[2].line + sp2;
191 uint8_t **src3 = desc->alpha ? desc->src->plane[3].line + sp3 : NULL;
192 uint8_t *dst[4] = { desc->dst->plane[0].line[dp0],
193 desc->dst->plane[1].line[dp1],
194 desc->dst->plane[2].line[dp2],
195 desc->alpha ? desc->dst->plane[3].line[dp3] : NULL };
197 av_assert1(!c->yuv2packed1 && !c->yuv2packed2);
198 ((yuv2anyX_fn)inst->pfn)(c, lum_filter + sliceY * lum_fsize,
199 (const int16_t**)src0, lum_fsize, chr_filter + sliceY * chr_fsize,
200 (const int16_t**)src1, (const int16_t**)src2, chr_fsize, (const int16_t**)src3, dst, dstW, sliceY);
206 int ff_init_vscale(SwsContext *c, SwsFilterDescriptor *desc, SwsSlice *src, SwsSlice *dst)
208 VScalerContext *lumCtx = NULL;
209 VScalerContext *chrCtx = NULL;
211 if (isPlanarYUV(c->dstFormat) || (isGray(c->dstFormat) && !isALPHA(c->dstFormat))) {
212 lumCtx = av_mallocz(sizeof(VScalerContext));
214 return AVERROR(ENOMEM);
217 desc[0].process = lum_planar_vscale;
218 desc[0].instance = lumCtx;
221 desc[0].alpha = c->needAlpha;
223 if (!isGray(c->dstFormat)) {
224 chrCtx = av_mallocz(sizeof(VScalerContext));
226 return AVERROR(ENOMEM);
227 desc[1].process = chr_planar_vscale;
228 desc[1].instance = chrCtx;
233 lumCtx = av_mallocz_array(sizeof(VScalerContext), 2);
235 return AVERROR(ENOMEM);
238 desc[0].process = c->yuv2packedX ? packed_vscale : any_vscale;
239 desc[0].instance = lumCtx;
242 desc[0].alpha = c->needAlpha;
245 ff_init_vscale_pfn(c, c->yuv2plane1, c->yuv2planeX, c->yuv2nv12cX,
246 c->yuv2packed1, c->yuv2packed2, c->yuv2packedX, c->yuv2anyX, c->use_mmx_vfilter);
250 void ff_init_vscale_pfn(SwsContext *c,
251 yuv2planar1_fn yuv2plane1,
252 yuv2planarX_fn yuv2planeX,
253 yuv2interleavedX_fn yuv2nv12cX,
254 yuv2packed1_fn yuv2packed1,
255 yuv2packed2_fn yuv2packed2,
256 yuv2packedX_fn yuv2packedX,
257 yuv2anyX_fn yuv2anyX, int use_mmx)
259 VScalerContext *lumCtx = NULL;
260 VScalerContext *chrCtx = NULL;
261 int idx = c->numDesc - (c->is_internal_gamma ? 2 : 1); //FIXME avoid hardcoding indexes
263 if (isPlanarYUV(c->dstFormat) || (isGray(c->dstFormat) && !isALPHA(c->dstFormat))) {
264 if (!isGray(c->dstFormat)) {
265 chrCtx = c->desc[idx].instance;
267 chrCtx->filter[0] = use_mmx ? (int16_t*)c->chrMmxFilter : c->vChrFilter;
268 chrCtx->filter_size = c->vChrFilterSize;
269 chrCtx->filter_pos = c->vChrFilterPos;
270 chrCtx->isMMX = use_mmx;
273 if (yuv2nv12cX) chrCtx->pfn = yuv2nv12cX;
274 else if (c->vChrFilterSize == 1) chrCtx->pfn = yuv2plane1;
275 else chrCtx->pfn = yuv2planeX;
278 lumCtx = c->desc[idx].instance;
280 lumCtx->filter[0] = use_mmx ? (int16_t*)c->lumMmxFilter : c->vLumFilter;
281 lumCtx->filter[1] = use_mmx ? (int16_t*)c->alpMmxFilter : c->vLumFilter;
282 lumCtx->filter_size = c->vLumFilterSize;
283 lumCtx->filter_pos = c->vLumFilterPos;
284 lumCtx->isMMX = use_mmx;
286 if (c->vLumFilterSize == 1) lumCtx->pfn = yuv2plane1;
287 else lumCtx->pfn = yuv2planeX;
290 lumCtx = c->desc[idx].instance;
293 lumCtx->filter[0] = c->vLumFilter;
294 lumCtx->filter_size = c->vLumFilterSize;
295 lumCtx->filter_pos = c->vLumFilterPos;
297 chrCtx->filter[0] = c->vChrFilter;
298 chrCtx->filter_size = c->vChrFilterSize;
299 chrCtx->filter_pos = c->vChrFilterPos;
301 lumCtx->isMMX = use_mmx;
302 chrCtx->isMMX = use_mmx;
305 if (c->yuv2packed1 && c->vLumFilterSize == 1 && c->vChrFilterSize <= 2)
306 lumCtx->pfn = yuv2packed1;
307 else if (c->yuv2packed2 && c->vLumFilterSize == 2 && c->vChrFilterSize == 2)
308 lumCtx->pfn = yuv2packed2;
309 lumCtx->yuv2packedX = yuv2packedX;
311 lumCtx->pfn = yuv2anyX;