xvid_bench.c

视频压缩编解码标准MPEG4商业级别的VC代码实现标准

/*****************************************************************************
 *
 *  XVID MPEG-4 VIDEO CODEC
 *  - Unit tests and benches -
 *
 *  Copyright(C) 2002 Pascal Massimino <[e-mail]>
 *
 *  This program is free software; you can redistribute it and/or modify
 *  it under the terms of the GNU General Public License as published by
 *  the Free Software Foundation; either version 2 of the License, or
 *  (at your option) any later version.
 *
 *  This program 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 General Public License for more details.
 *
 *  You should have received a copy of the GNU General Public License
 *  along with this program; if not, write to the Free Software
 *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
 *
 * $Id$
 *
 ****************************************************************************/

/*****************************************************************************
 *                            
 *  'Reference' output is at the end of file.
 *
 *   compiles with something like:
 *   gcc -o xvid_bench xvid_bench.c  -I../src/ -lxvidcore -lm
 *
 ****************************************************************************/

#include <stdio.h>
#include <stdlib.h>
#include <string.h>    /* for memset */
#include <assert.h>

#ifndef WIN32
#include <sys/time.h>	/* for gettimeofday */
#else
#include <time.h>
#endif


#include "xvid.h"

// inner guts
#include "dct/idct.h"
#include "dct/fdct.h"
#include "image/colorspace.h"
#include "image/interpolate8x8.h"
#include "utils/mem_transfer.h"
#include "quant/quant.h"
#include "motion/sad.h"
#include "utils/emms.h"
#include "utils/timer.h"
#include "quant/quant_matrix.c"
#include "bitstream/cbp.h"

#include <math.h>

#ifndef M_PI
#define M_PI		3.14159265358979323846
#endif

const int speed_ref = 100;  /* on slow machines, decrease this value */

/*********************************************************************
 * misc
 *********************************************************************/

/* returns time in micro-s*/
double gettime_usec()
{    
#ifndef WIN32
	struct timeval  tv;
	gettimeofday(&tv, 0);
	return tv.tv_sec*1.0e6 + tv.tv_usec;
#else
	clock_t clk;
	clk = clock();
	return clk * 1000000 / CLOCKS_PER_SEC;
#endif
}

/* returns squared deviates (mean(v*v)-mean(v)^2) of a 8x8 block */
double sqr_dev(uint8_t v[8*8])
{
	double sum=0.;
	double sum2=0.;
	int n;
	for (n=0;n<8*8;n++)
	{
		sum  += v[n];
		sum2 += v[n]*v[n];
	}
	sum2 /= n;
	sum /= n;
	return sum2-sum*sum;
}

/*********************************************************************
 * cpu init
 *********************************************************************/

typedef struct {
	const char *name;
	unsigned int cpu;
} CPU;

CPU cpu_list[] = 
{ { "PLAINC", 0 }
#ifdef ARCH_IS_IA32
  , { "MMX   ", XVID_CPU_MMX }
  , { "MMXEXT", XVID_CPU_MMXEXT | XVID_CPU_MMX }
  , { "SSE2  ", XVID_CPU_SSE2 | XVID_CPU_MMX }
  , { "3DNOW ", XVID_CPU_3DNOW }
  , { "3DNOWE", XVID_CPU_3DNOW | XVID_CPU_3DNOWEXT }
#endif
#ifdef ARCH_IS_PPC
  , { "ALTIVEC", XVID_CPU_ALTIVEC }
#endif
#ifdef ARCH_IS_X86_64
  , { "X86_64", XVID_CPU_ASM}
#endif
//, { "IA64  ", XVID_CPU_IA64 }  
//, { "TSC   ", XVID_CPU_TSC }
  , { 0, 0 } };

CPU  cpu_short_list[] =
{ { "PLAINC", 0 }
#ifdef ARCH_IS_IA32
  , { "MMX   ", XVID_CPU_MMX }
//, { "MMXEXT", XVID_CPU_MMXEXT | XVID_CPU_MMX }
#endif
//, { "IA64  ", XVID_CPU_IA64 }
  , { 0, 0 } };

CPU cpu_short_list2[] = 
{ { "PLAINC", 0 }
#ifdef ARCH_IS_IA32
  , { "MMX   ", XVID_CPU_MMX }
  , { "SSE2  ", XVID_CPU_SSE2 | XVID_CPU_MMX }
#endif
  , { 0, 0 } };


int init_cpu(CPU *cpu)
{
	xvid_gbl_info_t xinfo;

	/* Get the available CPU flags */
	memset(&xinfo, 0, sizeof(xinfo));
	xinfo.version = XVID_VERSION;
	xvid_global(NULL, XVID_GBL_INFO, &xinfo, NULL);

	/* Are we trying to test a subset of the host CPU features */
	if ((xinfo.cpu_flags & cpu->cpu) == cpu->cpu) {
		int xerr;
		xvid_gbl_init_t xinit;
		memset(&xinit, 0, sizeof(xinit));
		xinit.cpu_flags = cpu->cpu | XVID_CPU_FORCE;
		xinit.version = XVID_VERSION;
		xerr = xvid_global(NULL, XVID_GBL_INIT, &xinit, NULL);
		if (xerr==XVID_ERR_FAIL) {
			/* libxvidcore failed to init */
			return 0;
		}
	} else {
		/* The host CPU doesn't support some required feature for this test */
		return(0);
	}
	return 1;
}

#define CRC32_REMAINDER 0xCBF43926
#define CRC32_INITIAL 0xffffffff

#define DO1(c, crc) ((crc) = crc32tab[((unsigned int)((crc)>>24) ^ (*c++)) & 0xff] ^ ((crc) << 8))
#define DO2(c, crc)  DO1(c, crc); DO1(c, crc);
#define DO4(c, crc)  DO2(c, crc); DO2(c, crc);
#define DO8(c, crc)  DO4(c, crc); DO4(c, crc);

/******************************************************************************
* Precomputed AAL5 CRC32 lookup table
******************************************************************************/

static unsigned long crc32tab[256] = {

	0x00000000L, 0x04C11DB7L, 0x09823B6EL, 0x0D4326D9L,
	0x130476DCL, 0x17C56B6BL, 0x1A864DB2L, 0x1E475005L,
	0x2608EDB8L, 0x22C9F00FL, 0x2F8AD6D6L, 0x2B4BCB61L,
	0x350C9B64L, 0x31CD86D3L, 0x3C8EA00AL, 0x384FBDBDL,
	0x4C11DB70L, 0x48D0C6C7L, 0x4593E01EL, 0x4152FDA9L,
	0x5F15ADACL, 0x5BD4B01BL, 0x569796C2L, 0x52568B75L,
	0x6A1936C8L, 0x6ED82B7FL, 0x639B0DA6L, 0x675A1011L,
	0x791D4014L, 0x7DDC5DA3L, 0x709F7B7AL, 0x745E66CDL,
	0x9823B6E0L, 0x9CE2AB57L, 0x91A18D8EL, 0x95609039L,
	0x8B27C03CL, 0x8FE6DD8BL, 0x82A5FB52L, 0x8664E6E5L,
	0xBE2B5B58L, 0xBAEA46EFL, 0xB7A96036L, 0xB3687D81L,
	0xAD2F2D84L, 0xA9EE3033L, 0xA4AD16EAL, 0xA06C0B5DL,
	0xD4326D90L, 0xD0F37027L, 0xDDB056FEL, 0xD9714B49L,
	0xC7361B4CL, 0xC3F706FBL, 0xCEB42022L, 0xCA753D95L,
	0xF23A8028L, 0xF6FB9D9FL, 0xFBB8BB46L, 0xFF79A6F1L,
	0xE13EF6F4L, 0xE5FFEB43L, 0xE8BCCD9AL, 0xEC7DD02DL,
	0x34867077L, 0x30476DC0L, 0x3D044B19L, 0x39C556AEL,
	0x278206ABL, 0x23431B1CL, 0x2E003DC5L, 0x2AC12072L,
	0x128E9DCFL, 0x164F8078L, 0x1B0CA6A1L, 0x1FCDBB16L,
	0x018AEB13L, 0x054BF6A4L, 0x0808D07DL, 0x0CC9CDCAL,
	0x7897AB07L, 0x7C56B6B0L, 0x71159069L, 0x75D48DDEL,
	0x6B93DDDBL, 0x6F52C06CL, 0x6211E6B5L, 0x66D0FB02L,
	0x5E9F46BFL, 0x5A5E5B08L, 0x571D7DD1L, 0x53DC6066L,
	0x4D9B3063L, 0x495A2DD4L, 0x44190B0DL, 0x40D816BAL,
	0xACA5C697L, 0xA864DB20L, 0xA527FDF9L, 0xA1E6E04EL,
	0xBFA1B04BL, 0xBB60ADFCL, 0xB6238B25L, 0xB2E29692L,
	0x8AAD2B2FL, 0x8E6C3698L, 0x832F1041L, 0x87EE0DF6L,
	0x99A95DF3L, 0x9D684044L, 0x902B669DL, 0x94EA7B2AL,
	0xE0B41DE7L, 0xE4750050L, 0xE9362689L, 0xEDF73B3EL,
	0xF3B06B3BL, 0xF771768CL, 0xFA325055L, 0xFEF34DE2L,
	0xC6BCF05FL, 0xC27DEDE8L, 0xCF3ECB31L, 0xCBFFD686L,
	0xD5B88683L, 0xD1799B34L, 0xDC3ABDEDL, 0xD8FBA05AL,
	0x690CE0EEL, 0x6DCDFD59L, 0x608EDB80L, 0x644FC637L,
	0x7A089632L, 0x7EC98B85L, 0x738AAD5CL, 0x774BB0EBL,
	0x4F040D56L, 0x4BC510E1L, 0x46863638L, 0x42472B8FL,
	0x5C007B8AL, 0x58C1663DL, 0x558240E4L, 0x51435D53L,
	0x251D3B9EL, 0x21DC2629L, 0x2C9F00F0L, 0x285E1D47L,
	0x36194D42L, 0x32D850F5L, 0x3F9B762CL, 0x3B5A6B9BL,
	0x0315D626L, 0x07D4CB91L, 0x0A97ED48L, 0x0E56F0FFL,
	0x1011A0FAL, 0x14D0BD4DL, 0x19939B94L, 0x1D528623L,
	0xF12F560EL, 0xF5EE4BB9L, 0xF8AD6D60L, 0xFC6C70D7L,
	0xE22B20D2L, 0xE6EA3D65L, 0xEBA91BBCL, 0xEF68060BL,
	0xD727BBB6L, 0xD3E6A601L, 0xDEA580D8L, 0xDA649D6FL,
	0xC423CD6AL, 0xC0E2D0DDL, 0xCDA1F604L, 0xC960EBB3L,
	0xBD3E8D7EL, 0xB9FF90C9L, 0xB4BCB610L, 0xB07DABA7L,
	0xAE3AFBA2L, 0xAAFBE615L, 0xA7B8C0CCL, 0xA379DD7BL,
	0x9B3660C6L, 0x9FF77D71L, 0x92B45BA8L, 0x9675461FL,
	0x8832161AL, 0x8CF30BADL, 0x81B02D74L, 0x857130C3L,
	0x5D8A9099L, 0x594B8D2EL, 0x5408ABF7L, 0x50C9B640L,
	0x4E8EE645L, 0x4A4FFBF2L, 0x470CDD2BL, 0x43CDC09CL,
	0x7B827D21L, 0x7F436096L, 0x7200464FL, 0x76C15BF8L,
	0x68860BFDL, 0x6C47164AL, 0x61043093L, 0x65C52D24L,
	0x119B4BE9L, 0x155A565EL, 0x18197087L, 0x1CD86D30L,
	0x029F3D35L, 0x065E2082L, 0x0B1D065BL, 0x0FDC1BECL,
	0x3793A651L, 0x3352BBE6L, 0x3E119D3FL, 0x3AD08088L,
	0x2497D08DL, 0x2056CD3AL, 0x2D15EBE3L, 0x29D4F654L,
	0xC5A92679L, 0xC1683BCEL, 0xCC2B1D17L, 0xC8EA00A0L,
	0xD6AD50A5L, 0xD26C4D12L, 0xDF2F6BCBL, 0xDBEE767CL,
	0xE3A1CBC1L, 0xE760D676L, 0xEA23F0AFL, 0xEEE2ED18L,
	0xF0A5BD1DL, 0xF464A0AAL, 0xF9278673L, 0xFDE69BC4L,
	0x89B8FD09L, 0x8D79E0BEL, 0x803AC667L, 0x84FBDBD0L,
	0x9ABC8BD5L, 0x9E7D9662L, 0x933EB0BBL, 0x97FFAD0CL,
	0xAFB010B1L, 0xAB710D06L, 0xA6322BDFL, 0xA2F33668L,
	0xBCB4666DL, 0xB8757BDAL, 0xB5365D03L, 0xB1F740B4L

};

uint32_t
calc_crc(uint8_t *mem, int len, uint32_t initial)
{

	register unsigned int crc;

	crc = initial;

	while( len >= 8) {
		DO8(mem, crc);
		len -= 8;
	}

	while( len ) {
		DO1(mem, crc);
		len--;
	}

	return(crc);

}

/*********************************************************************
 * test DCT
 *********************************************************************/

#define ABS(X)  ((X)<0 ? -(X) : (X))

void test_dct()
{
	const int nb_tests = 300*speed_ref;
	int tst;
	CPU *cpu;
	int i;
	DECLARE_ALIGNED_MATRIX(iDst0, 8, 8, short, 16);
	DECLARE_ALIGNED_MATRIX(iDst,  8, 8, short, 16);
	DECLARE_ALIGNED_MATRIX(fDst,  8, 8, short, 16);
	double overhead;

	printf( "\n ===== test fdct/idct =====\n" );

	for(i=0; i<8*8; ++i) iDst0[i] = (i*7-i*i) & 0x7f;
	overhead = gettime_usec();
	for(tst=0; tst<nb_tests; ++tst)
	{
		for(i=0; i<8*8; ++i) fDst[i] = iDst0[i];
		for(i=0; i<8*8; ++i) iDst[i] = fDst[i];
	}
	overhead = gettime_usec() - overhead;

	for(cpu = cpu_list; cpu->name!=0; ++cpu)
	{
		double t, PSNR, MSE;

		if (!init_cpu(cpu))
			continue;

		t = gettime_usec();
		emms();
		for(tst=0; tst<nb_tests; ++tst)
		{
			for(i=0; i<8*8; ++i) fDst[i] = iDst0[i];
			fdct(fDst);
			for(i=0; i<8*8; ++i) iDst[i] = fDst[i];
			idct(iDst);
		}
		emms();
		t = (gettime_usec() - t - overhead) / nb_tests;
		MSE = 0.;
		for(i=0; i<8*8; ++i) {
			double delta = 1.0*(iDst[i] - iDst0[i]);
			MSE += delta*delta;
		}
		PSNR = (MSE==0.) ? 1.e6 : -4.3429448*log( MSE/64. );
		printf( "%s -  %.3f usec       PSNR=%.3f  MSE=%.3f %s\n",
				cpu->name, t, PSNR, MSE,
				(ABS(MSE)>=64)? "| ERROR" :"");
	}
}

/*********************************************************************
 * test SAD
 *********************************************************************/

void test_sad()
{
	const int nb_tests = 2000*speed_ref;
	int tst;
	CPU *cpu;
	int i;
	DECLARE_ALIGNED_MATRIX(Cur,  16, 16, uint8_t, 16);
	DECLARE_ALIGNED_MATRIX(Ref1, 16, 16, uint8_t, 16);
	DECLARE_ALIGNED_MATRIX(Ref2, 16, 16, uint8_t, 16);

	printf( "\n ======  test SAD ======\n" );
	for(i=0; i<16*16;++i) {
		Cur[i] = (i/5) ^ 0x05;
		Ref1[i] = (i + 0x0b) & 0xff;
		Ref2[i] = i ^ 0x76;
	}

	for(cpu = cpu_list; cpu->name!=0; ++cpu)
	{
		double t;
		uint32_t s;
		if (!init_cpu(cpu))
			continue;

		t = gettime_usec();
		emms();
		for(tst=0; tst<nb_tests; ++tst) s = sad8(Cur, Ref1, 16);
		emms();
		t = (gettime_usec() - t) / nb_tests;
		printf("%s - sad8    %.3f usec       sad=%d %s\n",
			   cpu->name, t, s,
			   (s!=3776)?"| ERROR": "" );

		t = gettime_usec();
		emms();
		for(tst=0; tst<nb_tests; ++tst) s = sad16(Cur, Ref1, 16, -1);
		emms();
		t = (gettime_usec() - t) / nb_tests;
		printf("%s - sad16   %.3f usec       sad=%d %s\n",
			   cpu->name, t, s,
			   (s!=27214)?"| ERROR": "" );

		t = gettime_usec();
		emms();
		for(tst=0; tst<nb_tests; ++tst) s = sad16bi(Cur, Ref1, Ref2, 16);
		emms();
		t = (gettime_usec() - t) / nb_tests;
		printf( "%s - sad16bi %.3f usec       sad=%d %s\n",
				cpu->name, t, s,
				(s!=26274)?"| ERROR": "" );

		t = gettime_usec();
		emms();
		for(tst=0; tst<nb_tests; ++tst) s = dev16(Cur, 16);
		emms();
		t = (gettime_usec() - t) / nb_tests;
		printf( "%s - dev16   %.3f usec       sad=%d %s\n",
				cpu->name, t, s,
				(s!=3344)?"| ERROR": "" );

		printf( " --- \n" );
	}
}

/*********************************************************************
 * test interpolation
 *********************************************************************/

#define ENTER \
for(i=0; i<16*8; ++i) Dst[i] = 0;   \
t = gettime_usec();                   \
emms();

#define LEAVE \
emms();                             \
t = (gettime_usec() - t) / nb_tests;  \
	iCrc = calc_crc((uint8_t*)Dst, sizeof(Dst), CRC32_INITIAL)

#define TEST_MB(FUNC, R)                \
ENTER                               \
for(tst=0; tst<nb_tests; ++tst) (FUNC)(Dst, Src0, 16, (R)); \
LEAVE

#define TEST_MB2(FUNC)                  \
ENTER                               \
for(tst=0; tst<nb_tests; ++tst) (FUNC)(Dst, Src0, 16); \
LEAVE


void test_mb()
{
	const int nb_tests = 2000*speed_ref;
	CPU *cpu;
	const uint8_t Src0[16*9] = {
		/* try to have every possible combinaison of rounding... */
		0, 0, 1, 0, 2, 0, 3, 0, 4             ,0,0,0, 0,0,0,0,
		0, 1, 1, 1, 2, 1, 3, 1, 3             ,0,0,0, 0,0,0,0,
		0, 2, 1, 2, 2, 2, 3, 2, 2             ,0,0,0, 0,0,0,0,
		0, 3, 1, 3, 2, 3, 3, 3, 1             ,0,0,0, 0,0,0,0,
		1, 3, 0, 2, 1, 0, 2, 3, 4             ,0,0,0, 0,0,0,0,
		2, 2, 1, 2, 0, 1, 3, 5, 3             ,0,0,0, 0,0,0,0,
		3, 1, 2, 3, 1, 2, 2, 6, 2             ,0,0,0, 0,0,0,0,
		1, 0, 1, 3, 0, 3, 1, 6, 1             ,0,0,0, 0,0,0,0,
		4, 3, 2, 1, 2, 3, 4, 0, 3             ,0,0,0, 0,0,0,0
	};
	uint8_t Dst[16*8] = {0};

	printf( "\n ===  test block motion ===\n" );

	for(cpu = cpu_list; cpu->name!=0; ++cpu)
	{
		double t;
		int tst, i, iCrc;

		if (!init_cpu(cpu))
			continue;

		TEST_MB(interpolate8x8_halfpel_h, 0);
		printf("%s - interp- h-round0 %.3f usec       crc32=0x%08x %s\n",
			   cpu->name, t, iCrc,