/*
* (c) 2001 Fabrice Bellard
+ * 2007 Marc Hoffman <marc.hoffman@analog.com>
*
- * This file is part of FFmpeg.
+ * This file is part of Libav.
*
- * FFmpeg is free software; you can redistribute it and/or
+ * Libav 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,
+ * Libav 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
+ * License along with Libav; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
/**
- * @file dct-test.c
- * DCT test. (c) 2001 Fabrice Bellard.
+ * @file
+ * DCT test (c) 2001 Fabrice Bellard
* Started from sample code by Juan J. Sierralta P.
*/
+#include "config.h"
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
-#include <sys/time.h>
+#if HAVE_UNISTD_H
#include <unistd.h>
+#endif
+#include <math.h>
-#include "dsputil.h"
+#include "libavutil/cpu.h"
+#include "libavutil/common.h"
+#include "libavutil/lfg.h"
+#include "libavutil/time.h"
+#include "dct.h"
#include "simple_idct.h"
+#include "aandcttab.h"
#include "faandct.h"
-
-#ifndef MAX
-#define MAX(a, b) (((a) > (b)) ? (a) : (b))
-#endif
+#include "faanidct.h"
+#include "x86/idct_xvid.h"
+#include "dctref.h"
#undef printf
-void *fast_memcpy(void *a, const void *b, size_t c){return memcpy(a,b,c);};
+// BFIN
+void ff_bfin_idct(int16_t *block);
+void ff_bfin_fdct(int16_t *block);
+
+// ALTIVEC
+void ff_fdct_altivec(int16_t *block);
+
+// ARM
+void ff_j_rev_dct_arm(int16_t *data);
+void ff_simple_idct_arm(int16_t *data);
+void ff_simple_idct_armv5te(int16_t *data);
+void ff_simple_idct_armv6(int16_t *data);
+void ff_simple_idct_neon(int16_t *data);
+
+void ff_simple_idct_axp(int16_t *data);
+
+struct algo {
+ const char *name;
+ void (*func)(int16_t *block);
+ enum formattag { NO_PERM, MMX_PERM, MMX_SIMPLE_PERM, SCALE_PERM,
+ SSE2_PERM, PARTTRANS_PERM } format;
+ int mm_support;
+ int nonspec;
+};
-/* reference fdct/idct */
-extern void fdct(DCTELEM *block);
-extern void idct(DCTELEM *block);
-extern void ff_idct_xvid_mmx(DCTELEM *block);
-extern void ff_idct_xvid_mmx2(DCTELEM *block);
-extern void init_fdct();
+static int cpu_flags;
-extern void j_rev_dct(DCTELEM *data);
-extern void ff_mmx_idct(DCTELEM *data);
-extern void ff_mmxext_idct(DCTELEM *data);
+static const struct algo fdct_tab[] = {
+ { "REF-DBL", ff_ref_fdct, NO_PERM },
+ { "FAAN", ff_faandct, NO_PERM },
+ { "IJG-AAN-INT", ff_fdct_ifast, SCALE_PERM },
+ { "IJG-LLM-INT", ff_jpeg_fdct_islow_8, NO_PERM },
-extern void odivx_idct_c (short *block);
+#if HAVE_MMX_INLINE
+ { "MMX", ff_fdct_mmx, NO_PERM, AV_CPU_FLAG_MMX },
+ { "MMXEXT", ff_fdct_mmxext, NO_PERM, AV_CPU_FLAG_MMXEXT },
+ { "SSE2", ff_fdct_sse2, NO_PERM, AV_CPU_FLAG_SSE2 },
+#endif
-#define AANSCALE_BITS 12
-static const unsigned short aanscales[64] = {
- /* precomputed values scaled up by 14 bits */
- 16384, 22725, 21407, 19266, 16384, 12873, 8867, 4520,
- 22725, 31521, 29692, 26722, 22725, 17855, 12299, 6270,
- 21407, 29692, 27969, 25172, 21407, 16819, 11585, 5906,
- 19266, 26722, 25172, 22654, 19266, 15137, 10426, 5315,
- 16384, 22725, 21407, 19266, 16384, 12873, 8867, 4520,
- 12873, 17855, 16819, 15137, 12873, 10114, 6967, 3552,
- 8867, 12299, 11585, 10426, 8867, 6967, 4799, 2446,
- 4520, 6270, 5906, 5315, 4520, 3552, 2446, 1247
+#if HAVE_ALTIVEC
+ { "altivecfdct", ff_fdct_altivec, NO_PERM, AV_CPU_FLAG_ALTIVEC },
+#endif
+
+#if ARCH_BFIN
+ { "BFINfdct", ff_bfin_fdct, NO_PERM },
+#endif
+
+ { 0 }
};
-uint8_t cropTbl[256 + 2 * MAX_NEG_CROP];
+static const struct algo idct_tab[] = {
+ { "FAANI", ff_faanidct, NO_PERM },
+ { "REF-DBL", ff_ref_idct, NO_PERM },
+ { "INT", ff_j_rev_dct, MMX_PERM },
+ { "SIMPLE-C", ff_simple_idct_8, NO_PERM },
+
+#if HAVE_MMX_INLINE
+ { "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_mmxext, NO_PERM, AV_CPU_FLAG_MMXEXT, 1 },
+ { "XVID-SSE2", ff_idct_xvid_sse2, SSE2_PERM, AV_CPU_FLAG_SSE2, 1 },
+#endif
-int64_t gettime(void)
-{
- struct timeval tv;
- gettimeofday(&tv,NULL);
- return (int64_t)tv.tv_sec * 1000000 + tv.tv_usec;
-}
+#if ARCH_BFIN
+ { "BFINidct", ff_bfin_idct, NO_PERM },
+#endif
+
+#if ARCH_ARM
+ { "SIMPLE-ARM", ff_simple_idct_arm, NO_PERM },
+ { "INT-ARM", ff_j_rev_dct_arm, MMX_PERM },
+#endif
+#if HAVE_ARMV5TE
+ { "SIMPLE-ARMV5TE", ff_simple_idct_armv5te,NO_PERM, AV_CPU_FLAG_ARMV5TE },
+#endif
+#if HAVE_ARMV6
+ { "SIMPLE-ARMV6", ff_simple_idct_armv6, MMX_PERM, AV_CPU_FLAG_ARMV6 },
+#endif
+#if HAVE_NEON
+ { "SIMPLE-NEON", ff_simple_idct_neon, PARTTRANS_PERM, AV_CPU_FLAG_NEON },
+#endif
+
+#if ARCH_ALPHA
+ { "SIMPLE-ALPHA", ff_simple_idct_axp, NO_PERM },
+#endif
+
+ { 0 }
+};
+
+#define AANSCALE_BITS 12
#define NB_ITS 20000
#define NB_ITS_SPEED 50000
static short idct_mmx_perm[64];
-static short idct_simple_mmx_perm[64]={
- 0x00, 0x08, 0x04, 0x09, 0x01, 0x0C, 0x05, 0x0D,
- 0x10, 0x18, 0x14, 0x19, 0x11, 0x1C, 0x15, 0x1D,
- 0x20, 0x28, 0x24, 0x29, 0x21, 0x2C, 0x25, 0x2D,
- 0x12, 0x1A, 0x16, 0x1B, 0x13, 0x1E, 0x17, 0x1F,
- 0x02, 0x0A, 0x06, 0x0B, 0x03, 0x0E, 0x07, 0x0F,
- 0x30, 0x38, 0x34, 0x39, 0x31, 0x3C, 0x35, 0x3D,
- 0x22, 0x2A, 0x26, 0x2B, 0x23, 0x2E, 0x27, 0x2F,
- 0x32, 0x3A, 0x36, 0x3B, 0x33, 0x3E, 0x37, 0x3F,
+static short idct_simple_mmx_perm[64] = {
+ 0x00, 0x08, 0x04, 0x09, 0x01, 0x0C, 0x05, 0x0D,
+ 0x10, 0x18, 0x14, 0x19, 0x11, 0x1C, 0x15, 0x1D,
+ 0x20, 0x28, 0x24, 0x29, 0x21, 0x2C, 0x25, 0x2D,
+ 0x12, 0x1A, 0x16, 0x1B, 0x13, 0x1E, 0x17, 0x1F,
+ 0x02, 0x0A, 0x06, 0x0B, 0x03, 0x0E, 0x07, 0x0F,
+ 0x30, 0x38, 0x34, 0x39, 0x31, 0x3C, 0x35, 0x3D,
+ 0x22, 0x2A, 0x26, 0x2B, 0x23, 0x2E, 0x27, 0x2F,
+ 0x32, 0x3A, 0x36, 0x3B, 0x33, 0x3E, 0x37, 0x3F,
};
-void idct_mmx_init(void)
+static const uint8_t idct_sse2_row_perm[8] = { 0, 4, 1, 5, 2, 6, 3, 7 };
+
+static void idct_mmx_init(void)
{
int i;
/* the mmx/mmxext idct uses a reordered input, so we patch scan tables */
for (i = 0; i < 64; i++) {
idct_mmx_perm[i] = (i & 0x38) | ((i & 6) >> 1) | ((i & 1) << 2);
-// idct_simple_mmx_perm[i] = simple_block_permute_op(i);
}
}
-static DCTELEM block[64] __attribute__ ((aligned (8)));
-static DCTELEM block1[64] __attribute__ ((aligned (8)));
-static DCTELEM block_org[64] __attribute__ ((aligned (8)));
+DECLARE_ALIGNED(16, static int16_t, block)[64];
+DECLARE_ALIGNED(8, static int16_t, block1)[64];
-void dct_error(const char *name, int is_idct,
- void (*fdct_func)(DCTELEM *block),
- void (*fdct_ref)(DCTELEM *block), int test)
+static void init_block(int16_t block[64], int test, int is_idct, AVLFG *prng)
{
- int it, i, scale;
- int err_inf, v;
- int64_t err2, ti, ti1, it1;
- int64_t sysErr[64], sysErrMax=0;
- int maxout=0;
- int blockSumErrMax=0, blockSumErr;
+ int i, j;
- srandom(0);
+ memset(block, 0, 64 * sizeof(*block));
- err_inf = 0;
- err2 = 0;
- for(i=0; i<64; i++) sysErr[i]=0;
- for(it=0;it<NB_ITS;it++) {
- for(i=0;i<64;i++)
- block1[i] = 0;
- switch(test){
- case 0:
- for(i=0;i<64;i++)
- block1[i] = (random() % 512) -256;
- if (is_idct){
- fdct(block1);
-
- for(i=0;i<64;i++)
- block1[i]>>=3;
- }
+ switch (test) {
+ case 0:
+ for (i = 0; i < 64; i++)
+ block[i] = (av_lfg_get(prng) % 512) - 256;
+ if (is_idct) {
+ ff_ref_fdct(block);
+ for (i = 0; i < 64; i++)
+ block[i] >>= 3;
+ }
break;
- case 1:{
- int num= (random()%10)+1;
- for(i=0;i<num;i++)
- block1[random()%64] = (random() % 512) -256;
- }break;
- case 2:
- block1[0]= (random()%4096)-2048;
- block1[63]= (block1[0]&1)^1;
+ 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;
break;
- }
-
-#if 0 // simulate mismatch control
-{ int sum=0;
- for(i=0;i<64;i++)
- sum+=block1[i];
-
- if((sum&1)==0) block1[63]^=1;
+ case 2:
+ block[ 0] = av_lfg_get(prng) % 4096 - 2048;
+ block[63] = (block[0] & 1) ^ 1;
+ break;
+ }
}
-#endif
- for(i=0; i<64; i++)
- block_org[i]= block1[i];
-
- if (fdct_func == ff_mmx_idct ||
- fdct_func == j_rev_dct || fdct_func == ff_mmxext_idct) {
- for(i=0;i<64;i++)
- block[idct_mmx_perm[i]] = block1[i];
- } else if(fdct_func == ff_simple_idct_mmx ) {
- for(i=0;i<64;i++)
- block[idct_simple_mmx_perm[i]] = block1[i];
-
- } else {
- for(i=0; i<64; i++)
- block[i]= block1[i];
- }
-#if 0 // simulate mismatch control for tested IDCT but not the ref
-{ int sum=0;
- for(i=0;i<64;i++)
- sum+=block[i];
+static void permute(int16_t dst[64], const int16_t src[64], int perm)
+{
+ int i;
- if((sum&1)==0) block[63]^=1;
+ if (perm == MMX_PERM) {
+ for (i = 0; i < 64; i++)
+ dst[idct_mmx_perm[i]] = src[i];
+ } else if (perm == MMX_SIMPLE_PERM) {
+ for (i = 0; i < 64; i++)
+ dst[idct_simple_mmx_perm[i]] = src[i];
+ } else if (perm == SSE2_PERM) {
+ for (i = 0; i < 64; i++)
+ dst[(i & 0x38) | idct_sse2_row_perm[i & 7]] = src[i];
+ } else if (perm == PARTTRANS_PERM) {
+ for (i = 0; i < 64; i++)
+ dst[(i & 0x24) | ((i & 3) << 3) | ((i >> 3) & 3)] = src[i];
+ } else {
+ for (i = 0; i < 64; i++)
+ dst[i] = src[i];
+ }
}
-#endif
- fdct_func(block);
- emms(); /* for ff_mmx_idct */
+static int dct_error(const struct algo *dct, int test, int is_idct, int speed)
+{
+ void (*ref)(int16_t *block) = is_idct ? ff_ref_idct : ff_ref_fdct;
+ int it, i, scale;
+ int err_inf, v;
+ int64_t err2, ti, ti1, it1, err_sum = 0;
+ int64_t sysErr[64], sysErrMax = 0;
+ int maxout = 0;
+ int blockSumErrMax = 0, blockSumErr;
+ AVLFG prng;
+ double omse, ome;
+ int spec_err;
+
+ av_lfg_init(&prng, 1);
- if (fdct_func == fdct_ifast
-#ifndef FAAN_POSTSCALE
- || fdct_func == ff_faandct
-#endif
- ) {
- for(i=0; i<64; i++) {
- scale = 8*(1 << (AANSCALE_BITS + 11)) / aanscales[i];
- block[i] = (block[i] * scale /*+ (1<<(AANSCALE_BITS-1))*/) >> AANSCALE_BITS;
+ err_inf = 0;
+ err2 = 0;
+ for (i = 0; i < 64; i++)
+ sysErr[i] = 0;
+ for (it = 0; it < NB_ITS; it++) {
+ init_block(block1, test, is_idct, &prng);
+ permute(block, block1, dct->format);
+
+ dct->func(block);
+ emms_c();
+
+ if (dct->format == SCALE_PERM) {
+ for (i = 0; i < 64; i++) {
+ scale = 8 * (1 << (AANSCALE_BITS + 11)) / ff_aanscales[i];
+ block[i] = (block[i] * scale) >> AANSCALE_BITS;
}
}
- fdct_ref(block1);
+ ref(block1);
- blockSumErr=0;
- for(i=0;i<64;i++) {
- v = abs(block[i] - block1[i]);
+ blockSumErr = 0;
+ for (i = 0; i < 64; i++) {
+ int err = block[i] - block1[i];
+ err_sum += err;
+ v = abs(err);
if (v > err_inf)
err_inf = v;
err2 += v * v;
sysErr[i] += block[i] - block1[i];
blockSumErr += v;
- if( abs(block[i])>maxout) maxout=abs(block[i]);
+ if (abs(block[i]) > maxout)
+ maxout = abs(block[i]);
}
- if(blockSumErrMax < blockSumErr) blockSumErrMax= blockSumErr;
-#if 0 // print different matrix pairs
- if(blockSumErr){
- printf("\n");
- for(i=0; i<64; i++){
- if((i&7)==0) printf("\n");
- printf("%4d ", block_org[i]);
- }
- for(i=0; i<64; i++){
- if((i&7)==0) printf("\n");
- printf("%4d ", block[i] - block1[i]);
- }
- }
-#endif
+ if (blockSumErrMax < blockSumErr)
+ blockSumErrMax = blockSumErr;
}
- for(i=0; i<64; i++) sysErrMax= MAX(sysErrMax, FFABS(sysErr[i]));
+ for (i = 0; i < 64; i++)
+ sysErrMax = FFMAX(sysErrMax, FFABS(sysErr[i]));
-#if 1 // dump systematic errors
- for(i=0; i<64; i++){
- if(i%8==0) printf("\n");
- printf("%5d ", (int)sysErr[i]);
+ for (i = 0; i < 64; i++) {
+ if (i % 8 == 0)
+ printf("\n");
+ printf("%7d ", (int) sysErr[i]);
}
printf("\n");
-#endif
- printf("%s %s: err_inf=%d err2=%0.8f syserr=%0.8f maxout=%d blockSumErr=%d\n",
- is_idct ? "IDCT" : "DCT",
- name, err_inf, (double)err2 / NB_ITS / 64.0, (double)sysErrMax / NB_ITS, maxout, blockSumErrMax);
-#if 1 //Speed test
- /* speed test */
- for(i=0;i<64;i++)
- block1[i] = 0;
- switch(test){
- case 0:
- for(i=0;i<64;i++)
- block1[i] = (random() % 512) -256;
- if (is_idct){
- fdct(block1);
+ omse = (double) err2 / NB_ITS / 64;
+ ome = (double) err_sum / NB_ITS / 64;
- for(i=0;i<64;i++)
- block1[i]>>=3;
- }
- break;
- case 1:{
- case 2:
- block1[0] = (random() % 512) -256;
- block1[1] = (random() % 512) -256;
- block1[2] = (random() % 512) -256;
- block1[3] = (random() % 512) -256;
- }break;
- }
+ spec_err = is_idct && (err_inf > 1 || omse > 0.02 || fabs(ome) > 0.0015);
- if (fdct_func == ff_mmx_idct ||
- fdct_func == j_rev_dct || fdct_func == ff_mmxext_idct) {
- for(i=0;i<64;i++)
- block[idct_mmx_perm[i]] = block1[i];
- } else if(fdct_func == ff_simple_idct_mmx ) {
- for(i=0;i<64;i++)
- block[idct_simple_mmx_perm[i]] = block1[i];
- } else {
- for(i=0; i<64; i++)
- block[i]= block1[i];
- }
+ printf("%s %s: ppe=%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);
+
+ if (spec_err && !dct->nonspec)
+ return 1;
+
+ if (!speed)
+ return 0;
+
+ /* speed test */
+ init_block(block, test, is_idct, &prng);
+ permute(block1, block, dct->format);
- ti = gettime();
+ ti = av_gettime();
it1 = 0;
do {
- for(it=0;it<NB_ITS_SPEED;it++) {
- for(i=0; i<64; i++)
- block[i]= block1[i];
-// memcpy(block, block1, sizeof(DCTELEM) * 64);
-// dont memcpy especially not fastmemcpy because it does movntq !!!
- fdct_func(block);
+ for (it = 0; it < NB_ITS_SPEED; it++) {
+ memcpy(block, block1, sizeof(block));
+ dct->func(block);
}
it1 += NB_ITS_SPEED;
- ti1 = gettime() - ti;
+ ti1 = av_gettime() - ti;
} while (ti1 < 1000000);
- emms();
+ emms_c();
- printf("%s %s: %0.1f kdct/s\n",
- is_idct ? "IDCT" : "DCT",
- name, (double)it1 * 1000.0 / (double)ti1);
-#endif
+ printf("%s %s: %0.1f kdct/s\n", is_idct ? "IDCT" : "DCT", dct->name,
+ (double) it1 * 1000.0 / (double) ti1);
+
+ return 0;
}
-static uint8_t img_dest[64] __attribute__ ((aligned (8)));
-static uint8_t img_dest1[64] __attribute__ ((aligned (8)));
+DECLARE_ALIGNED(8, static uint8_t, img_dest)[64];
+DECLARE_ALIGNED(8, static uint8_t, img_dest1)[64];
-void idct248_ref(uint8_t *dest, int linesize, int16_t *block)
+static void idct248_ref(uint8_t *dest, int linesize, int16_t *block)
{
static int init;
static double c8[8][8];
if (!init) {
init = 1;
- for(i=0;i<8;i++) {
+ for (i = 0; i < 8; i++) {
sum = 0;
- for(j=0;j<8;j++) {
- s = (i==0) ? sqrt(1.0/8.0) : sqrt(1.0/4.0);
+ for (j = 0; j < 8; j++) {
+ s = (i == 0) ? sqrt(1.0 / 8.0) : sqrt(1.0 / 4.0);
c8[i][j] = s * cos(M_PI * i * (j + 0.5) / 8.0);
sum += c8[i][j] * c8[i][j];
}
}
- for(i=0;i<4;i++) {
+ for (i = 0; i < 4; i++) {
sum = 0;
- for(j=0;j<4;j++) {
- s = (i==0) ? sqrt(1.0/4.0) : sqrt(1.0/2.0);
+ for (j = 0; j < 4; j++) {
+ s = (i == 0) ? sqrt(1.0 / 4.0) : sqrt(1.0 / 2.0);
c4[i][j] = s * cos(M_PI * i * (j + 0.5) / 4.0);
sum += c4[i][j] * c4[i][j];
}
/* butterfly */
s = 0.5 * sqrt(2.0);
- for(i=0;i<4;i++) {
- for(j=0;j<8;j++) {
- block1[8*(2*i)+j] = (block[8*(2*i)+j] + block[8*(2*i+1)+j]) * s;
- block1[8*(2*i+1)+j] = (block[8*(2*i)+j] - block[8*(2*i+1)+j]) * s;
+ for (i = 0; i < 4; i++) {
+ for (j = 0; j < 8; j++) {
+ block1[8 * (2 * i) + j] =
+ (block[8 * (2 * i) + j] + block[8 * (2 * i + 1) + j]) * s;
+ block1[8 * (2 * i + 1) + j] =
+ (block[8 * (2 * i) + j] - block[8 * (2 * i + 1) + j]) * s;
}
}
/* idct8 on lines */
- for(i=0;i<8;i++) {
- for(j=0;j<8;j++) {
+ for (i = 0; i < 8; i++) {
+ for (j = 0; j < 8; j++) {
sum = 0;
- for(k=0;k<8;k++)
- sum += c8[k][j] * block1[8*i+k];
- block2[8*i+j] = sum;
+ for (k = 0; k < 8; k++)
+ sum += c8[k][j] * block1[8 * i + k];
+ block2[8 * i + j] = sum;
}
}
/* idct4 */
- for(i=0;i<8;i++) {
- for(j=0;j<4;j++) {
+ for (i = 0; i < 8; i++) {
+ for (j = 0; j < 4; j++) {
/* top */
sum = 0;
- for(k=0;k<4;k++)
- sum += c4[k][j] * block2[8*(2*k)+i];
- block3[8*(2*j)+i] = sum;
+ for (k = 0; k < 4; k++)
+ sum += c4[k][j] * block2[8 * (2 * k) + i];
+ block3[8 * (2 * j) + i] = sum;
/* bottom */
sum = 0;
- for(k=0;k<4;k++)
- sum += c4[k][j] * block2[8*(2*k+1)+i];
- block3[8*(2*j+1)+i] = sum;
+ for (k = 0; k < 4; k++)
+ sum += c4[k][j] * block2[8 * (2 * k + 1) + i];
+ block3[8 * (2 * j + 1) + i] = sum;
}
}
/* clamp and store the result */
- for(i=0;i<8;i++) {
- for(j=0;j<8;j++) {
- v = block3[8*i+j];
- if (v < 0)
- v = 0;
- else if (v > 255)
- v = 255;
- dest[i * linesize + j] = (int)rint(v);
+ for (i = 0; i < 8; i++) {
+ for (j = 0; j < 8; j++) {
+ v = block3[8 * i + j];
+ if (v < 0) v = 0;
+ else if (v > 255) v = 255;
+ dest[i * linesize + j] = (int) rint(v);
}
}
}
-void idct248_error(const char *name,
- void (*idct248_put)(uint8_t *dest, int line_size, int16_t *block))
+static void idct248_error(const char *name,
+ void (*idct248_put)(uint8_t *dest, int line_size,
+ int16_t *block),
+ int speed)
{
int it, i, it1, ti, ti1, err_max, v;
+ AVLFG prng;
- srandom(0);
+ av_lfg_init(&prng, 1);
/* just one test to see if code is correct (precision is less
important here) */
err_max = 0;
- for(it=0;it<NB_ITS;it++) {
-
+ for (it = 0; it < NB_ITS; it++) {
/* XXX: use forward transform to generate values */
- for(i=0;i<64;i++)
- block1[i] = (random() % 256) - 128;
+ for (i = 0; i < 64; i++)
+ block1[i] = av_lfg_get(&prng) % 256 - 128;
block1[0] += 1024;
- for(i=0; i<64; i++)
- block[i]= block1[i];
+ for (i = 0; i < 64; i++)
+ block[i] = block1[i];
idct248_ref(img_dest1, 8, block);
- for(i=0; i<64; i++)
- block[i]= block1[i];
+ for (i = 0; i < 64; i++)
+ block[i] = block1[i];
idct248_put(img_dest, 8, block);
- for(i=0;i<64;i++) {
- v = abs((int)img_dest[i] - (int)img_dest1[i]);
+ for (i = 0; i < 64; i++) {
+ v = abs((int) img_dest[i] - (int) img_dest1[i]);
if (v == 255)
printf("%d %d\n", img_dest[i], img_dest1[i]);
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);
+ printf("%s %s: err_inf=%d\n", 1 ? "IDCT248" : "DCT248", name, err_max);
- ti = gettime();
+ if (!speed)
+ return;
+
+ ti = av_gettime();
it1 = 0;
do {
- for(it=0;it<NB_ITS_SPEED;it++) {
- for(i=0; i<64; i++)
- block[i]= block1[i];
-// memcpy(block, block1, sizeof(DCTELEM) * 64);
-// dont memcpy especially not fastmemcpy because it does movntq !!!
+ for (it = 0; it < NB_ITS_SPEED; it++) {
+ for (i = 0; i < 64; i++)
+ block[i] = block1[i];
idct248_put(img_dest, 8, block);
}
it1 += NB_ITS_SPEED;
- ti1 = gettime() - ti;
+ ti1 = av_gettime() - ti;
} while (ti1 < 1000000);
- emms();
+ emms_c();
- printf("%s %s: %0.1f kdct/s\n",
- 1 ? "IDCT248" : "DCT248",
- name, (double)it1 * 1000.0 / (double)ti1);
+ printf("%s %s: %0.1f kdct/s\n", 1 ? "IDCT248" : "DCT248", name,
+ (double) it1 * 1000.0 / (double) ti1);
}
-void help(void)
+static void help(void)
{
printf("dct-test [-i] [<test-number>]\n"
"test-number 0 -> test with random matrixes\n"
" 1 -> test with random sparse matrixes\n"
" 2 -> do 3. test from mpeg4 std\n"
"-i test IDCT implementations\n"
- "-4 test IDCT248 implementations\n");
- exit(1);
+ "-4 test IDCT248 implementations\n"
+ "-t speed test\n");
}
+#if !HAVE_GETOPT
+#include "compat/getopt.c"
+#endif
+
int main(int argc, char **argv)
{
int test_idct = 0, test_248_dct = 0;
- int c,i;
- int test=1;
+ int c, i;
+ int test = 1;
+ int speed = 0;
+ int err = 0;
- init_fdct();
- idct_mmx_init();
+ cpu_flags = av_get_cpu_flags();
- 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;
- }
+ ff_ref_dct_init();
+ idct_mmx_init();
- for(;;) {
- c = getopt(argc, argv, "ih4");
+ for (;;) {
+ c = getopt(argc, argv, "ih4t");
if (c == -1)
break;
- switch(c) {
+ switch (c) {
case 'i':
test_idct = 1;
break;
case '4':
test_248_dct = 1;
break;
- default :
+ case 't':
+ speed = 1;
+ break;
+ default:
case 'h':
help();
- break;
+ return 0;
}
}
- if(optind <argc) test= atoi(argv[optind]);
+ if (optind < argc)
+ test = atoi(argv[optind]);
- printf("ffmpeg DCT/IDCT test\n");
+ printf("Libav DCT/IDCT test\n");
if (test_248_dct) {
- idct248_error("SIMPLE-C", simple_idct248_put);
+ idct248_error("SIMPLE-C", ff_simple_idct248_put, speed);
} else {
- if (!test_idct) {
- dct_error("REF-DBL", 0, fdct, fdct, test); /* only to verify code ! */
- dct_error("IJG-AAN-INT", 0, fdct_ifast, fdct, test);
- dct_error("IJG-LLM-INT", 0, ff_jpeg_fdct_islow, fdct, test);
- dct_error("MMX", 0, ff_fdct_mmx, fdct, test);
- dct_error("MMX2", 0, ff_fdct_mmx2, fdct, test);
- dct_error("FAAN", 0, ff_faandct, fdct, test);
- } else {
- dct_error("REF-DBL", 1, idct, idct, test);
- dct_error("INT", 1, j_rev_dct, idct, test);
- dct_error("LIBMPEG2-MMX", 1, ff_mmx_idct, idct, test);
- dct_error("LIBMPEG2-MMXEXT", 1, ff_mmxext_idct, idct, test);
- dct_error("SIMPLE-C", 1, simple_idct, idct, test);
- dct_error("SIMPLE-MMX", 1, ff_simple_idct_mmx, idct, test);
- dct_error("XVID-MMX", 1, ff_idct_xvid_mmx, idct, test);
- dct_error("XVID-MMX2", 1, ff_idct_xvid_mmx2, idct, test);
- // dct_error("ODIVX-C", 1, odivx_idct_c, idct);
- //printf(" test against odivx idct\n");
- // dct_error("REF", 1, idct, odivx_idct_c);
- // dct_error("INT", 1, j_rev_dct, odivx_idct_c);
- // dct_error("MMX", 1, ff_mmx_idct, odivx_idct_c);
- // dct_error("MMXEXT", 1, ff_mmxext_idct, odivx_idct_c);
- // dct_error("SIMPLE-C", 1, simple_idct, odivx_idct_c);
- // dct_error("SIMPLE-MMX", 1, ff_simple_idct_mmx, odivx_idct_c);
- // dct_error("ODIVX-C", 1, odivx_idct_c, odivx_idct_c);
- }
+ 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);
+ }
}
- return 0;
+
+ return err;
}
projects / ffmpeg / blobdiff