--- xvid_bench.c	2006/10/11 14:55:28	1.29
+++ xvid_bench.c	2006/11/01 07:12:26	1.34
@@ -19,7 +19,7 @@
  *  along with this program; if not, write to the Free Software
  *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
  *
- * $Id: xvid_bench.c,v 1.29 2006/10/11 14:55:28 Skal Exp $
+ * $Id: xvid_bench.c,v 1.34 2006/11/01 07:12:26 Skal Exp $
  *
  ****************************************************************************/
 
@@ -58,6 +58,7 @@
 #include "utils/timer.h"
 #include "quant/quant_matrix.c"
 #include "bitstream/cbp.h"
+#include "bitstream/bitstream.h"
 
 #include <math.h>
 
@@ -1725,8 +1726,8 @@
 for(tst=0; tst<nb_tests; ++tst) (FUNC)(Dst0[0], S*WIDTH, Src0[0], Src0[1], Src0[2], WIDTH, WIDTH/2, WIDTH, HEIGHT, 0); \
 LEAVE
 
-#define WIDTH 64
-#define HEIGHT 64
+#define WIDTH 128
+#define HEIGHT 32
 void test_yuv()
 {
 	const int nb_tests = 200*speed_ref;
@@ -1740,6 +1741,7 @@
 	colorspace_init();
 	ieee_reseed(1);
 	for(i=0; i<(int)sizeof(Src0); ++i) Src0[0][i] = ieee_rand(0,255);
+	for(i=0; i<(int)sizeof(Dst0); ++i) Dst0[0][i] = 0x5a;
 
 	printf( "\n ===  test YUV ===\n" );
 
@@ -1774,6 +1776,70 @@
 #endif
 	printf( " --- \n" );
 }
+
+#define TEST_YV2(FUNC, WITH_UV, WITH_FLIP)        \
+ENTER                               \
+for(tst=0; tst<nb_tests; ++tst) (FUNC)(Dst0[0], Dst0[1], Dst0[2], WIDTH, WIDTH, \
+	Src0[0], (WITH_UV) ? Src0[1] : 0, (WITH_UV) ? Src0[2] : 0,  WIDTH, WIDTH, \
+	WIDTH-2, HEIGHT-2, WITH_FLIP); \
+LEAVE
+
+#define PRINT_NxN(DATA,W,H,STR)   {   \
+	int i,j; \
+	for(j=0; j<(H); ++j) { \
+		for(i=0; i<(W); ++i) printf( "0x%.2x ", (DATA)[i+j*(STR)] );\
+		printf("\n"); \
+	} \
+	printf("---\n"); \
+}
+
+static const int yv12_CRCs[2][2] = { 
+	{0x5cab7cf0,0xdab46541} 
+,       {0xe8bae865,0x1faf77b7}
+};
+
+void test_yuv2()
+{
+	const int nb_tests = 800*speed_ref;
+	CPU *cpu;
+	uint8_t Src0[3][WIDTH*HEIGHT];
+	uint8_t Dst0[3][WIDTH*HEIGHT];
+	int with_uv, with_flip;
+	int i, j;
+	double t;
+	int tst, iCrc;
+
+	colorspace_init();
+	ieee_reseed(1);
+	for(i=0; i<(int)sizeof(Src0); ++i) Src0[0][i] = ieee_rand(0,255);
+
+	printf( "\n ===  test YV2 ===\n" );
+        for(with_flip=0; with_flip<=1; ++with_flip) {
+        	for(with_uv=0; with_uv<=1; ++with_uv) {
+			init_cpu(&cpu_list[0]);
+			TEST_YV2(yv12_to_yv12_c, with_uv, with_flip);
+			printf(" yv12_to_yv12_c   %.3f usec      \tcrc32=0x%08x %s\n",
+				t, iCrc, (iCrc!=yv12_CRCs[with_flip][with_uv])?"| ERROR": "" );
+			/* if (!with_uv) PRINT_NxN(Dst0[1], WIDTH/2, HEIGHT/2, WIDTH ); */
+
+#ifdef ARCH_IS_IA32
+			init_cpu(&cpu_list[1]);
+			TEST_YV2(yv12_to_yv12_mmx, with_uv, with_flip);
+			printf(" yv12_to_yv12_mmx %.3f usec     \tcrc32=0x%08x %s\n",
+				t, iCrc, (iCrc!=yv12_CRCs[with_flip][with_uv])?"| ERROR": "" );
+			/* if (!with_uv) PRINT_NxN(Dst0[1], WIDTH/2, HEIGHT/2, WIDTH ); */
+
+			TEST_YV2(yv12_to_yv12_xmm, with_uv, with_flip);
+			printf(" yv12_to_yv12_xmm %.3f usec     \tcrc32=0x%08x %s\n",
+				t, iCrc, (iCrc!=yv12_CRCs[with_flip][with_uv])?"| ERROR": "" );
+#endif
+		}
+
+		printf( " --- \n" );
+	}
+	printf( " ===== \n" );
+}
+
 #undef WIDTH
 #undef HEIGHT
 #undef ENTER
@@ -1867,7 +1933,7 @@
 
 void test_gcd()
 {
-	const int nb_tests = 10*speed_ref;
+  const int nb_tests = 10*speed_ref;
   int i;
   uint32_t crc1=0, crc2=0;
   uint32_t n0, n, d0, d;
@@ -1944,6 +2010,7 @@
     printf( "ERROR! please post your platform/compiler specs to xvid-devel@xvid.org !\n" );
   }
 }
+
 /*********************************************************************
  * test SSIM functions
  *********************************************************************/
@@ -1954,7 +2021,7 @@
 extern int lum_8x8_c(uint8_t* ptr, int stride);
 extern int lum_8x8_mmx(uint8_t* ptr, int stride);
 extern int lum_2x8_c(uint8_t* ptr, int stride);
-extern void iconsim_c(uint8_t* ptro, uint8_t* ptrc, int stride, int lumo, int lumc, int* pdevo, int* pdevc, int* pcorr);
+extern void consim_c(uint8_t* ptro, uint8_t* ptrc, int stride, int lumo, int lumc, int* pdevo, int* pdevc, int* pcorr);
 extern void consim_mmx(uint8_t* ptro, uint8_t* ptrc, int stride, int lumo, int lumc, int* pdevo, int* pdevc, int* pcorr);
 extern void consim_sse2(uint8_t* ptro, uint8_t* ptrc, int stride, int lumo, int lumc, int* pdevo, int* pdevc, int* pcorr);
 
@@ -1992,7 +2059,7 @@
 			continue;
 		lum8x8 = lum_8x8_c;
 		lum2x8 = lum_2x8_c;
-		csim   = iconsim_c;
+		csim   = consim_c;
 		if (cpu->cpu & XVID_CPU_MMX){
 			lum8x8 = lum_8x8_mmx;
 			csim = consim_mmx;
@@ -2012,27 +2079,82 @@
 
 		t = gettime_usec();
 		emms();
-		for(tst=0; tst<nb_tests; ++tst) m = lum2x8(Ref1, 16);
+		for(tst=0; tst<nb_tests; ++tst) m = lum2x8(Ref1+8, 16);
 		emms();
 		t = (gettime_usec() - t) / nb_tests;
 		printf("%s - ssim-lum2x8    %.3f usec       m=%d %s\n",
 			   cpu->name, t, m,
-			   (m!=-841)?"| ERROR": "" );
+			   (m!=681)?"| ERROR": "" );
 
 		t = gettime_usec();
 		emms();
 		for(tst=0; tst<nb_tests; ++tst) csim(Ref1, Ref2, 16, lumo, lumc, devs+0, devs+1, devs+2);
 		emms();
 		t = (gettime_usec() - t) / nb_tests;
-		printf("%s - ssim-lum2x8    %.3f usec       devs=[0x%x 0x%x 0x%x] %s\n",
+		printf("%s - ssim-consim    %.3f usec       devs=[0x%x 0x%x 0x%x] %s\n",
 			   cpu->name, t, devs[0], devs[1], devs[2],
-			   (devs[0]!=0xeba80 || devs[1]!=0x1053e7 ||  devs[2]!=0x51215)?"| ERROR": "" );
-
+			   (devs[0]!=0x1bdf0f || devs[1]!=0x137258 ||  devs[2]!=0xcdb13)?"| ERROR": "" );
 		printf( " --- \n" );
 	}
 }
 
 /*********************************************************************
+ * test bitstream functions
+ *********************************************************************/
+
+#define BIT_BUF_SIZE 2000
+
+static void test_bits()
+{
+  const int nb_tests = 50*speed_ref;
+  int tst;
+  uint32_t Crc;
+  uint8_t Buf[BIT_BUF_SIZE];
+  uint32_t Extracted[BIT_BUF_SIZE*8]; /* worst case: bits read 1 by 1 */
+  int Lens[BIT_BUF_SIZE*8];     
+  double t1;
+
+
+  printf( "\n ===  test bitstream ===\n" );
+  ieee_reseed(1);
+  Crc = 0;
+
+  t1 = gettime_usec();
+  for(tst=0; tst<nb_tests; ++tst) {
+  	Bitstream bs;
+  	int m, m2, l, l2;
+
+	for(l=0; l<BIT_BUF_SIZE; ++l)
+		Buf[l] = (uint8_t)ieee_rand(0,255);
+
+  	l = BIT_BUF_SIZE - ieee_rand(1,BIT_BUF_SIZE/10);
+  	BitstreamInit(&bs, (void*)(Buf+BIT_BUF_SIZE-l), l);
+
+
+	BitstreamReset(&bs);
+	for(l2=l*8, m=0; l2>0; m++) {
+		const int b = ieee_rand(1,32);
+		Lens[m] = b;
+		l2 -= b;
+		if (l2<0) break;
+		Extracted[m] = BitstreamShowBits(&bs, b);
+		BitstreamSkip(&bs, b);
+//		printf( "<= %d: %d 0x%x\n", m, b, Extracted[m]);
+	}
+
+	BitstreamReset(&bs);
+	for(m2=0; m2<m; ++m2) {
+		const int b = Lens[m2];
+		const uint32_t v = BitstreamGetBits(&bs, b);
+		Crc |= (v!=Extracted[m2]);
+//		printf( "=> %d: %d 0x%x %c\n", m2, b, v, " *"[Crc]);
+	}
+  }
+  t1 = (gettime_usec() - t1) / nb_tests;
+  printf(" test_bits   %.3f usec   %s\n", t1, (Crc!=0)?"| ERROR": "" );
+}
+
+/*********************************************************************
  * main
  *********************************************************************/
 
@@ -2047,7 +2169,7 @@
 	int c, what = 0;
 	int width, height;
 	uint32_t chksum = 0;
-  const char * test_bitstream = 0;
+	const char * test_bitstream = 0;
 
 	cpu_mask = 0;  // default => will use autodectect
 	for(c=1; c<argc; ++c)
@@ -2103,7 +2225,8 @@
 	if (what==0 || what==13) test_compiler();
 	if (what==0 || what==14) test_yuv();
 	if (what==0 || what==15) test_SSIM();
-
+	if (what==0 || what==16) test_yuv2();
+	if (what==0 || what==17) test_bits();
 
 	if (what==7) {
 		test_IEEE1180_compliance(-256, 255, 1);