44 0x0200020002000200LL,};
48 0x0004000400040004LL,};
61 DECLARE_ALIGNED(8,
const uint64_t, ff_bgr2YOffset) = 0x1010101010101010ULL;
62 DECLARE_ALIGNED(8,
const uint64_t, ff_bgr2UVOffset) = 0x8080808080808080ULL;
69 #define COMPILE_TEMPLATE_MMXEXT 0
70 #define RENAME(a) a ## _mmx
75 #if HAVE_MMXEXT_INLINE
77 #undef COMPILE_TEMPLATE_MMXEXT
78 #define COMPILE_TEMPLATE_MMXEXT 1
79 #define RENAME(a) a ## _mmxext
103 const int firstLumSrcY= vLumFilterPos[
dstY];
104 const int firstChrSrcY= vChrFilterPos[chrDstY];
112 if (dstY < dstH - 2) {
113 const int16_t **lumSrcPtr= (
const int16_t **)(
void*) lumPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf +
vLumBufSize;
114 const int16_t **chrUSrcPtr= (
const int16_t **)(
void*) chrUPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf +
vChrBufSize;
115 const int16_t **alpSrcPtr= (CONFIG_SWSCALE_ALPHA &&
alpPixBuf) ? (
const int16_t **)(
void*) alpPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize :
NULL;
118 if (firstLumSrcY < 0 || firstLumSrcY + vLumFilterSize > c->
srcH) {
119 const int16_t **tmpY = (
const int16_t **) lumPixBuf + 2 * vLumBufSize;
120 int neg = -firstLumSrcY, i,
end =
FFMIN(c->
srcH - firstLumSrcY, vLumFilterSize);
121 for (i = 0; i < neg; i++)
122 tmpY[i] = lumSrcPtr[neg];
123 for ( ; i <
end; i++)
124 tmpY[i] = lumSrcPtr[i];
130 const int16_t **tmpA = (
const int16_t **) alpPixBuf + 2 * vLumBufSize;
131 for (i = 0; i < neg; i++)
132 tmpA[i] = alpSrcPtr[neg];
133 for ( ; i <
end; i++)
134 tmpA[i] = alpSrcPtr[i];
136 tmpA[i] = tmpA[i - 1];
140 if (firstChrSrcY < 0 || firstChrSrcY + vChrFilterSize > c->
chrSrcH) {
141 const int16_t **tmpU = (
const int16_t **) chrUPixBuf + 2 * vChrBufSize;
142 int neg = -firstChrSrcY, i, end =
FFMIN(c->
chrSrcH - firstChrSrcY, vChrFilterSize);
143 for (i = 0; i < neg; i++) {
144 tmpU[i] = chrUSrcPtr[neg];
146 for ( ; i <
end; i++) {
147 tmpU[i] = chrUSrcPtr[i];
150 tmpU[i] = tmpU[i - 1];
158 *(
const void**)&lumMmxFilter[s*i ]= lumSrcPtr[i ];
159 *(
const void**)&lumMmxFilter[s*i+
APCK_PTR2/4 ]= lumSrcPtr[i+(vLumFilterSize>1)];
161 lumMmxFilter[s*i+
APCK_COEF/4+1]= vLumFilter[dstY*vLumFilterSize + i ]
162 + (vLumFilterSize>1 ? vLumFilter[dstY*vLumFilterSize + i + 1]<<16 : 0);
163 if (CONFIG_SWSCALE_ALPHA && alpPixBuf) {
164 *(
const void**)&alpMmxFilter[s*i ]= alpSrcPtr[i ];
165 *(
const void**)&alpMmxFilter[s*i+
APCK_PTR2/4 ]= alpSrcPtr[i+(vLumFilterSize>1)];
171 *(
const void**)&chrMmxFilter[s*i ]= chrUSrcPtr[i ];
172 *(
const void**)&chrMmxFilter[s*i+
APCK_PTR2/4 ]= chrUSrcPtr[i+(vChrFilterSize>1)];
174 chrMmxFilter[s*i+
APCK_COEF/4+1]= vChrFilter[chrDstY*vChrFilterSize + i ]
175 + (vChrFilterSize>1 ? vChrFilter[chrDstY*vChrFilterSize + i + 1]<<16 : 0);
179 *(
const void**)&lumMmxFilter[4*i+0]= lumSrcPtr[i];
182 ((uint16_t)vLumFilter[dstY*vLumFilterSize + i])*0x10001
U;
183 if (CONFIG_SWSCALE_ALPHA && alpPixBuf) {
184 *(
const void**)&alpMmxFilter[4*i+0]= alpSrcPtr[i];
186 alpMmxFilter[4*i+3]= lumMmxFilter[4*i+2];
190 *(
const void**)&chrMmxFilter[4*i+0]= chrUSrcPtr[i];
193 ((uint16_t)vChrFilter[chrDstY*vChrFilterSize + i])*0x10001
U;
200 static void yuv2yuvX_sse3(
const int16_t *
filter,
int filterSize,
204 if(((uintptr_t)dest) & 15){
205 yuv2yuvX_mmxext(filter, filterSize, src, dest, dstW, dither, offset);
209 #define MAIN_FUNCTION \
210 "pxor %%xmm0, %%xmm0 \n\t" \
211 "punpcklbw %%xmm0, %%xmm3 \n\t" \
212 "movd %4, %%xmm1 \n\t" \
213 "punpcklwd %%xmm1, %%xmm1 \n\t" \
214 "punpckldq %%xmm1, %%xmm1 \n\t" \
215 "punpcklqdq %%xmm1, %%xmm1 \n\t" \
216 "psllw $3, %%xmm1 \n\t" \
217 "paddw %%xmm1, %%xmm3 \n\t" \
218 "psraw $4, %%xmm3 \n\t" \
219 "movdqa %%xmm3, %%xmm4 \n\t" \
220 "movdqa %%xmm3, %%xmm7 \n\t" \
221 "movl %3, %%ecx \n\t" \
222 "mov %0, %%"REG_d" \n\t"\
223 "mov (%%"REG_d"), %%"REG_S" \n\t"\
226 "movddup 8(%%"REG_d"), %%xmm0 \n\t" \
227 "movdqa (%%"REG_S", %%"REG_c", 2), %%xmm2 \n\t" \
228 "movdqa 16(%%"REG_S", %%"REG_c", 2), %%xmm5 \n\t" \
229 "add $16, %%"REG_d" \n\t"\
230 "mov (%%"REG_d"), %%"REG_S" \n\t"\
231 "test %%"REG_S", %%"REG_S" \n\t"\
232 "pmulhw %%xmm0, %%xmm2 \n\t"\
233 "pmulhw %%xmm0, %%xmm5 \n\t"\
234 "paddw %%xmm2, %%xmm3 \n\t"\
235 "paddw %%xmm5, %%xmm4 \n\t"\
237 "psraw $3, %%xmm3 \n\t"\
238 "psraw $3, %%xmm4 \n\t"\
239 "packuswb %%xmm4, %%xmm3 \n\t"\
240 "movntdq %%xmm3, (%1, %%"REG_c")\n\t"\
241 "add $16, %%"REG_c" \n\t"\
242 "cmp %2, %%"REG_c" \n\t"\
243 "movdqa %%xmm7, %%xmm3 \n\t" \
244 "movdqa %%xmm7, %%xmm4 \n\t" \
245 "mov %0, %%"REG_d" \n\t"\
246 "mov (%%"REG_d"), %%"REG_S" \n\t"\
251 "movq %5, %%xmm3 \n\t"
252 "movdqa %%xmm3, %%xmm4 \n\t"
253 "psrlq $24, %%xmm3 \n\t"
254 "psllq $40, %%xmm4 \n\t"
255 "por %%xmm4, %%xmm3 \n\t"
258 "r" (dest-offset),
"g" ((
x86_reg)(dstW+offset)),
"m" (offset),
259 "m"(filterSize),
"m"(((uint64_t *)
dither)[0])
260 :
XMM_CLOBBERS(
"%xmm0" ,
"%xmm1" ,
"%xmm2" ,
"%xmm3" ,
"%xmm4" ,
"%xmm5" ,
"%xmm7" ,)
261 "%"REG_d,
"%"REG_S,
"%"REG_c
265 "movq %5, %%xmm3 \n\t"
268 "r" (dest-offset),
"g" ((
x86_reg)(dstW+offset)),
"m" (offset),
269 "m"(filterSize),
"m"(((uint64_t *)
dither)[0])
270 :
XMM_CLOBBERS(
"%xmm0" ,
"%xmm1" ,
"%xmm2" ,
"%xmm3" ,
"%xmm4" ,
"%xmm5" ,
"%xmm7" ,)
271 "%"REG_d,
"%"REG_S,
"%"REG_c
279 #define SCALE_FUNC(filter_n, from_bpc, to_bpc, opt) \
280 void ff_hscale ## from_bpc ## to ## to_bpc ## _ ## filter_n ## _ ## opt( \
281 SwsContext *c, int16_t *data, \
282 int dstW, const uint8_t *src, \
283 const int16_t *filter, \
284 const int32_t *filterPos, int filterSize)
286 #define SCALE_FUNCS(filter_n, opt) \
287 SCALE_FUNC(filter_n, 8, 15, opt); \
288 SCALE_FUNC(filter_n, 9, 15, opt); \
289 SCALE_FUNC(filter_n, 10, 15, opt); \
290 SCALE_FUNC(filter_n, 12, 15, opt); \
291 SCALE_FUNC(filter_n, 14, 15, opt); \
292 SCALE_FUNC(filter_n, 16, 15, opt); \
293 SCALE_FUNC(filter_n, 8, 19, opt); \
294 SCALE_FUNC(filter_n, 9, 19, opt); \
295 SCALE_FUNC(filter_n, 10, 19, opt); \
296 SCALE_FUNC(filter_n, 12, 19, opt); \
297 SCALE_FUNC(filter_n, 14, 19, opt); \
298 SCALE_FUNC(filter_n, 16, 19, opt)
300 #define SCALE_FUNCS_MMX(opt) \
301 SCALE_FUNCS(4, opt); \
302 SCALE_FUNCS(8, opt); \
305 #define SCALE_FUNCS_SSE(opt) \
306 SCALE_FUNCS(4, opt); \
307 SCALE_FUNCS(8, opt); \
308 SCALE_FUNCS(X4, opt); \
318 #define VSCALEX_FUNC(size, opt) \
319 void ff_yuv2planeX_ ## size ## _ ## opt(const int16_t *filter, int filterSize, \
320 const int16_t **src, uint8_t *dest, int dstW, \
321 const uint8_t *dither, int offset)
322 #define VSCALEX_FUNCS(opt) \
323 VSCALEX_FUNC(8, opt); \
324 VSCALEX_FUNC(9, opt); \
325 VSCALEX_FUNC(10, opt)
335 #define VSCALE_FUNC(size, opt) \
336 void ff_yuv2plane1_ ## size ## _ ## opt(const int16_t *src, uint8_t *dst, int dstW, \
337 const uint8_t *dither, int offset)
338 #define VSCALE_FUNCS(opt1, opt2) \
339 VSCALE_FUNC(8, opt1); \
340 VSCALE_FUNC(9, opt2); \
341 VSCALE_FUNC(10, opt2); \
342 VSCALE_FUNC(16, opt1)
351 #define INPUT_Y_FUNC(fmt, opt) \
352 void ff_ ## fmt ## ToY_ ## opt(uint8_t *dst, const uint8_t *src, \
353 const uint8_t *unused1, const uint8_t *unused2, \
354 int w, uint32_t *unused)
355 #define INPUT_UV_FUNC(fmt, opt) \
356 void ff_ ## fmt ## ToUV_ ## opt(uint8_t *dstU, uint8_t *dstV, \
357 const uint8_t *unused0, \
358 const uint8_t *src1, \
359 const uint8_t *src2, \
360 int w, uint32_t *unused)
361 #define INPUT_FUNC(fmt, opt) \
362 INPUT_Y_FUNC(fmt, opt); \
363 INPUT_UV_FUNC(fmt, opt)
364 #define INPUT_FUNCS(opt) \
365 INPUT_FUNC(uyvy, opt); \
366 INPUT_FUNC(yuyv, opt); \
367 INPUT_UV_FUNC(nv12, opt); \
368 INPUT_UV_FUNC(nv21, opt); \
369 INPUT_FUNC(rgba, opt); \
370 INPUT_FUNC(bgra, opt); \
371 INPUT_FUNC(argb, opt); \
372 INPUT_FUNC(abgr, opt); \
373 INPUT_FUNC(rgb24, opt); \
374 INPUT_FUNC(bgr24, opt)
389 sws_init_swscale_mmx(c);
391 #if HAVE_MMXEXT_INLINE
393 sws_init_swscale_mmxext(c);
400 #define ASSIGN_SCALE_FUNC2(hscalefn, filtersize, opt1, opt2) do { \
401 if (c->srcBpc == 8) { \
402 hscalefn = c->dstBpc <= 14 ? ff_hscale8to15_ ## filtersize ## _ ## opt2 : \
403 ff_hscale8to19_ ## filtersize ## _ ## opt1; \
404 } else if (c->srcBpc == 9) { \
405 hscalefn = c->dstBpc <= 14 ? ff_hscale9to15_ ## filtersize ## _ ## opt2 : \
406 ff_hscale9to19_ ## filtersize ## _ ## opt1; \
407 } else if (c->srcBpc == 10) { \
408 hscalefn = c->dstBpc <= 14 ? ff_hscale10to15_ ## filtersize ## _ ## opt2 : \
409 ff_hscale10to19_ ## filtersize ## _ ## opt1; \
410 } else if (c->srcBpc == 12) { \
411 hscalefn = c->dstBpc <= 14 ? ff_hscale12to15_ ## filtersize ## _ ## opt2 : \
412 ff_hscale12to19_ ## filtersize ## _ ## opt1; \
413 } else if (c->srcBpc == 14 || ((c->srcFormat==AV_PIX_FMT_PAL8||isAnyRGB(c->srcFormat)) && av_pix_fmt_desc_get(c->srcFormat)->comp[0].depth_minus1<15)) { \
414 hscalefn = c->dstBpc <= 14 ? ff_hscale14to15_ ## filtersize ## _ ## opt2 : \
415 ff_hscale14to19_ ## filtersize ## _ ## opt1; \
417 av_assert0(c->srcBpc == 16);\
418 hscalefn = c->dstBpc <= 14 ? ff_hscale16to15_ ## filtersize ## _ ## opt2 : \
419 ff_hscale16to19_ ## filtersize ## _ ## opt1; \
422 #define ASSIGN_MMX_SCALE_FUNC(hscalefn, filtersize, opt1, opt2) \
423 switch (filtersize) { \
424 case 4: ASSIGN_SCALE_FUNC2(hscalefn, 4, opt1, opt2); break; \
425 case 8: ASSIGN_SCALE_FUNC2(hscalefn, 8, opt1, opt2); break; \
426 default: ASSIGN_SCALE_FUNC2(hscalefn, X, opt1, opt2); break; \
428 #define ASSIGN_VSCALEX_FUNC(vscalefn, opt, do_16_case, condition_8bit) \
430 case 16: do_16_case; break; \
431 case 10: if (!isBE(c->dstFormat)) vscalefn = ff_yuv2planeX_10_ ## opt; break; \
432 case 9: if (!isBE(c->dstFormat)) vscalefn = ff_yuv2planeX_9_ ## opt; break; \
433 case 8: if ((condition_8bit) && !c->use_mmx_vfilter) vscalefn = ff_yuv2planeX_8_ ## opt; break; \
435 #define ASSIGN_VSCALE_FUNC(vscalefn, opt1, opt2, opt2chk) \
437 case 16: if (!isBE(c->dstFormat)) vscalefn = ff_yuv2plane1_16_ ## opt1; break; \
438 case 10: if (!isBE(c->dstFormat) && opt2chk) vscalefn = ff_yuv2plane1_10_ ## opt2; break; \
439 case 9: if (!isBE(c->dstFormat) && opt2chk) vscalefn = ff_yuv2plane1_9_ ## opt2; break; \
440 case 8: vscalefn = ff_yuv2plane1_8_ ## opt1; break; \
441 default: av_assert0(c->dstBpc>8); \
443 #define case_rgb(x, X, opt) \
444 case AV_PIX_FMT_ ## X: \
445 c->lumToYV12 = ff_ ## x ## ToY_ ## opt; \
446 if (!c->chrSrcHSubSample) \
447 c->chrToYV12 = ff_ ## x ## ToUV_ ## opt; \
489 #define ASSIGN_SSE_SCALE_FUNC(hscalefn, filtersize, opt1, opt2) \
490 switch (filtersize) { \
491 case 4: ASSIGN_SCALE_FUNC2(hscalefn, 4, opt1, opt2); break; \
492 case 8: ASSIGN_SCALE_FUNC2(hscalefn, 8, opt1, opt2); break; \
493 default: if (filtersize & 4) ASSIGN_SCALE_FUNC2(hscalefn, X4, opt1, opt2); \
494 else ASSIGN_SCALE_FUNC2(hscalefn, X8, opt1, opt2); \
501 HAVE_ALIGNED_STACK || ARCH_X86_64);
550 HAVE_ALIGNED_STACK || ARCH_X86_64);
557 HAVE_ALIGNED_STACK || ARCH_X86_64);
#define EXTERNAL_MMX(flags)
packed YUV 4:2:2, 16bpp, Cb Y0 Cr Y1
int16_t ** alpPixBuf
Ring buffer for scaled horizontal alpha plane lines to be fed to the vertical scaler.
const uint64_t ff_dither8[2]
int chrBufIndex
Index in ring buffer of the last scaled horizontal chroma line from source.
int chrSrcH
Height of source chroma planes.
#define VSCALE_FUNC(size, opt)
#define SCALE_FUNCS_MMX(opt)
#define DECLARE_ALIGNED(n, t, v)
int dstY
Last destination vertical line output from last slice.
#define case_rgb(x, X, opt)
Macro definitions for various function/variable attributes.
#define ASSIGN_SSE_SCALE_FUNC(hscalefn, filtersize, opt1, opt2)
int srcH
Height of source luma/alpha planes.
#define VSCALE_FUNCS(opt1, opt2)
#define EXTERNAL_SSE4(flags)
#define DECLARE_ASM_CONST(n, t, v)
int chrDstVSubSample
Binary logarithm of vertical subsampling factor between luma/alpha and chroma planes in destination i...
int vChrFilterSize
Vertical filter size for chroma pixels.
static av_cold int end(AVCodecContext *avctx)
int16_t ** lumPixBuf
Ring buffer for scaled horizontal luma plane lines to be fed to the vertical scaler.
#define AV_CPU_FLAG_MMXEXT
SSE integer functions or AMD MMX ext.
void(* lumToYV12)(uint8_t *dst, const uint8_t *src, const uint8_t *src2, const uint8_t *src3, int width, uint32_t *pal)
Unscaled conversion of luma plane to YV12 for horizontal scaler.
av_cold void ff_sws_init_swscale_x86(SwsContext *c)
int lastInLumBuf
Last scaled horizontal luma/alpha line from source in the ring buffer.
enum AVPixelFormat dstFormat
Destination pixel format.
#define EXTERNAL_SSE2(flags)
#define VSCALEX_FUNCS(opt)
int32_t * vChrFilterPos
Array of vertical filter starting positions for each dst[i] for chroma planes.
int dstH
Height of destination luma/alpha planes.
#define INLINE_MMX(flags)
const uint64_t ff_dither4[2]
int hLumFilterSize
Horizontal filter size for luma/alpha pixels.
static const uint8_t dither[8][8]
planar YUV 4:2:0, 12bpp, 1 plane for Y and 1 plane for the UV components, which are interleaved (firs...
simple assert() macros that are a bit more flexible than ISO C assert().
int vChrBufSize
Number of vertical chroma lines allocated in the ring buffer.
static const uint8_t offset[127][2]
int32_t alpMmxFilter[4 *MAX_FILTER_SIZE]
int hChrFilterSize
Horizontal filter size for chroma pixels.
as above, but U and V bytes are swapped
#define AV_CPU_FLAG_SSE3
Prescott SSE3 functions.
yuv2planar1_fn yuv2plane1
int vLumBufSize
Number of vertical luma/alpha lines allocated in the ring buffer.
#define SCALE_FUNCS_SSE(opt)
int16_t ** chrUPixBuf
Ring buffer for scaled horizontal chroma plane lines to be fed to the vertical scaler.
void(* hcScale)(struct SwsContext *c, int16_t *dst, int dstW, const uint8_t *src, const int16_t *filter, const int32_t *filterPos, int filterSize)
#define XMM_CLOBBERS(...)
int dstW
Width of destination luma/alpha planes.
int32_t * vLumFilterPos
Array of vertical filter starting positions for each dst[i] for luma/alpha planes.
#define AV_PIX_FMT_BGR555
static av_always_inline int isBE(enum AVPixelFormat pix_fmt)
int32_t lumMmxFilter[4 *MAX_FILTER_SIZE]
#define ASSIGN_MMX_SCALE_FUNC(hscalefn, filtersize, opt1, opt2)
yuv2planarX_fn yuv2planeX
packed YUV 4:2:2, 16bpp, Y0 Cb Y1 Cr
#define EXTERNAL_SSSE3(flags)
int vLumFilterSize
Vertical filter size for luma/alpha pixels.
int16_t * vChrFilter
Array of vertical filter coefficients for chroma planes.
int av_get_cpu_flags(void)
Return the flags which specify extensions supported by the CPU.
#define EXTERNAL_MMXEXT(flags)
int lumBufIndex
Index in ring buffer of the last scaled horizontal luma/alpha line from source.
#define VSCALEX_FUNC(size, opt)
#define INLINE_MMXEXT(flags)
int lastInChrBuf
Last scaled horizontal chroma line from source in the ring buffer.
static void filter(MpegAudioContext *s, int ch, const short *samples, int incr)
enum AVPixelFormat srcFormat
Source pixel format.
int32_t chrMmxFilter[4 *MAX_FILTER_SIZE]
#define AV_PIX_FMT_RGB555
#define ASSIGN_VSCALEX_FUNC(vscalefn, opt, do_16_case, condition_8bit)
#define ASSIGN_VSCALE_FUNC(vscalefn, opt1, opt2, opt2chk)
void(* alpToYV12)(uint8_t *dst, const uint8_t *src, const uint8_t *src2, const uint8_t *src3, int width, uint32_t *pal)
Unscaled conversion of alpha plane to YV12 for horizontal scaler.
void ff_updateMMXDitherTables(SwsContext *c, int dstY, int lumBufIndex, int chrBufIndex, int lastInLumBuf, int lastInChrBuf)
void(* chrToYV12)(uint8_t *dstU, uint8_t *dstV, const uint8_t *src1, const uint8_t *src2, const uint8_t *src3, int width, uint32_t *pal)
Unscaled conversion of chroma planes to YV12 for horizontal scaler.
int16_t * vLumFilter
Array of vertical filter coefficients for luma/alpha planes.
int flags
Flags passed by the user to select scaler algorithm, optimizations, subsampling, etc...
#define EXTERNAL_AVX(flags)
void(* hyScale)(struct SwsContext *c, int16_t *dst, int dstW, const uint8_t *src, const int16_t *filter, const int32_t *filterPos, int filterSize)
Scale one horizontal line of input data using a filter over the input lines, to produce one (differen...