* (c) 2001 Fabrice Bellard
* 2007 Marc Hoffman <marc.hoffman@analog.com>
*
- * This file is part of Libav.
+ * This file is part of FFmpeg.
*
- * Libav is free software; you can redistribute it and/or
+ * 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.
*
- * Libav is distributed in the hope that it will be useful,
+ * 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 Libav; if not, write to the Free Software
+ * License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#undef printf
+void ff_mmx_idct(DCTELEM *data);
+void ff_mmxext_idct(DCTELEM *data);
+
// BFIN
void ff_bfin_idct(DCTELEM *block);
void ff_bfin_fdct(DCTELEM *block);
// ALTIVEC
void ff_fdct_altivec(DCTELEM *block);
- //void ff_idct_altivec(DCTELEM *block);?? no routine
// ARM
void ff_j_rev_dct_arm(DCTELEM *data);
const char *name;
void (*func)(DCTELEM *block);
enum formattag { NO_PERM, MMX_PERM, MMX_SIMPLE_PERM, SCALE_PERM,
- SSE2_PERM, PARTTRANS_PERM } format;
+ SSE2_PERM, PARTTRANS_PERM, TRANSPOSE_PERM } format;
int mm_support;
int nonspec;
};
{ 0 }
};
+#if ARCH_X86_64 && HAVE_MMX && HAVE_YASM
+void ff_prores_idct_put_10_sse2(uint16_t *dst, int linesize,
+ DCTELEM *block, int16_t *qmat);
+
+static void ff_prores_idct_put_10_sse2_wrap(DCTELEM *dst){
+ DECLARE_ALIGNED(16, static int16_t, qmat)[64];
+ DECLARE_ALIGNED(16, static int16_t, tmp)[64];
+ int i;
+
+ for(i=0; i<64; i++){
+ qmat[i]=4;
+ tmp[i]= dst[i];
+ }
+ ff_prores_idct_put_10_sse2(dst, 16, tmp, qmat);
+}
+#endif
+
static const struct algo idct_tab[] = {
{ "FAANI", ff_faanidct, NO_PERM },
{ "REF-DBL", ff_ref_idct, NO_PERM },
{ "SIMPLE-C", ff_simple_idct_8, NO_PERM },
#if HAVE_MMX_INLINE
+#if CONFIG_GPL
+ { "LIBMPEG2-MMX", ff_mmx_idct, MMX_PERM, AV_CPU_FLAG_MMX, 1 },
+ { "LIBMPEG2-MMX2", ff_mmxext_idct, MMX_PERM, AV_CPU_FLAG_MMX2, 1 },
+#endif
{ "SIMPLE-MMX", ff_simple_idct_mmx, MMX_SIMPLE_PERM, AV_CPU_FLAG_MMX },
{ "XVID-MMX", ff_idct_xvid_mmx, NO_PERM, AV_CPU_FLAG_MMX, 1 },
{ "XVID-MMXEXT", ff_idct_xvid_mmx2, NO_PERM, AV_CPU_FLAG_MMXEXT, 1 },
{ "XVID-SSE2", ff_idct_xvid_sse2, SSE2_PERM, AV_CPU_FLAG_SSE2, 1 },
+#if ARCH_X86_64 && HAVE_YASM
+ { "PR-SSE2", ff_prores_idct_put_10_sse2_wrap, TRANSPOSE_PERM, AV_CPU_FLAG_SSE2, 1 },
+#endif
#endif
#if ARCH_BFIN
DECLARE_ALIGNED(16, static DCTELEM, block)[64];
DECLARE_ALIGNED(8, static DCTELEM, block1)[64];
-static void init_block(DCTELEM block[64], int test, int is_idct, AVLFG *prng)
+static void init_block(DCTELEM block[64], int test, int is_idct, AVLFG *prng, int vals)
{
int i, j;
switch (test) {
case 0:
for (i = 0; i < 64; i++)
- block[i] = (av_lfg_get(prng) % 512) - 256;
+ block[i] = (av_lfg_get(prng) % (2*vals)) -vals;
if (is_idct) {
ff_ref_fdct(block);
for (i = 0; i < 64; i++)
break;
case 1:
j = av_lfg_get(prng) % 10 + 1;
- for (i = 0; i < j; i++)
- block[av_lfg_get(prng) % 64] = av_lfg_get(prng) % 512 - 256;
+ for (i = 0; i < j; i++) {
+ int idx = av_lfg_get(prng) % 64;
+ block[idx] = av_lfg_get(prng) % (2*vals) -vals;
+ }
break;
case 2:
- block[ 0] = av_lfg_get(prng) % 4096 - 2048;
+ block[ 0] = av_lfg_get(prng) % (16*vals) - (8*vals);
block[63] = (block[0] & 1) ^ 1;
break;
}
} else if (perm == PARTTRANS_PERM) {
for (i = 0; i < 64; i++)
dst[(i & 0x24) | ((i & 3) << 3) | ((i >> 3) & 3)] = src[i];
+ } else if (perm == TRANSPOSE_PERM) {
+ for (i = 0; i < 64; i++)
+ dst[(i>>3) | ((i<<3)&0x38)] = src[i];
} else {
for (i = 0; i < 64; i++)
dst[i] = src[i];
}
}
-static int dct_error(const struct algo *dct, int test, int is_idct, int speed)
+static int dct_error(const struct algo *dct, int test, int is_idct, int speed, const int bits)
{
void (*ref)(DCTELEM *block) = is_idct ? ff_ref_idct : ff_ref_fdct;
int it, i, scale;
int maxout = 0;
int blockSumErrMax = 0, blockSumErr;
AVLFG prng;
+ const int vals=1<<bits;
double omse, ome;
int spec_err;
for (i = 0; i < 64; i++)
sysErr[i] = 0;
for (it = 0; it < NB_ITS; it++) {
- init_block(block1, test, is_idct, &prng);
+ init_block(block1, test, is_idct, &prng, vals);
permute(block, block1, dct->format);
dct->func(block);
spec_err = is_idct && (err_inf > 1 || omse > 0.02 || fabs(ome) > 0.0015);
- printf("%s %s: ppe=%d omse=%0.8f ome=%0.8f syserr=%0.8f maxout=%d blockSumErr=%d\n",
+ printf("%s %s: max_err=%d omse=%0.8f ome=%0.8f syserr=%0.8f maxout=%d blockSumErr=%d\n",
is_idct ? "IDCT" : "DCT", dct->name, err_inf,
omse, ome, (double) sysErrMax / NB_ITS,
maxout, blockSumErrMax);
return 0;
/* speed test */
- init_block(block, test, is_idct, &prng);
+
+ init_block(block, test, is_idct, &prng, vals);
permute(block1, block, dct->format);
ti = av_gettime();
memcpy(block, block1, sizeof(block));
dct->func(block);
}
+ emms_c();
it1 += NB_ITS_SPEED;
ti1 = av_gettime() - ti;
} while (ti1 < 1000000);
- emms_c();
printf("%s %s: %0.1f kdct/s\n", is_idct ? "IDCT" : "DCT", dct->name,
(double) it1 * 1000.0 / (double) ti1);
if (v > err_max)
err_max = v;
}
+#if 0
+ printf("ref=\n");
+ for(i=0;i<8;i++) {
+ int j;
+ for(j=0;j<8;j++) {
+ printf(" %3d", img_dest1[i*8+j]);
+ }
+ printf("\n");
+ }
+
+ printf("out=\n");
+ for(i=0;i<8;i++) {
+ int j;
+ for(j=0;j<8;j++) {
+ printf(" %3d", img_dest[i*8+j]);
+ }
+ printf("\n");
+ }
+#endif
}
printf("%s %s: err_inf=%d\n", 1 ? "IDCT248" : "DCT248", name, err_max);
block[i] = block1[i];
idct248_put(img_dest, 8, block);
}
+ emms_c();
it1 += NB_ITS_SPEED;
ti1 = av_gettime() - ti;
} while (ti1 < 1000000);
- emms_c();
printf("%s %s: %0.1f kdct/s\n", 1 ? "IDCT248" : "DCT248", name,
(double) it1 * 1000.0 / (double) ti1);
static void help(void)
{
- printf("dct-test [-i] [<test-number>]\n"
+ printf("dct-test [-i] [<test-number>] [<bits>]\n"
"test-number 0 -> test with random matrixes\n"
" 1 -> test with random sparse matrixes\n"
" 2 -> do 3. test from mpeg4 std\n"
+ "bits Number of time domain bits to use, 8 is default\n"
"-i test IDCT implementations\n"
"-4 test IDCT248 implementations\n"
"-t speed test\n");
int test = 1;
int speed = 0;
int err = 0;
+ int bits=8;
cpu_flags = av_get_cpu_flags();
if (optind < argc)
test = atoi(argv[optind]);
+ if(optind+1 < argc) bits= atoi(argv[optind+1]);
- printf("Libav DCT/IDCT test\n");
+ printf("ffmpeg DCT/IDCT test\n");
if (test_248_dct) {
idct248_error("SIMPLE-C", ff_simple_idct248_put, speed);
const struct algo *algos = test_idct ? idct_tab : fdct_tab;
for (i = 0; algos[i].name; i++)
if (!(~cpu_flags & algos[i].mm_support)) {
- err |= dct_error(&algos[i], test, test_idct, speed);
+ err |= dct_error(&algos[i], test, test_idct, speed, bits);
}
}
* Copyright (c) 2000, 2001 Fabrice Bellard
* Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
*
- * MMX optimization by Nick Kurshev <nickols_k@mail.ru>
- *
- * This file is part of Libav.
+ * This file is part of FFmpeg.
*
- * Libav is free software; you can redistribute it and/or
+ * 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.
*
- * Libav is distributed in the hope that it will be useful,
+ * 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 Libav; if not, write to the Free Software
+ * License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ *
+ * MMX optimization by Nick Kurshev <nickols_k@mail.ru>
*/
#include "libavutil/cpu.h"
#include "libavcodec/simple_idct.h"
#include "dsputil_mmx.h"
#include "idct_xvid.h"
+#include "diracdsp_mmx.h"
//#undef NDEBUG
//#include <assert.h>
: "+r"(ptr)
: "r"((x86_reg)wrap), "r"((x86_reg)width), "r"(ptr + wrap * height)
);
- } else {
+ } else if(w==16){
__asm__ volatile (
"1: \n\t"
"movd (%0), %%mm0 \n\t"
: "+r"(ptr)
: "r"((x86_reg)wrap), "r"((x86_reg)width), "r"(ptr + wrap * height)
);
+ } else {
+ av_assert1(w == 4);
+ __asm__ volatile (
+ "1: \n\t"
+ "movd (%0), %%mm0 \n\t"
+ "punpcklbw %%mm0, %%mm0 \n\t"
+ "punpcklwd %%mm0, %%mm0 \n\t"
+ "movd %%mm0, -4(%0) \n\t"
+ "movd -4(%0, %2), %%mm1 \n\t"
+ "punpcklbw %%mm1, %%mm1 \n\t"
+ "punpckhwd %%mm1, %%mm1 \n\t"
+ "punpckhdq %%mm1, %%mm1 \n\t"
+ "movd %%mm1, (%0, %2) \n\t"
+ "add %1, %0 \n\t"
+ "cmp %3, %0 \n\t"
+ "jb 1b \n\t"
+ : "+r"(ptr)
+ : "r"((x86_reg)wrap), "r"((x86_reg)width), "r"(ptr + wrap * height)
+ );
}
/* top and bottom (and hopefully also the corners) */
avg_pixels8_mmx2(dst, src, stride, 8);
}
- static void ff_idct_xvid_mmx_put(uint8_t *dest, int line_size, DCTELEM *block)
- {
- ff_idct_xvid_mmx(block);
- ff_put_pixels_clamped_mmx(block, dest, line_size);
- }
-
- static void ff_idct_xvid_mmx_add(uint8_t *dest, int line_size, DCTELEM *block)
- {
- ff_idct_xvid_mmx(block);
- ff_add_pixels_clamped_mmx(block, dest, line_size);
- }
-
- static void ff_idct_xvid_mmx2_put(uint8_t *dest, int line_size, DCTELEM *block)
- {
- ff_idct_xvid_mmx2(block);
- ff_put_pixels_clamped_mmx(block, dest, line_size);
- }
-
- static void ff_idct_xvid_mmx2_add(uint8_t *dest, int line_size, DCTELEM *block)
- {
- ff_idct_xvid_mmx2(block);
- ff_add_pixels_clamped_mmx(block, dest, line_size);
- }
-
+/* only used in VP3/5/6 */
+static void put_vp_no_rnd_pixels8_l2_mmx(uint8_t *dst, const uint8_t *a, const uint8_t *b, int stride, int h)
+{
+// START_TIMER
+ MOVQ_BFE(mm6);
+ __asm__ volatile(
+ "1: \n\t"
+ "movq (%1), %%mm0 \n\t"
+ "movq (%2), %%mm1 \n\t"
+ "movq (%1,%4), %%mm2 \n\t"
+ "movq (%2,%4), %%mm3 \n\t"
+ PAVGBP_MMX_NO_RND(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5)
+ "movq %%mm4, (%3) \n\t"
+ "movq %%mm5, (%3,%4) \n\t"
+
+ "movq (%1,%4,2), %%mm0 \n\t"
+ "movq (%2,%4,2), %%mm1 \n\t"
+ "movq (%1,%5), %%mm2 \n\t"
+ "movq (%2,%5), %%mm3 \n\t"
+ "lea (%1,%4,4), %1 \n\t"
+ "lea (%2,%4,4), %2 \n\t"
+ PAVGBP_MMX_NO_RND(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5)
+ "movq %%mm4, (%3,%4,2) \n\t"
+ "movq %%mm5, (%3,%5) \n\t"
+ "lea (%3,%4,4), %3 \n\t"
+ "subl $4, %0 \n\t"
+ "jnz 1b \n\t"
+ :"+r"(h), "+r"(a), "+r"(b), "+r"(dst)
+ :"r"((x86_reg)stride), "r"((x86_reg)3L*stride)
+ :"memory");
+// STOP_TIMER("put_vp_no_rnd_pixels8_l2_mmx")
+}
+static void put_vp_no_rnd_pixels16_l2_mmx(uint8_t *dst, const uint8_t *a, const uint8_t *b, int stride, int h)
+{
+ put_vp_no_rnd_pixels8_l2_mmx(dst, a, b, stride, h);
+ put_vp_no_rnd_pixels8_l2_mmx(dst+8, a+8, b+8, stride, h);
+}
+
+#if CONFIG_DIRAC_DECODER
+#define DIRAC_PIXOP(OPNAME, EXT)\
+void ff_ ## OPNAME ## _dirac_pixels8_ ## EXT(uint8_t *dst, const uint8_t *src[5], int stride, int h)\
+{\
+ OPNAME ## _pixels8_ ## EXT(dst, src[0], stride, h);\
+}\
+void ff_ ## OPNAME ## _dirac_pixels16_ ## EXT(uint8_t *dst, const uint8_t *src[5], int stride, int h)\
+{\
+ OPNAME ## _pixels16_ ## EXT(dst, src[0], stride, h);\
+}\
+void ff_ ## OPNAME ## _dirac_pixels32_ ## EXT(uint8_t *dst, const uint8_t *src[5], int stride, int h)\
+{\
+ OPNAME ## _pixels16_ ## EXT(dst , src[0] , stride, h);\
+ OPNAME ## _pixels16_ ## EXT(dst+16, src[0]+16, stride, h);\
+}
+
+DIRAC_PIXOP(put, mmx)
+DIRAC_PIXOP(avg, mmx)
+DIRAC_PIXOP(avg, mmx2)
+
+void ff_put_dirac_pixels16_sse2(uint8_t *dst, const uint8_t *src[5], int stride, int h)
+{
+ put_pixels16_sse2(dst, src[0], stride, h);
+}
+void ff_avg_dirac_pixels16_sse2(uint8_t *dst, const uint8_t *src[5], int stride, int h)
+{
+ avg_pixels16_sse2(dst, src[0], stride, h);
+}
+void ff_put_dirac_pixels32_sse2(uint8_t *dst, const uint8_t *src[5], int stride, int h)
+{
+ put_pixels16_sse2(dst , src[0] , stride, h);
+ put_pixels16_sse2(dst+16, src[0]+16, stride, h);
+}
+void ff_avg_dirac_pixels32_sse2(uint8_t *dst, const uint8_t *src[5], int stride, int h)
+{
+ avg_pixels16_sse2(dst , src[0] , stride, h);
+ avg_pixels16_sse2(dst+16, src[0]+16, stride, h);
+}
+#endif
+
+/* XXX: Those functions should be suppressed ASAP when all IDCTs are
+ * converted. */
+#if CONFIG_GPL
+static void ff_libmpeg2mmx_idct_put(uint8_t *dest, int line_size,
+ DCTELEM *block)
+{
+ ff_mmx_idct(block);
+ ff_put_pixels_clamped_mmx(block, dest, line_size);
+}
+
+static void ff_libmpeg2mmx_idct_add(uint8_t *dest, int line_size,
+ DCTELEM *block)
+{
+ ff_mmx_idct(block);
+ ff_add_pixels_clamped_mmx(block, dest, line_size);
+}
+
+static void ff_libmpeg2mmx2_idct_put(uint8_t *dest, int line_size,
+ DCTELEM *block)
+{
+ ff_mmxext_idct(block);
+ ff_put_pixels_clamped_mmx(block, dest, line_size);
+}
+
+static void ff_libmpeg2mmx2_idct_add(uint8_t *dest, int line_size,
+ DCTELEM *block)
+{
+ ff_mmxext_idct(block);
+ ff_add_pixels_clamped_mmx(block, dest, line_size);
+}
+#endif
+
static void vorbis_inverse_coupling_3dnow(float *mag, float *ang, int blocksize)
{
int i;
SET_HPEL_FUNCS(put_no_rnd, 1, 8, mmx);
SET_HPEL_FUNCS(avg, 1, 8, mmx);
SET_HPEL_FUNCS(avg_no_rnd, 1, 8, mmx);
-
- switch (avctx->idct_algo) {
- case FF_IDCT_AUTO:
- case FF_IDCT_SIMPLEMMX:
- c->idct_put = ff_simple_idct_put_mmx;
- c->idct_add = ff_simple_idct_add_mmx;
- c->idct = ff_simple_idct_mmx;
- c->idct_permutation_type = FF_SIMPLE_IDCT_PERM;
- break;
- case FF_IDCT_XVIDMMX:
- c->idct_put = ff_idct_xvid_mmx_put;
- c->idct_add = ff_idct_xvid_mmx_add;
- c->idct = ff_idct_xvid_mmx;
- break;
- }
}
#if ARCH_X86_32 || !HAVE_YASM
c->add_bytes = add_bytes_mmx;
+ c->put_no_rnd_pixels_l2[0]= put_vp_no_rnd_pixels16_l2_mmx;
+ c->put_no_rnd_pixels_l2[1]= put_vp_no_rnd_pixels8_l2_mmx;
+
if (CONFIG_H263_DECODER || CONFIG_H263_ENCODER) {
c->h263_v_loop_filter = h263_v_loop_filter_mmx;
c->h263_h_loop_filter = h263_h_loop_filter_mmx;
}
}
- if (!high_bit_depth && avctx->idct_algo == FF_IDCT_XVIDMMX) {
- c->idct_put = ff_idct_xvid_mmx2_put;
- c->idct_add = ff_idct_xvid_mmx2_add;
- c->idct = ff_idct_xvid_mmx2;
- }
-
if (CONFIG_VP3_DECODER && (avctx->codec_id == AV_CODEC_ID_VP3 ||
avctx->codec_id == AV_CODEC_ID_THEORA)) {
c->put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels8_x2_exact_mmx2;
H264_QPEL_FUNCS(3, 2, sse2);
H264_QPEL_FUNCS(3, 3, sse2);
}
-
- if (!high_bit_depth && avctx->idct_algo == FF_IDCT_XVIDMMX) {
- c->idct_put = ff_idct_xvid_sse2_put;
- c->idct_add = ff_idct_xvid_sse2_add;
- c->idct = ff_idct_xvid_sse2;
- c->idct_permutation_type = FF_SSE2_IDCT_PERM;
- }
#endif /* HAVE_INLINE_ASM */
#if HAVE_YASM
c->add_hfyu_median_prediction = add_hfyu_median_prediction_cmov;
#endif
- if (mm_flags & AV_CPU_FLAG_MMX)
+ if (mm_flags & AV_CPU_FLAG_MMX) {
+#if HAVE_INLINE_ASM
+ const int idct_algo = avctx->idct_algo;
+
+ if (avctx->lowres == 0 && avctx->bits_per_raw_sample <= 8) {
+ if (idct_algo == FF_IDCT_AUTO || idct_algo == FF_IDCT_SIMPLEMMX) {
+ c->idct_put = ff_simple_idct_put_mmx;
+ c->idct_add = ff_simple_idct_add_mmx;
+ c->idct = ff_simple_idct_mmx;
+ c->idct_permutation_type = FF_SIMPLE_IDCT_PERM;
+#if CONFIG_GPL
+ } else if (idct_algo == FF_IDCT_LIBMPEG2MMX) {
+ if (mm_flags & AV_CPU_FLAG_MMX2) {
+ c->idct_put = ff_libmpeg2mmx2_idct_put;
+ c->idct_add = ff_libmpeg2mmx2_idct_add;
+ c->idct = ff_mmxext_idct;
+ } else {
+ c->idct_put = ff_libmpeg2mmx_idct_put;
+ c->idct_add = ff_libmpeg2mmx_idct_add;
+ c->idct = ff_mmx_idct;
+ }
+ c->idct_permutation_type = FF_LIBMPEG2_IDCT_PERM;
+#endif
+ } else if (idct_algo == FF_IDCT_XVIDMMX) {
+ if (mm_flags & AV_CPU_FLAG_SSE2) {
+ c->idct_put = ff_idct_xvid_sse2_put;
+ c->idct_add = ff_idct_xvid_sse2_add;
+ c->idct = ff_idct_xvid_sse2;
+ c->idct_permutation_type = FF_SSE2_IDCT_PERM;
+ } else if (mm_flags & AV_CPU_FLAG_MMXEXT) {
+ c->idct_put = ff_idct_xvid_mmx2_put;
+ c->idct_add = ff_idct_xvid_mmx2_add;
+ c->idct = ff_idct_xvid_mmx2;
+ } else {
+ c->idct_put = ff_idct_xvid_mmx_put;
+ c->idct_add = ff_idct_xvid_mmx_add;
+ c->idct = ff_idct_xvid_mmx;
+ }
+ }
+ }
+#endif /* HAVE_INLINE_ASM */
+
dsputil_init_mmx(c, avctx, mm_flags);
+ }
if (mm_flags & AV_CPU_FLAG_MMXEXT)
dsputil_init_mmx2(c, avctx, mm_flags);
*
* conversion to gcc syntax by Michael Niedermayer
*
- * This file is part of Libav.
+ * This file is part of FFmpeg.
*
- * Libav is free software; you can redistribute it and/or
+ * 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.
*
- * Libav is distributed in the hope that it will be useful,
+ * 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 Libav; if not, write to the Free Software Foundation,
+ * along with FFmpeg; if not, write to the Free Software Foundation,
* Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "config.h"
#include "libavcodec/avcodec.h"
#include "libavutil/mem.h"
+ #include "dsputil_mmx.h"
#include "idct_xvid.h"
#if HAVE_INLINE_ASM
:: "r"(block), "r"(rounder_0), "r"(tab_i_04_xmm), "r"(tg_1_16));
}
+ void ff_idct_xvid_mmx_put(uint8_t *dest, int line_size, DCTELEM *block)
+ {
+ ff_idct_xvid_mmx(block);
+ ff_put_pixels_clamped_mmx(block, dest, line_size);
+ }
+
+ void ff_idct_xvid_mmx_add(uint8_t *dest, int line_size, DCTELEM *block)
+ {
+ ff_idct_xvid_mmx(block);
+ ff_add_pixels_clamped_mmx(block, dest, line_size);
+ }
+
+ void ff_idct_xvid_mmx2_put(uint8_t *dest, int line_size, DCTELEM *block)
+ {
+ ff_idct_xvid_mmx2(block);
+ ff_put_pixels_clamped_mmx(block, dest, line_size);
+ }
+
+ void ff_idct_xvid_mmx2_add(uint8_t *dest, int line_size, DCTELEM *block)
+ {
+ ff_idct_xvid_mmx2(block);
+ ff_add_pixels_clamped_mmx(block, dest, line_size);
+ }
+
#endif /* HAVE_INLINE_ASM */
/*
* XVID MPEG-4 VIDEO CODEC
*
- * This file is part of Libav.
+ * This file is part of FFmpeg.
*
- * Libav is free software; you can redistribute it and/or
+ * 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.
*
- * Libav is distributed in the hope that it will be useful,
+ * 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 Libav; if not, write to the Free Software
+ * License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include <stdint.h>
+ #include "libavcodec/dsputil.h"
+
void ff_idct_xvid_mmx(short *block);
+ void ff_idct_xvid_mmx_put(uint8_t *dest, int line_size, DCTELEM *block);
+ void ff_idct_xvid_mmx_add(uint8_t *dest, int line_size, DCTELEM *block);
+
void ff_idct_xvid_mmx2(short *block);
+ void ff_idct_xvid_mmx2_put(uint8_t *dest, int line_size, DCTELEM *block);
+ void ff_idct_xvid_mmx2_add(uint8_t *dest, int line_size, DCTELEM *block);
+
void ff_idct_xvid_sse2(short *block);
void ff_idct_xvid_sse2_put(uint8_t *dest, int line_size, short *block);
void ff_idct_xvid_sse2_add(uint8_t *dest, int line_size, short *block);
--- /dev/null
- * This file is part of Libav.
+ /*
- * Libav is free software; you can redistribute it and/or
++ * This file is part of FFmpeg.
+ *
- * Libav is distributed in the hope that it will be useful,
++ * 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.
+ *
- * License along with Libav; if not, write to the Free Software
++ * 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 <stdint.h>
+
+ #include "config.h"
+ #include "libavutil/attributes.h"
+ #include "libavutil/cpu.h"
+ #include "libavutil/arm/cpu.h"
+ #include "libavutil/samplefmt.h"
+ #include "libavresample/audio_convert.h"
+
+ void ff_conv_flt_to_s16_neon(int16_t *dst, const float *src, int len);
+ void ff_conv_fltp_to_s16_neon(int16_t *dst, float *const *src,
+ int len, int channels);
+ void ff_conv_fltp_to_s16_2ch_neon(int16_t *dst, float *const *src,
+ int len, int channels);
+
+ av_cold void ff_audio_convert_init_arm(AudioConvert *ac)
+ {
+ int cpu_flags = av_get_cpu_flags();
+
+ if (have_neon(cpu_flags)) {
+ ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_FLT,
+ 0, 16, 8, "NEON",
+ ff_conv_flt_to_s16_neon);
+ ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_FLTP,
+ 0, 16, 8, "NEON",
+ ff_conv_fltp_to_s16_neon);
+ ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_FLTP,
+ 2, 16, 8, "NEON",
+ ff_conv_fltp_to_s16_2ch_neon);
+ }
+ }
--- /dev/null
- * This file is part of Libav.
+ /*
+ * Copyright (c) 2008 Mans Rullgard <mans@mansr.com>
+ *
- * Libav is free software; you can redistribute it and/or
++ * This file is part of FFmpeg
+ *
- * Libav is distributed in the hope that it will be useful,
++ * 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.
+ *
- * License along with Libav; if not, write to the Free Software
++ * 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 "config.h"
+ #include "libavutil/arm/asm.S"
+
+ function ff_conv_flt_to_s16_neon, export=1
+ subs r2, r2, #8
+ vld1.32 {q0}, [r1,:128]!
+ vcvt.s32.f32 q8, q0, #31
+ vld1.32 {q1}, [r1,:128]!
+ vcvt.s32.f32 q9, q1, #31
+ beq 3f
+ bics r12, r2, #15
+ beq 2f
+ 1: subs r12, r12, #16
+ vqrshrn.s32 d4, q8, #16
+ vld1.32 {q0}, [r1,:128]!
+ vcvt.s32.f32 q0, q0, #31
+ vqrshrn.s32 d5, q9, #16
+ vld1.32 {q1}, [r1,:128]!
+ vcvt.s32.f32 q1, q1, #31
+ vqrshrn.s32 d6, q0, #16
+ vst1.16 {q2}, [r0,:128]!
+ vqrshrn.s32 d7, q1, #16
+ vld1.32 {q8}, [r1,:128]!
+ vcvt.s32.f32 q8, q8, #31
+ vld1.32 {q9}, [r1,:128]!
+ vcvt.s32.f32 q9, q9, #31
+ vst1.16 {q3}, [r0,:128]!
+ bne 1b
+ ands r2, r2, #15
+ beq 3f
+ 2: vld1.32 {q0}, [r1,:128]!
+ vqrshrn.s32 d4, q8, #16
+ vcvt.s32.f32 q0, q0, #31
+ vld1.32 {q1}, [r1,:128]!
+ vqrshrn.s32 d5, q9, #16
+ vcvt.s32.f32 q1, q1, #31
+ vqrshrn.s32 d6, q0, #16
+ vst1.16 {q2}, [r0,:128]!
+ vqrshrn.s32 d7, q1, #16
+ vst1.16 {q3}, [r0,:128]!
+ bx lr
+ 3: vqrshrn.s32 d4, q8, #16
+ vqrshrn.s32 d5, q9, #16
+ vst1.16 {q2}, [r0,:128]!
+ bx lr
+ endfunc
+
+ function ff_conv_fltp_to_s16_2ch_neon, export=1
+ ldm r1, {r1, r3}
+ subs r2, r2, #8
+ vld1.32 {q0}, [r1,:128]!
+ vcvt.s32.f32 q8, q0, #31
+ vld1.32 {q1}, [r1,:128]!
+ vcvt.s32.f32 q9, q1, #31
+ vld1.32 {q10}, [r3,:128]!
+ vcvt.s32.f32 q10, q10, #31
+ vld1.32 {q11}, [r3,:128]!
+ vcvt.s32.f32 q11, q11, #31
+ beq 3f
+ bics r12, r2, #15
+ beq 2f
+ 1: subs r12, r12, #16
+ vld1.32 {q0}, [r1,:128]!
+ vcvt.s32.f32 q0, q0, #31
+ vsri.32 q10, q8, #16
+ vld1.32 {q1}, [r1,:128]!
+ vcvt.s32.f32 q1, q1, #31
+ vld1.32 {q12}, [r3,:128]!
+ vcvt.s32.f32 q12, q12, #31
+ vld1.32 {q13}, [r3,:128]!
+ vsri.32 q11, q9, #16
+ vst1.16 {q10}, [r0,:128]!
+ vcvt.s32.f32 q13, q13, #31
+ vst1.16 {q11}, [r0,:128]!
+ vsri.32 q12, q0, #16
+ vld1.32 {q8}, [r1,:128]!
+ vsri.32 q13, q1, #16
+ vst1.16 {q12}, [r0,:128]!
+ vcvt.s32.f32 q8, q8, #31
+ vld1.32 {q9}, [r1,:128]!
+ vcvt.s32.f32 q9, q9, #31
+ vld1.32 {q10}, [r3,:128]!
+ vcvt.s32.f32 q10, q10, #31
+ vld1.32 {q11}, [r3,:128]!
+ vcvt.s32.f32 q11, q11, #31
+ vst1.16 {q13}, [r0,:128]!
+ bne 1b
+ ands r2, r2, #15
+ beq 3f
+ 2: vsri.32 q10, q8, #16
+ vld1.32 {q0}, [r1,:128]!
+ vcvt.s32.f32 q0, q0, #31
+ vld1.32 {q1}, [r1,:128]!
+ vcvt.s32.f32 q1, q1, #31
+ vld1.32 {q12}, [r3,:128]!
+ vcvt.s32.f32 q12, q12, #31
+ vsri.32 q11, q9, #16
+ vld1.32 {q13}, [r3,:128]!
+ vcvt.s32.f32 q13, q13, #31
+ vst1.16 {q10}, [r0,:128]!
+ vsri.32 q12, q0, #16
+ vst1.16 {q11}, [r0,:128]!
+ vsri.32 q13, q1, #16
+ vst1.16 {q12-q13},[r0,:128]!
+ bx lr
+ 3: vsri.32 q10, q8, #16
+ vsri.32 q11, q9, #16
+ vst1.16 {q10-q11},[r0,:128]!
+ bx lr
+ endfunc
+
+ function ff_conv_fltp_to_s16_neon, export=1
+ cmp r3, #2
+ itt lt
+ ldrlt r1, [r1]
+ blt ff_conv_flt_to_s16_neon
+ beq ff_conv_fltp_to_s16_2ch_neon
+
+ push {r4-r8, lr}
+ cmp r3, #4
+ lsl r12, r3, #1
+ blt 4f
+
+ @ 4 channels
+ 5: ldm r1!, {r4-r7}
+ mov lr, r2
+ mov r8, r0
+ vld1.32 {q8}, [r4,:128]!
+ vcvt.s32.f32 q8, q8, #31
+ vld1.32 {q9}, [r5,:128]!
+ vcvt.s32.f32 q9, q9, #31
+ vld1.32 {q10}, [r6,:128]!
+ vcvt.s32.f32 q10, q10, #31
+ vld1.32 {q11}, [r7,:128]!
+ vcvt.s32.f32 q11, q11, #31
+ 6: subs lr, lr, #8
+ vld1.32 {q0}, [r4,:128]!
+ vcvt.s32.f32 q0, q0, #31
+ vsri.32 q9, q8, #16
+ vld1.32 {q1}, [r5,:128]!
+ vcvt.s32.f32 q1, q1, #31
+ vsri.32 q11, q10, #16
+ vld1.32 {q2}, [r6,:128]!
+ vcvt.s32.f32 q2, q2, #31
+ vzip.32 d18, d22
+ vld1.32 {q3}, [r7,:128]!
+ vcvt.s32.f32 q3, q3, #31
+ vzip.32 d19, d23
+ vst1.16 {d18}, [r8], r12
+ vsri.32 q1, q0, #16
+ vst1.16 {d22}, [r8], r12
+ vsri.32 q3, q2, #16
+ vst1.16 {d19}, [r8], r12
+ vzip.32 d2, d6
+ vst1.16 {d23}, [r8], r12
+ vzip.32 d3, d7
+ beq 7f
+ vld1.32 {q8}, [r4,:128]!
+ vcvt.s32.f32 q8, q8, #31
+ vst1.16 {d2}, [r8], r12
+ vld1.32 {q9}, [r5,:128]!
+ vcvt.s32.f32 q9, q9, #31
+ vst1.16 {d6}, [r8], r12
+ vld1.32 {q10}, [r6,:128]!
+ vcvt.s32.f32 q10, q10, #31
+ vst1.16 {d3}, [r8], r12
+ vld1.32 {q11}, [r7,:128]!
+ vcvt.s32.f32 q11, q11, #31
+ vst1.16 {d7}, [r8], r12
+ b 6b
+ 7: vst1.16 {d2}, [r8], r12
+ vst1.16 {d6}, [r8], r12
+ vst1.16 {d3}, [r8], r12
+ vst1.16 {d7}, [r8], r12
+ subs r3, r3, #4
+ it eq
+ popeq {r4-r8, pc}
+ cmp r3, #4
+ add r0, r0, #8
+ bge 5b
+
+ @ 2 channels
+ 4: cmp r3, #2
+ blt 4f
+ ldm r1!, {r4-r5}
+ mov lr, r2
+ mov r8, r0
+ tst lr, #8
+ vld1.32 {q8}, [r4,:128]!
+ vcvt.s32.f32 q8, q8, #31
+ vld1.32 {q9}, [r5,:128]!
+ vcvt.s32.f32 q9, q9, #31
+ vld1.32 {q10}, [r4,:128]!
+ vcvt.s32.f32 q10, q10, #31
+ vld1.32 {q11}, [r5,:128]!
+ vcvt.s32.f32 q11, q11, #31
+ beq 6f
+ subs lr, lr, #8
+ beq 7f
+ vsri.32 d18, d16, #16
+ vsri.32 d19, d17, #16
+ vld1.32 {q8}, [r4,:128]!
+ vcvt.s32.f32 q8, q8, #31
+ vst1.32 {d18[0]}, [r8], r12
+ vsri.32 d22, d20, #16
+ vst1.32 {d18[1]}, [r8], r12
+ vsri.32 d23, d21, #16
+ vst1.32 {d19[0]}, [r8], r12
+ vst1.32 {d19[1]}, [r8], r12
+ vld1.32 {q9}, [r5,:128]!
+ vcvt.s32.f32 q9, q9, #31
+ vst1.32 {d22[0]}, [r8], r12
+ vst1.32 {d22[1]}, [r8], r12
+ vld1.32 {q10}, [r4,:128]!
+ vcvt.s32.f32 q10, q10, #31
+ vst1.32 {d23[0]}, [r8], r12
+ vst1.32 {d23[1]}, [r8], r12
+ vld1.32 {q11}, [r5,:128]!
+ vcvt.s32.f32 q11, q11, #31
+ 6: subs lr, lr, #16
+ vld1.32 {q0}, [r4,:128]!
+ vcvt.s32.f32 q0, q0, #31
+ vsri.32 d18, d16, #16
+ vld1.32 {q1}, [r5,:128]!
+ vcvt.s32.f32 q1, q1, #31
+ vsri.32 d19, d17, #16
+ vld1.32 {q2}, [r4,:128]!
+ vcvt.s32.f32 q2, q2, #31
+ vld1.32 {q3}, [r5,:128]!
+ vcvt.s32.f32 q3, q3, #31
+ vst1.32 {d18[0]}, [r8], r12
+ vsri.32 d22, d20, #16
+ vst1.32 {d18[1]}, [r8], r12
+ vsri.32 d23, d21, #16
+ vst1.32 {d19[0]}, [r8], r12
+ vsri.32 d2, d0, #16
+ vst1.32 {d19[1]}, [r8], r12
+ vsri.32 d3, d1, #16
+ vst1.32 {d22[0]}, [r8], r12
+ vsri.32 d6, d4, #16
+ vst1.32 {d22[1]}, [r8], r12
+ vsri.32 d7, d5, #16
+ vst1.32 {d23[0]}, [r8], r12
+ vst1.32 {d23[1]}, [r8], r12
+ beq 6f
+ vld1.32 {q8}, [r4,:128]!
+ vcvt.s32.f32 q8, q8, #31
+ vst1.32 {d2[0]}, [r8], r12
+ vst1.32 {d2[1]}, [r8], r12
+ vld1.32 {q9}, [r5,:128]!
+ vcvt.s32.f32 q9, q9, #31
+ vst1.32 {d3[0]}, [r8], r12
+ vst1.32 {d3[1]}, [r8], r12
+ vld1.32 {q10}, [r4,:128]!
+ vcvt.s32.f32 q10, q10, #31
+ vst1.32 {d6[0]}, [r8], r12
+ vst1.32 {d6[1]}, [r8], r12
+ vld1.32 {q11}, [r5,:128]!
+ vcvt.s32.f32 q11, q11, #31
+ vst1.32 {d7[0]}, [r8], r12
+ vst1.32 {d7[1]}, [r8], r12
+ bgt 6b
+ 6: vst1.32 {d2[0]}, [r8], r12
+ vst1.32 {d2[1]}, [r8], r12
+ vst1.32 {d3[0]}, [r8], r12
+ vst1.32 {d3[1]}, [r8], r12
+ vst1.32 {d6[0]}, [r8], r12
+ vst1.32 {d6[1]}, [r8], r12
+ vst1.32 {d7[0]}, [r8], r12
+ vst1.32 {d7[1]}, [r8], r12
+ b 8f
+ 7: vsri.32 d18, d16, #16
+ vsri.32 d19, d17, #16
+ vst1.32 {d18[0]}, [r8], r12
+ vsri.32 d22, d20, #16
+ vst1.32 {d18[1]}, [r8], r12
+ vsri.32 d23, d21, #16
+ vst1.32 {d19[0]}, [r8], r12
+ vst1.32 {d19[1]}, [r8], r12
+ vst1.32 {d22[0]}, [r8], r12
+ vst1.32 {d22[1]}, [r8], r12
+ vst1.32 {d23[0]}, [r8], r12
+ vst1.32 {d23[1]}, [r8], r12
+ 8: subs r3, r3, #2
+ add r0, r0, #4
+ it eq
+ popeq {r4-r8, pc}
+
+ @ 1 channel
+ 4: ldr r4, [r1]
+ tst r2, #8
+ mov lr, r2
+ mov r5, r0
+ vld1.32 {q0}, [r4,:128]!
+ vcvt.s32.f32 q0, q0, #31
+ vld1.32 {q1}, [r4,:128]!
+ vcvt.s32.f32 q1, q1, #31
+ bne 8f
+ 6: subs lr, lr, #16
+ vld1.32 {q2}, [r4,:128]!
+ vcvt.s32.f32 q2, q2, #31
+ vld1.32 {q3}, [r4,:128]!
+ vcvt.s32.f32 q3, q3, #31
+ vst1.16 {d0[1]}, [r5,:16], r12
+ vst1.16 {d0[3]}, [r5,:16], r12
+ vst1.16 {d1[1]}, [r5,:16], r12
+ vst1.16 {d1[3]}, [r5,:16], r12
+ vst1.16 {d2[1]}, [r5,:16], r12
+ vst1.16 {d2[3]}, [r5,:16], r12
+ vst1.16 {d3[1]}, [r5,:16], r12
+ vst1.16 {d3[3]}, [r5,:16], r12
+ beq 7f
+ vld1.32 {q0}, [r4,:128]!
+ vcvt.s32.f32 q0, q0, #31
+ vld1.32 {q1}, [r4,:128]!
+ vcvt.s32.f32 q1, q1, #31
+ 7: vst1.16 {d4[1]}, [r5,:16], r12
+ vst1.16 {d4[3]}, [r5,:16], r12
+ vst1.16 {d5[1]}, [r5,:16], r12
+ vst1.16 {d5[3]}, [r5,:16], r12
+ vst1.16 {d6[1]}, [r5,:16], r12
+ vst1.16 {d6[3]}, [r5,:16], r12
+ vst1.16 {d7[1]}, [r5,:16], r12
+ vst1.16 {d7[3]}, [r5,:16], r12
+ bgt 6b
+ pop {r4-r8, pc}
+ 8: subs lr, lr, #8
+ vst1.16 {d0[1]}, [r5,:16], r12
+ vst1.16 {d0[3]}, [r5,:16], r12
+ vst1.16 {d1[1]}, [r5,:16], r12
+ vst1.16 {d1[3]}, [r5,:16], r12
+ vst1.16 {d2[1]}, [r5,:16], r12
+ vst1.16 {d2[3]}, [r5,:16], r12
+ vst1.16 {d3[1]}, [r5,:16], r12
+ vst1.16 {d3[3]}, [r5,:16], r12
+ it eq
+ popeq {r4-r8, pc}
+ vld1.32 {q0}, [r4,:128]!
+ vcvt.s32.f32 q0, q0, #31
+ vld1.32 {q1}, [r4,:128]!
+ vcvt.s32.f32 q1, q1, #31
+ b 6b
+ endfunc
/*
* Copyright (c) 2007 Michael Niedermayer
*
- * This file is part of Libav.
+ * This file is part of FFmpeg.
*
- * Libav is free software; you can redistribute it and/or
+ * 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.
*
- * Libav is distributed in the hope that it will be useful,
+ * 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 Libav; if not, write to the Free Software
+ * License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include <stdio.h>
#include <stdlib.h>
#include <inttypes.h>
+ #include <errno.h>
+ #include <string.h>
static uint32_t state;
static uint32_t ran(void)
projects / ffmpeg.git / commitdiff