X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavcodec%2Fdct-test.c;h=9fdf7d179db63aff3d7454a0805a5d8792014b03;hb=018a6645bc606b545e6efdd980d9449c33e09b00;hp=1665682698fabf0630e1396808b75f2066a1a016;hpb=4dccfff9dd01d6c45731cda21fc6ea0a649d6581;p=ffmpeg diff --git a/libavcodec/dct-test.c b/libavcodec/dct-test.c index 1665682698f..9fdf7d179db 100644 --- a/libavcodec/dct-test.c +++ b/libavcodec/dct-test.c @@ -2,20 +2,20 @@ * (c) 2001 Fabrice Bellard * 2007 Marc Hoffman * - * 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 */ @@ -71,7 +71,7 @@ struct algo { 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; }; @@ -107,6 +107,22 @@ static const struct algo fdct_tab[] = { { 0 } }; +#if HAVE_MMX +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(uint16_t *dst){ + int16_t qmat[64]; int i; + int16_t tmp[64]; + + 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 }, @@ -122,6 +138,9 @@ static const struct algo idct_tab[] = { { "XVID-MMX", ff_idct_xvid_mmx, NO_PERM, AV_CPU_FLAG_MMX, 1 }, { "XVID-MMX2", ff_idct_xvid_mmx2, NO_PERM, AV_CPU_FLAG_MMX2, 1 }, { "XVID-SSE2", ff_idct_xvid_sse2, SSE2_PERM, AV_CPU_FLAG_SSE2, 1 }, +#if ARCH_X86_64 + { "PR-SSE2", ff_prores_idct_put_10_sse2_wrap, TRANSPOSE_PERM, AV_CPU_FLAG_SSE2, 1 }, +#endif #endif #if ARCH_BFIN @@ -151,8 +170,6 @@ static const struct algo idct_tab[] = { #define AANSCALE_BITS 12 -static uint8_t cropTbl[256 + 2 * MAX_NEG_CROP]; - static int64_t gettime(void) { struct timeval tv; @@ -199,7 +216,7 @@ static inline void mmx_emms(void) #endif } -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; @@ -208,7 +225,7 @@ static void init_block(DCTELEM block[64], int test, int is_idct, AVLFG *prng) 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++) @@ -218,10 +235,10 @@ static void init_block(DCTELEM block[64], int test, int is_idct, AVLFG *prng) 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; + block[av_lfg_get(prng) % 64] = 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; } @@ -243,13 +260,16 @@ static void permute(DCTELEM dst[64], const DCTELEM src[64], int perm) } 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; @@ -259,6 +279,7 @@ static int dct_error(const struct algo *dct, int test, int is_idct, int speed) int maxout = 0; int blockSumErrMax = 0, blockSumErr; AVLFG prng; + const int vals=1<format); dct->func(block); @@ -315,7 +336,7 @@ static int dct_error(const struct algo *dct, int test, int is_idct, int speed) 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); @@ -327,7 +348,8 @@ static int dct_error(const struct algo *dct, int test, int is_idct, int speed) 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 = gettime(); @@ -465,6 +487,25 @@ static void idct248_error(const char *name, 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); @@ -490,10 +531,11 @@ static void idct248_error(const char *name, static void help(void) { - printf("dct-test [-i] []\n" + printf("dct-test [-i] [] []\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"); @@ -506,19 +548,13 @@ int main(int argc, char **argv) int test = 1; int speed = 0; int err = 0; + int bits=8; cpu_flags = av_get_cpu_flags(); ff_ref_dct_init(); idct_mmx_init(); - for (i = 0; i < 256; i++) - cropTbl[i + MAX_NEG_CROP] = i; - for (i = 0; i < MAX_NEG_CROP; i++) { - cropTbl[i] = 0; - cropTbl[i + MAX_NEG_CROP + 256] = 255; - } - for (;;) { c = getopt(argc, argv, "ih4t"); if (c == -1) @@ -542,8 +578,9 @@ int main(int argc, char **argv) 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); @@ -551,7 +588,7 @@ int main(int argc, char **argv) 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); } }