--- trunk/xvidcore/examples/xvid_bench.c	2006/10/30 10:52:00	1742
+++ trunk/xvidcore/examples/xvid_bench.c	2008/11/14 15:43:28	1794
@@ -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 1.32 2006-10-30 10:52:00 Skal Exp $
+ * $Id: xvid_bench.c,v 1.36 2008-11-14 15:43:27 Isibaar 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>
 
@@ -118,7 +119,9 @@
 	{ "MMX    ", XVID_CPU_MMX },
 	{ "MMXEXT ", XVID_CPU_MMXEXT | XVID_CPU_MMX },
 	{ "SSE2   ", XVID_CPU_SSE2 | XVID_CPU_MMX },
-	{ "3DNOW  ", XVID_CPU_3DNOW },
+	{ "SSE3   ", XVID_CPU_SSE3 | XVID_CPU_SSE2 | XVID_CPU_MMX },
+	{ "SSE41  ", XVID_CPU_SSE41| XVID_CPU_SSE3 | XVID_CPU_SSE2 | XVID_CPU_MMX },
+    { "3DNOW  ", XVID_CPU_3DNOW },
 	{ "3DNOWE ", XVID_CPU_3DNOW | XVID_CPU_3DNOWEXT },
 #endif
 #ifdef ARCH_IS_PPC
@@ -1932,7 +1935,7 @@
 
 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;
@@ -2009,6 +2012,7 @@
     printf( "ERROR! please post your platform/compiler specs to xvid-devel@xvid.org !\n" );
   }
 }
+
 /*********************************************************************
  * test SSIM functions
  *********************************************************************/
@@ -2019,7 +2023,7 @@
 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);
 
@@ -2057,7 +2061,8 @@
 			continue;
 		lum8x8 = lum_8x8_c;
 		lum2x8 = lum_2x8_c;
-		csim   = iconsim_c;
+		csim   = consim_c;
+#ifdef ARCH_IS_IA32
 		if (cpu->cpu & XVID_CPU_MMX){
 			lum8x8 = lum_8x8_mmx;
 			csim = consim_mmx;
@@ -2065,7 +2070,7 @@
 		if (cpu->cpu & XVID_CPU_MMX){
 			csim = consim_sse2;
 		}
-
+#endif
 		t = gettime_usec();
 		emms();
 		for(tst=0; tst<nb_tests; ++tst) m = lum8x8(Ref1, 16);
@@ -2091,13 +2096,68 @@
 		t = (gettime_usec() - t) / nb_tests;
 		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
  *********************************************************************/
 
@@ -2112,7 +2172,7 @@
 	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)
@@ -2122,7 +2182,9 @@
 	  else if (!strcmp(argv[c], "-mmx"))    cpu_mask = XVID_CPU_MMX    | XVID_CPU_FORCE;
 	  else if (!strcmp(argv[c], "-mmxext")) cpu_mask = XVID_CPU_MMXEXT | XVID_CPU_MMX | XVID_CPU_FORCE;
 	  else if (!strcmp(argv[c], "-sse2"))   cpu_mask = XVID_CPU_SSE2   | XVID_CPU_MMXEXT | XVID_CPU_MMX | XVID_CPU_FORCE;
-	  else if (!strcmp(argv[c], "-3dnow"))  cpu_mask = XVID_CPU_3DNOW  | XVID_CPU_FORCE;
+	  else if (!strcmp(argv[c], "-sse3"))   cpu_mask = XVID_CPU_SSE3   | XVID_CPU_SSE2 | XVID_CPU_MMXEXT | XVID_CPU_MMX | XVID_CPU_FORCE;
+	  else if (!strcmp(argv[c], "-sse4"))   cpu_mask = XVID_CPU_SSE41  | XVID_CPU_SSE3 | XVID_CPU_SSE2 | XVID_CPU_MMXEXT | XVID_CPU_MMX | XVID_CPU_FORCE;
+      else if (!strcmp(argv[c], "-3dnow"))  cpu_mask = XVID_CPU_3DNOW  | XVID_CPU_FORCE;
 	  else if (!strcmp(argv[c], "-3dnowe")) cpu_mask = XVID_CPU_3DNOW  | XVID_CPU_3DNOWEXT | XVID_CPU_FORCE;
 	  else if (!strcmp(argv[c], "-altivec")) cpu_mask = XVID_CPU_ALTIVEC | XVID_CPU_FORCE;
 	  else if (!strcmp(argv[c], "-spd")) {
@@ -2169,7 +2231,7 @@
 	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);