vscale.c 12.9 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
/*
 * Copyright (C) 2015 Pedro Arthur <bygrandao@gmail.com>
 *
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * FFmpeg is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with FFmpeg; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */
#include "swscale_internal.h"

22 23 24 25 26 27 28
typedef struct VScalerContext
{
    uint16_t *filter[2];
    int32_t  *filter_pos;
    int filter_size;
    int isMMX;
    void *pfn;
29
    yuv2packedX_fn yuv2packedX;
30 31 32
} VScalerContext;


33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72
static int lum_planar_vscale(SwsContext *c, SwsFilterDescriptor *desc, int sliceY, int sliceH)
{
    VScalerContext *inst = desc->instance;
    int dstW = desc->dst->width;

    int first = FFMAX(1-inst->filter_size, inst->filter_pos[sliceY]);
    int sp = first - desc->src->plane[0].sliceY;
    int dp = sliceY - desc->dst->plane[0].sliceY;
    uint8_t **src = desc->src->plane[0].line + sp;
    uint8_t **dst = desc->dst->plane[0].line + dp;
    uint16_t *filter = inst->filter[0] + (inst->isMMX ? 0 : sliceY * inst->filter_size);

    if (inst->filter_size == 1)
        ((yuv2planar1_fn)inst->pfn)((const int16_t*)src[0], dst[0], dstW, c->lumDither8, 0);
    else
        ((yuv2planarX_fn)inst->pfn)(filter, inst->filter_size, (const int16_t**)src, dst[0], dstW, c->lumDither8, 0);

    if (desc->alpha) {
        int sp = first - desc->src->plane[3].sliceY;
        int dp = sliceY - desc->dst->plane[3].sliceY;
        uint8_t **src = desc->src->plane[3].line + sp;
        uint8_t **dst = desc->dst->plane[3].line + dp;
        uint16_t *filter = inst->filter[1] + (inst->isMMX ? 0 : sliceY * inst->filter_size);

        if (inst->filter_size == 1)
            ((yuv2planar1_fn)inst->pfn)((const int16_t*)src[0], dst[0], dstW, c->lumDither8, 0);
        else
            ((yuv2planarX_fn)inst->pfn)(filter, inst->filter_size, (const int16_t**)src, dst[0], dstW, c->lumDither8, 0);
    }

    return 1;
}

static int chr_planar_vscale(SwsContext *c, SwsFilterDescriptor *desc, int sliceY, int sliceH)
{
    const int chrSkipMask = (1 << desc->dst->v_chr_sub_sample) - 1;
    if (sliceY & chrSkipMask)
        return 0;
    else {
        VScalerContext *inst = desc->instance;
73
        int dstW = AV_CEIL_RSHIFT(desc->dst->width, desc->dst->h_chr_sub_sample);
74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111
        int chrSliceY = sliceY >> desc->dst->v_chr_sub_sample;

        int first = FFMAX(1-inst->filter_size, inst->filter_pos[chrSliceY]);
        int sp1 = first - desc->src->plane[1].sliceY;
        int sp2 = first - desc->src->plane[2].sliceY;
        int dp1 = chrSliceY - desc->dst->plane[1].sliceY;
        int dp2 = chrSliceY - desc->dst->plane[2].sliceY;
        uint8_t **src1 = desc->src->plane[1].line + sp1;
        uint8_t **src2 = desc->src->plane[2].line + sp2;
        uint8_t **dst1 = desc->dst->plane[1].line + dp1;
        uint8_t **dst2 = desc->dst->plane[2].line + dp2;
        uint16_t *filter = inst->filter[0] + (inst->isMMX ? 0 : chrSliceY * inst->filter_size);

        if (c->yuv2nv12cX) {
            ((yuv2interleavedX_fn)inst->pfn)(c, filter, inst->filter_size, (const int16_t**)src1, (const int16_t**)src2, dst1[0], dstW);
        } else if (inst->filter_size == 1) {
            ((yuv2planar1_fn)inst->pfn)((const int16_t*)src1[0], dst1[0], dstW, c->chrDither8, 0);
            ((yuv2planar1_fn)inst->pfn)((const int16_t*)src2[0], dst2[0], dstW, c->chrDither8, 3);
        } else {
            ((yuv2planarX_fn)inst->pfn)(filter, inst->filter_size, (const int16_t**)src1, dst1[0], dstW, c->chrDither8, 0);
            ((yuv2planarX_fn)inst->pfn)(filter, inst->filter_size, (const int16_t**)src2, dst2[0], dstW, c->chrDither8, inst->isMMX ? (c->uv_offx2 >> 1) : 3);
        }
    }

    return 1;
}

static int packed_vscale(SwsContext *c, SwsFilterDescriptor *desc, int sliceY, int sliceH)
{
    VScalerContext *inst = desc->instance;
    int dstW = desc->dst->width;
    int chrSliceY = sliceY >> desc->dst->v_chr_sub_sample;

    int lum_fsize = inst[0].filter_size;
    int chr_fsize = inst[1].filter_size;
    uint16_t *lum_filter = inst[0].filter[0];
    uint16_t *chr_filter = inst[1].filter[0];

112
    int firstLum = FFMAX(1-lum_fsize, inst[0].filter_pos[   sliceY]);
113 114 115 116 117 118 119 120 121 122 123 124 125 126
    int firstChr = FFMAX(1-chr_fsize, inst[1].filter_pos[chrSliceY]);

    int sp0 = firstLum - desc->src->plane[0].sliceY;
    int sp1 = firstChr - desc->src->plane[1].sliceY;
    int sp2 = firstChr - desc->src->plane[2].sliceY;
    int sp3 = firstLum - desc->src->plane[3].sliceY;
    int dp = sliceY - desc->dst->plane[0].sliceY;
    uint8_t **src0 = desc->src->plane[0].line + sp0;
    uint8_t **src1 = desc->src->plane[1].line + sp1;
    uint8_t **src2 = desc->src->plane[2].line + sp2;
    uint8_t **src3 = desc->alpha ? desc->src->plane[3].line + sp3 : NULL;
    uint8_t **dst = desc->dst->plane[0].line + dp;


127 128 129 130
    if (c->yuv2packed1 && lum_fsize == 1 && chr_fsize == 1) { // unscaled RGB
        ((yuv2packed1_fn)inst->pfn)(c, (const int16_t*)*src0, (const int16_t**)src1, (const int16_t**)src2,
                                    (const int16_t*)(desc->alpha ? *src3 : NULL),  *dst, dstW, 0, sliceY);
    } else if (c->yuv2packed1 && lum_fsize == 1 && chr_fsize == 2 &&
131 132 133
               chr_filter[2 * chrSliceY + 1] + chr_filter[2 * chrSliceY] == 4096 &&
               chr_filter[2 * chrSliceY + 1] <= 4096U) { // unscaled RGB
        int chrAlpha = chr_filter[2 * chrSliceY + 1];
134 135 136 137 138 139 140 141
        ((yuv2packed1_fn)inst->pfn)(c, (const int16_t*)*src0, (const int16_t**)src1, (const int16_t**)src2,
                                    (const int16_t*)(desc->alpha ? *src3 : NULL),  *dst, dstW, chrAlpha, sliceY);
    } else if (c->yuv2packed2 && lum_fsize == 2 && chr_fsize == 2 &&
               lum_filter[2 * sliceY + 1] + lum_filter[2 * sliceY] == 4096 &&
               lum_filter[2 * sliceY + 1] <= 4096U &&
               chr_filter[2 * chrSliceY + 1] + chr_filter[2 * chrSliceY] == 4096 &&
               chr_filter[2 * chrSliceY + 1] <= 4096U
    ) { // bilinear upscale RGB
142
        int lumAlpha = lum_filter[2 * sliceY + 1];
143
        int chrAlpha = chr_filter[2 * chrSliceY + 1];
144 145 146 147 148 149 150
        c->lumMmxFilter[2] =
        c->lumMmxFilter[3] = lum_filter[2 * sliceY]    * 0x10001;
        c->chrMmxFilter[2] =
        c->chrMmxFilter[3] = chr_filter[2 * chrSliceY] * 0x10001;
        ((yuv2packed2_fn)inst->pfn)(c, (const int16_t**)src0, (const int16_t**)src1, (const int16_t**)src2, (const int16_t**)src3,
                    *dst, dstW, lumAlpha, chrAlpha, sliceY);
    } else { // general RGB
151 152 153 154 155 156 157 158
        if ((c->yuv2packed1 && lum_fsize == 1 && chr_fsize == 2) ||
            (c->yuv2packed2 && lum_fsize == 2 && chr_fsize == 2)) {
            if (!c->warned_unuseable_bilinear)
                av_log(c, AV_LOG_INFO, "Optimized 2 tap filter code cannot be used\n");
            c->warned_unuseable_bilinear = 1;
        }

        inst->yuv2packedX(c, lum_filter + sliceY * lum_fsize,
159
                    (const int16_t**)src0, lum_fsize, chr_filter + chrSliceY * chr_fsize,
160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175
                    (const int16_t**)src1, (const int16_t**)src2, chr_fsize, (const int16_t**)src3, *dst, dstW, sliceY);
    }
    return 1;
}

static int any_vscale(SwsContext *c, SwsFilterDescriptor *desc, int sliceY, int sliceH)
{
    VScalerContext *inst = desc->instance;
    int dstW = desc->dst->width;
    int chrSliceY = sliceY >> desc->dst->v_chr_sub_sample;

    int lum_fsize = inst[0].filter_size;
    int chr_fsize = inst[1].filter_size;
    uint16_t *lum_filter = inst[0].filter[0];
    uint16_t *chr_filter = inst[1].filter[0];

176
    int firstLum = FFMAX(1-lum_fsize, inst[0].filter_pos[   sliceY]);
177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220
    int firstChr = FFMAX(1-chr_fsize, inst[1].filter_pos[chrSliceY]);

    int sp0 = firstLum - desc->src->plane[0].sliceY;
    int sp1 = firstChr - desc->src->plane[1].sliceY;
    int sp2 = firstChr - desc->src->plane[2].sliceY;
    int sp3 = firstLum - desc->src->plane[3].sliceY;
    int dp0 = sliceY - desc->dst->plane[0].sliceY;
    int dp1 = chrSliceY - desc->dst->plane[1].sliceY;
    int dp2 = chrSliceY - desc->dst->plane[2].sliceY;
    int dp3 = sliceY - desc->dst->plane[3].sliceY;

    uint8_t **src0 = desc->src->plane[0].line + sp0;
    uint8_t **src1 = desc->src->plane[1].line + sp1;
    uint8_t **src2 = desc->src->plane[2].line + sp2;
    uint8_t **src3 = desc->alpha ? desc->src->plane[3].line + sp3 : NULL;
    uint8_t *dst[4] = { desc->dst->plane[0].line[dp0],
                        desc->dst->plane[1].line[dp1],
                        desc->dst->plane[2].line[dp2],
                        desc->alpha ? desc->dst->plane[3].line[dp3] : NULL };

    av_assert1(!c->yuv2packed1 && !c->yuv2packed2);
    ((yuv2anyX_fn)inst->pfn)(c, lum_filter + sliceY * lum_fsize,
             (const int16_t**)src0, lum_fsize, chr_filter + sliceY * chr_fsize,
             (const int16_t**)src1, (const int16_t**)src2, chr_fsize, (const int16_t**)src3, dst, dstW, sliceY);

    return 1;

}

int ff_init_vscale(SwsContext *c, SwsFilterDescriptor *desc, SwsSlice *src, SwsSlice *dst)
{
    VScalerContext *lumCtx = NULL;
    VScalerContext *chrCtx = NULL;

    if (isPlanarYUV(c->dstFormat) || (isGray(c->dstFormat) && !isALPHA(c->dstFormat))) {
        lumCtx = av_mallocz(sizeof(VScalerContext));
        if (!lumCtx)
            return AVERROR(ENOMEM);


        desc[0].process = lum_planar_vscale;
        desc[0].instance = lumCtx;
        desc[0].src = src;
        desc[0].dst = dst;
221
        desc[0].alpha = c->needAlpha;
222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241

        if (!isGray(c->dstFormat)) {
            chrCtx = av_mallocz(sizeof(VScalerContext));
            if (!chrCtx)
                return AVERROR(ENOMEM);
            desc[1].process = chr_planar_vscale;
            desc[1].instance = chrCtx;
            desc[1].src = src;
            desc[1].dst = dst;
        }
    } else {
        lumCtx = av_mallocz_array(sizeof(VScalerContext), 2);
        if (!lumCtx)
            return AVERROR(ENOMEM);
        chrCtx = &lumCtx[1];

        desc[0].process = c->yuv2packedX ? packed_vscale : any_vscale;
        desc[0].instance = lumCtx;
        desc[0].src = src;
        desc[0].dst = dst;
242
        desc[0].alpha = c->needAlpha;
243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260
    }

    ff_init_vscale_pfn(c, c->yuv2plane1, c->yuv2planeX, c->yuv2nv12cX,
        c->yuv2packed1, c->yuv2packed2, c->yuv2packedX, c->yuv2anyX, c->use_mmx_vfilter);
    return 0;
}

void ff_init_vscale_pfn(SwsContext *c,
    yuv2planar1_fn yuv2plane1,
    yuv2planarX_fn yuv2planeX,
    yuv2interleavedX_fn yuv2nv12cX,
    yuv2packed1_fn yuv2packed1,
    yuv2packed2_fn yuv2packed2,
    yuv2packedX_fn yuv2packedX,
    yuv2anyX_fn yuv2anyX, int use_mmx)
{
    VScalerContext *lumCtx = NULL;
    VScalerContext *chrCtx = NULL;
261
    int idx = c->numDesc - (c->is_internal_gamma ? 2 : 1); //FIXME avoid hardcoding indexes
262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308

    if (isPlanarYUV(c->dstFormat) || (isGray(c->dstFormat) && !isALPHA(c->dstFormat))) {
        if (!isGray(c->dstFormat)) {
            chrCtx = c->desc[idx].instance;

            chrCtx->filter[0] = use_mmx ? (int16_t*)c->chrMmxFilter : c->vChrFilter;
            chrCtx->filter_size = c->vChrFilterSize;
            chrCtx->filter_pos = c->vChrFilterPos;
            chrCtx->isMMX = use_mmx;

            --idx;
            if (yuv2nv12cX)               chrCtx->pfn = yuv2nv12cX;
            else if (c->vChrFilterSize == 1) chrCtx->pfn = yuv2plane1;
            else                             chrCtx->pfn = yuv2planeX;
        }

        lumCtx = c->desc[idx].instance;

        lumCtx->filter[0] = use_mmx ? (int16_t*)c->lumMmxFilter : c->vLumFilter;
        lumCtx->filter[1] = use_mmx ? (int16_t*)c->alpMmxFilter : c->vLumFilter;
        lumCtx->filter_size = c->vLumFilterSize;
        lumCtx->filter_pos = c->vLumFilterPos;
        lumCtx->isMMX = use_mmx;

        if (c->vLumFilterSize == 1) lumCtx->pfn = yuv2plane1;
        else                        lumCtx->pfn = yuv2planeX;

    } else {
        lumCtx = c->desc[idx].instance;
        chrCtx = &lumCtx[1];

        lumCtx->filter[0] = c->vLumFilter;
        lumCtx->filter_size = c->vLumFilterSize;
        lumCtx->filter_pos = c->vLumFilterPos;

        chrCtx->filter[0] = c->vChrFilter;
        chrCtx->filter_size = c->vChrFilterSize;
        chrCtx->filter_pos = c->vChrFilterPos;

        lumCtx->isMMX = use_mmx;
        chrCtx->isMMX = use_mmx;

        if (yuv2packedX) {
            if (c->yuv2packed1 && c->vLumFilterSize == 1 && c->vChrFilterSize <= 2)
                lumCtx->pfn = yuv2packed1;
            else if (c->yuv2packed2 && c->vLumFilterSize == 2 && c->vChrFilterSize == 2)
                lumCtx->pfn = yuv2packed2;
309
            lumCtx->yuv2packedX = yuv2packedX;
310 311 312 313 314 315
        } else
            lumCtx->pfn = yuv2anyX;
    }
}