diff --git a/Makefile.system b/Makefile.system
index 239047f36..eac61e961 100644
--- a/Makefile.system
+++ b/Makefile.system
@@ -277,14 +277,14 @@ ifeq ($(ARCH), x86)
 DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
 	       CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
 ifneq ($(NO_AVX), 1)
-DYNAMIC_CORE += SANDYBRIDGE 
+DYNAMIC_CORE += SANDYBRIDGE BULLDOZER
 endif
 endif
 
 ifeq ($(ARCH), x86_64)
 DYNAMIC_CORE = PRESCOTT CORE2 PENRYN DUNNINGTON NEHALEM OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
 ifneq ($(NO_AVX), 1)
-DYNAMIC_CORE += SANDYBRIDGE 
+DYNAMIC_CORE += SANDYBRIDGE BULLDOZER
 endif
 endif
 
diff --git a/TargetList.txt b/TargetList.txt
index 1a212e6ca..c859db082 100644
--- a/TargetList.txt
+++ b/TargetList.txt
@@ -29,6 +29,7 @@ BARCELONA
 SHANGHAI
 ISTANBUL
 BOBCAT
+BULLDOZER
 
 c)VIA CPU:
 SSE_GENERIC
diff --git a/cpuid.h b/cpuid.h
index bb57ad92d..c52d503cc 100644
--- a/cpuid.h
+++ b/cpuid.h
@@ -125,7 +125,8 @@
 #define HAVE_MISALIGNSSE (1 << 15)
 #define HAVE_128BITFPU   (1 << 16)
 #define HAVE_FASTMOVU    (1 << 17)
-#define HAVE_AVX     (1 <<  18)
+#define HAVE_AVX      (1 <<  18)
+#define HAVE_FMA4     (1 <<  19)
 
 #define CACHE_INFO_L1_I     1
 #define CACHE_INFO_L1_D     2
diff --git a/cpuid_x86.c b/cpuid_x86.c
index a19dedeee..317774691 100644
--- a/cpuid_x86.c
+++ b/cpuid_x86.c
@@ -43,6 +43,8 @@
 #ifdef NO_AVX
 #define CPUTYPE_SANDYBRIDGE CPUTYPE_NEHALEM
 #define CORE_SANDYBRIDGE CORE_NEHALEM
+#define CPUTYPE_BULLDOZER CPUTYPE_BARCELONA
+#define CORE_BULLDOZER CORE_BARCELONA
 #endif
 
 #ifndef CPUIDEMU
@@ -229,6 +231,9 @@ int get_cputype(int gettype){
       cpuid(0x80000001, &eax, &ebx, &ecx, &edx);
       if ((ecx & (1 <<  6)) != 0) feature |= HAVE_SSE4A;
       if ((ecx & (1 <<  7)) != 0) feature |= HAVE_MISALIGNSSE;
+#ifndef NO_AVX
+      if ((ecx & (1 <<  16)) != 0) feature |= HAVE_FMA4;
+#endif
       if ((edx & (1 << 30)) != 0) feature |= HAVE_3DNOWEX;
       if ((edx & (1 << 31)) != 0) feature |= HAVE_3DNOW;
     }
@@ -1078,8 +1083,12 @@ int get_cpuname(void){
 	return CPUTYPE_OPTERON;
       case  1:
       case 10:
-      case  6:   //AMD Bulldozer Opteron 6200 / Opteron 4200 / AMD FX-Series
 	return CPUTYPE_BARCELONA;
+      case  6:   //AMD Bulldozer Opteron 6200 / Opteron 4200 / AMD FX-Series
+	if(support_avx())
+	  return CPUTYPE_BULLDOZER;
+	else
+	  return CPUTYPE_BARCELONA; //OS don't support AVX.
       case  5:
 	return CPUTYPE_BOBCAT;
       }
@@ -1432,8 +1441,13 @@ int get_coretype(void){
     if (family == 0xf){
       if ((exfamily == 0) || (exfamily == 2)) return CORE_OPTERON; 
       else if (exfamily == 5) return CORE_BOBCAT; 
-      else if (exfamily == 6) return CORE_BARCELONA;  //AMD Bulldozer Opteron 6200 / Opteron 4200 / AMD FX-Series
-      else return CORE_BARCELONA;
+      else if (exfamily == 6) {
+	//AMD Bulldozer Opteron 6200 / Opteron 4200 / AMD FX-Series
+	if(support_avx())
+	  return CORE_BULLDOZER;
+	else
+	  return CORE_BARCELONA; //OS don't support AVX. Use old kernels.
+      }else return CORE_BARCELONA;
     }
   }
 
@@ -1519,6 +1533,7 @@ void get_cpuconfig(void){
     if (features & HAVE_AVX )    printf("#define HAVE_AVX\n");
     if (features & HAVE_3DNOWEX) printf("#define HAVE_3DNOWEX\n");
     if (features & HAVE_3DNOW)   printf("#define HAVE_3DNOW\n");
+    if (features & HAVE_FMA4 )    printf("#define HAVE_FMA4\n");
     if (features & HAVE_CFLUSH)  printf("#define HAVE_CFLUSH\n");
     if (features & HAVE_HIT)     printf("#define HAVE_HIT 1\n");
     if (features & HAVE_MISALIGNSSE) printf("#define HAVE_MISALIGNSSE\n");
@@ -1585,5 +1600,6 @@ void get_sse(void){
   if (features & HAVE_AVX )    printf("HAVE_AVX=1\n");
   if (features & HAVE_3DNOWEX) printf("HAVE_3DNOWEX=1\n");
   if (features & HAVE_3DNOW)   printf("HAVE_3DNOW=1\n");
+  if (features & HAVE_FMA4 )    printf("HAVE_FMA4=1\n");
 
 }
diff --git a/driver/others/dynamic.c b/driver/others/dynamic.c
index b6f27d0ad..6523abb4d 100644
--- a/driver/others/dynamic.c
+++ b/driver/others/dynamic.c
@@ -63,9 +63,11 @@ extern gotoblas_t  gotoblas_BARCELONA;
 extern gotoblas_t  gotoblas_BOBCAT;
 #ifndef NO_AVX
 extern gotoblas_t  gotoblas_SANDYBRIDGE;
+extern gotoblas_t  gotoblas_BULLDOZER;
 #else
 //Use NEHALEM kernels for sandy bridge
 #define gotoblas_SANDYBRIDGE gotoblas_NEHALEM
+#define gotoblas_BULLDOZER gotoblas_BARCELONA
 #endif
 
 
@@ -204,6 +206,14 @@ static gotoblas_t *get_coretype(void){
 	else return &gotoblas_OPTERON;
       }  else if (exfamily == 5) {
 	return &gotoblas_BOBCAT;
+      } else if (exfamily == 6) {
+	//AMD Bulldozer Opteron 6200 / Opteron 4200 / AMD FX-Series
+	  if(support_avx())
+	    return &gotoblas_BULLDOZER;
+	  else{
+	    fprintf(stderr, "OpenBLAS : Your OS doesn't support AVX. Use Barcelona kernels.\n");
+	    return &gotoblas_BARCELONA; //OS doesn't support AVX. Use old kernels.
+	  }	
       } else {
 	return &gotoblas_BARCELONA;
       }
@@ -240,6 +250,7 @@ static char *corename[] = {
     "Nano",
     "Sandybridge",
     "Bobcat",
+    "Bulldozer",
 };
 
 char *gotoblas_corename(void) {
@@ -261,6 +272,7 @@ char *gotoblas_corename(void) {
   if (gotoblas == &gotoblas_NANO)         return corename[15];
   if (gotoblas == &gotoblas_SANDYBRIDGE)  return corename[16];
   if (gotoblas == &gotoblas_BOBCAT)       return corename[17];
+  if (gotoblas == &gotoblas_BULLDOZER)    return corename[18];
 
   return corename[0];
 }
diff --git a/driver/others/parameter.c b/driver/others/parameter.c
index d261e5a4e..58e5fb11d 100644
--- a/driver/others/parameter.c
+++ b/driver/others/parameter.c
@@ -163,7 +163,7 @@ int get_L2_size(void){
 
   int eax, ebx, ecx, edx;
 
-#if defined(ATHLON) || defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || \
+#if defined(ATHLON) || defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER) || \
     defined(CORE_PRESCOTT) || defined(CORE_CORE2) || defined(PENRYN) || defined(DUNNINGTON) || \
   defined(CORE_NEHALEM) || defined(CORE_SANDYBRIDGE) || defined(ATOM) || defined(GENERIC)
 
diff --git a/getarch.c b/getarch.c
index 5916a9a04..2b9856338 100644
--- a/getarch.c
+++ b/getarch.c
@@ -350,7 +350,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #define CORENAME  "OPTERON"
 #endif
 
-#if defined(FORCE_BARCELONA) || defined(FORCE_SHANGHAI) || defined(FORCE_ISTANBUL) || defined (FORCE_BULLDOZER)
+#if defined(FORCE_BARCELONA) || defined(FORCE_SHANGHAI) || defined(FORCE_ISTANBUL)
 #define FORCE
 #define FORCE_INTEL
 #define ARCHITECTURE    "X86"
@@ -380,6 +380,22 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #define CORENAME  "BOBCAT"
 #endif
 
+#if defined (FORCE_BULLDOZER)
+#define FORCE
+#define FORCE_INTEL
+#define ARCHITECTURE    "X86"
+#define SUBARCHITECTURE "BULLDOZER"
+#define ARCHCONFIG   "-DBULLDOZER " \
+		     "-DL1_DATA_SIZE=49152 -DL1_DATA_LINESIZE=64 " \
+		     "-DL2_SIZE=1024000 -DL2_LINESIZE=64  -DL3_SIZE=16777216 " \
+		     "-DDTB_DEFAULT_ENTRIES=32 -DDTB_SIZE=4096 " \
+		     "-DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3 " \
+		     "-DHAVE_SSE4A -DHAVE_MISALIGNSSE -DHAVE_128BITFPU -DHAVE_FASTMOVU " \
+                     "-DHAVE_AVX -DHAVE_FMA4"
+#define LIBNAME   "bulldozer"
+#define CORENAME  "BULLDOZER"
+#endif
+
 #ifdef FORCE_SSE_GENERIC
 #define FORCE
 #define FORCE_INTEL
diff --git a/kernel/setparam-ref.c b/kernel/setparam-ref.c
index 4deabdacb..83f2b047f 100644
--- a/kernel/setparam-ref.c
+++ b/kernel/setparam-ref.c
@@ -810,6 +810,22 @@ static void init_parameter(void) {
 #endif
 #endif
 
+#ifdef BULLDOZER
+
+#ifdef DEBUG
+  fprintf(stderr, "Bulldozer\n");
+#endif
+
+  TABLE_NAME.sgemm_p = SGEMM_DEFAULT_P;
+  TABLE_NAME.dgemm_p = DGEMM_DEFAULT_P;
+  TABLE_NAME.cgemm_p = CGEMM_DEFAULT_P;
+  TABLE_NAME.zgemm_p = ZGEMM_DEFAULT_P;
+#ifdef EXPRECISION
+  TABLE_NAME.qgemm_p = QGEMM_DEFAULT_P;
+  TABLE_NAME.xgemm_p = XGEMM_DEFAULT_P;
+#endif
+#endif
+
 #ifdef NANO
 
 #ifdef DEBUG
diff --git a/kernel/x86/KERNEL.BULLDOZER b/kernel/x86/KERNEL.BULLDOZER
new file mode 100644
index 000000000..231350a62
--- /dev/null
+++ b/kernel/x86/KERNEL.BULLDOZER
@@ -0,0 +1,59 @@
+SGEMMKERNEL    =  gemm_kernel_4x4_barcelona.S
+SGEMMINCOPY    =  
+SGEMMITCOPY    =  
+SGEMMONCOPY    =  ../generic/gemm_ncopy_4.c
+SGEMMOTCOPY    =  ../generic/gemm_tcopy_4.c
+SGEMMINCOPYOBJ =  
+SGEMMITCOPYOBJ =  
+SGEMMONCOPYOBJ =  sgemm_oncopy$(TSUFFIX).$(SUFFIX)
+SGEMMOTCOPYOBJ =  sgemm_otcopy$(TSUFFIX).$(SUFFIX)
+DGEMMKERNEL    =  gemm_kernel_2x4_barcelona.S
+DGEMMINCOPY    =  ../generic/gemm_ncopy_2.c
+DGEMMITCOPY    =  ../generic/gemm_tcopy_2.c
+DGEMMONCOPY    =  ../generic/gemm_ncopy_4.c
+DGEMMOTCOPY    =  ../generic/gemm_tcopy_4.c
+DGEMMINCOPYOBJ =  dgemm_incopy$(TSUFFIX).$(SUFFIX)
+DGEMMITCOPYOBJ =  dgemm_itcopy$(TSUFFIX).$(SUFFIX)
+DGEMMONCOPYOBJ =  dgemm_oncopy$(TSUFFIX).$(SUFFIX)
+DGEMMOTCOPYOBJ =  dgemm_otcopy$(TSUFFIX).$(SUFFIX)
+CGEMMKERNEL    =  zgemm_kernel_2x2_barcelona.S
+CGEMMINCOPY    =  
+CGEMMITCOPY    =  
+CGEMMONCOPY    =  ../generic/zgemm_ncopy_2.c
+CGEMMOTCOPY    =  ../generic/zgemm_tcopy_2.c
+CGEMMINCOPYOBJ =
+CGEMMITCOPYOBJ =  
+CGEMMONCOPYOBJ =  cgemm_oncopy$(TSUFFIX).$(SUFFIX)
+CGEMMOTCOPYOBJ =  cgemm_otcopy$(TSUFFIX).$(SUFFIX)
+ZGEMMKERNEL    =  zgemm_kernel_1x2_barcelona.S
+ZGEMMINCOPY    =  ../generic/zgemm_ncopy_1.c
+ZGEMMITCOPY    =  ../generic/zgemm_tcopy_1.c
+ZGEMMONCOPY    =  ../generic/zgemm_ncopy_2.c
+ZGEMMOTCOPY    =  ../generic/zgemm_tcopy_2.c
+ZGEMMINCOPYOBJ =  zgemm_incopy$(TSUFFIX).$(SUFFIX)
+ZGEMMITCOPYOBJ =  zgemm_itcopy$(TSUFFIX).$(SUFFIX)
+ZGEMMONCOPYOBJ =  zgemm_oncopy$(TSUFFIX).$(SUFFIX)
+ZGEMMOTCOPYOBJ =  zgemm_otcopy$(TSUFFIX).$(SUFFIX)
+
+STRSMKERNEL_LN	=  trsm_kernel_LN_4x4_sse.S
+STRSMKERNEL_LT	=  trsm_kernel_LT_4x4_sse.S
+STRSMKERNEL_RN	=  trsm_kernel_LT_4x4_sse.S
+STRSMKERNEL_RT	=  trsm_kernel_RT_4x4_sse.S
+
+DTRSMKERNEL_LN	=  trsm_kernel_LN_2x4_sse2.S
+DTRSMKERNEL_LT	=  trsm_kernel_LT_2x4_sse2.S
+DTRSMKERNEL_RN	=  trsm_kernel_LT_2x4_sse2.S
+DTRSMKERNEL_RT	=  trsm_kernel_RT_2x4_sse2.S
+
+CTRSMKERNEL_LN	=  ztrsm_kernel_LN_2x2_sse.S
+CTRSMKERNEL_LT	=  ztrsm_kernel_LT_2x2_sse.S
+CTRSMKERNEL_RN	=  ztrsm_kernel_LT_2x2_sse.S
+CTRSMKERNEL_RT	=  ztrsm_kernel_RT_2x2_sse.S
+
+ZTRSMKERNEL_LN	=  ztrsm_kernel_LT_1x2_sse2.S
+ZTRSMKERNEL_LT	=  ztrsm_kernel_LT_1x2_sse2.S
+ZTRSMKERNEL_RN	=  ztrsm_kernel_LT_1x2_sse2.S
+ZTRSMKERNEL_RT	=  ztrsm_kernel_RT_1x2_sse2.S
+
+CGEMM3MKERNEL    =  zgemm3m_kernel_4x4_barcelona.S
+ZGEMM3MKERNEL    =  zgemm3m_kernel_2x4_barcelona.S
diff --git a/kernel/x86/gemm_kernel_4x4_barcelona.S b/kernel/x86/gemm_kernel_4x4_barcelona.S
index 18b9a43bd..f081aec2a 100644
--- a/kernel/x86/gemm_kernel_4x4_barcelona.S
+++ b/kernel/x86/gemm_kernel_4x4_barcelona.S
@@ -596,7 +596,7 @@
 .L22:
 	mulps	%xmm0, %xmm2
 	addps	%xmm2, %xmm4
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movsd	 4 * SIZE(BB), %xmm2
@@ -842,7 +842,7 @@
 .L32:
 	mulss	%xmm0, %xmm2
 	addss	%xmm2, %xmm4
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movss	 4 * SIZE(BB), %xmm2
@@ -1168,7 +1168,7 @@
 
 .L52:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	mulps	 4 * SIZE(BB), %xmm0
@@ -1198,7 +1198,7 @@
 	addps	%xmm0, %xmm5
 	movaps	32 * SIZE(AA), %xmm0
 
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 16) * SIZE(AA)
 #endif
 	mulps	%xmm1, %xmm2
@@ -1347,7 +1347,7 @@
 	ALIGN_4
 
 .L62:
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 
@@ -1531,7 +1531,7 @@
 
 .L72:
 	mulss	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	mulss	 4 * SIZE(BB), %xmm0
@@ -1778,7 +1778,7 @@
 
 .L92:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movaps	 4 * SIZE(AA), %xmm0
@@ -1793,7 +1793,7 @@
 	mulps	12 * SIZE(BB), %xmm0
 	addps	%xmm0, %xmm7
 	movaps	32 * SIZE(AA), %xmm0
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 16) * SIZE(AA)
 #endif
 	mulps	%xmm1, %xmm3
@@ -1924,7 +1924,7 @@
 
 .L102:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movsd	 2 * SIZE(AA), %xmm0
@@ -2069,7 +2069,7 @@
 
 .L112:
 	mulss	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movss	 1 * SIZE(AA), %xmm0
diff --git a/kernel/x86/scal_sse.S b/kernel/x86/scal_sse.S
index aa5ab760e..48edfc585 100644
--- a/kernel/x86/scal_sse.S
+++ b/kernel/x86/scal_sse.S
@@ -269,7 +269,7 @@
 	sarl	$5, I
 	jle	.L113
 
-#if defined(BARCELONA)
+#if defined(BARCELONA) || defined(BULLDOZER)
 
 	movaps	%xmm0, %xmm1
 	mulps	-32 * SIZE(X), %xmm1
diff --git a/kernel/x86/scal_sse2.S b/kernel/x86/scal_sse2.S
index 67c1f437b..35b79132c 100644
--- a/kernel/x86/scal_sse2.S
+++ b/kernel/x86/scal_sse2.S
@@ -253,7 +253,7 @@
 	sarl	$4, I
 	jle	.L113
 
-#if defined(BARCELONA)
+#if defined(BARCELONA) || defined(BULLDOZER)
 
 	movaps  %xmm0, %xmm1
 	mulpd	-16 * SIZE(X), %xmm1
diff --git a/kernel/x86/trsm_kernel_LN_2x4_sse2.S b/kernel/x86/trsm_kernel_LN_2x4_sse2.S
index 2b6877a31..036e17338 100644
--- a/kernel/x86/trsm_kernel_LN_2x4_sse2.S
+++ b/kernel/x86/trsm_kernel_LN_2x4_sse2.S
@@ -69,7 +69,7 @@
 #define STACK_ALIGN	4096
 #define STACK_OFFSET	1024
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCH     prefetch
 #define PREFETCHSIZE (8 * 10 + 4)
 #endif
@@ -439,7 +439,7 @@
 .L22:
 	mulsd	%xmm0, %xmm2
 	addsd	%xmm2, %xmm4
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	PREFETCH (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movlpd	 2 * SIZE(BB), %xmm2
@@ -488,7 +488,7 @@
 	movlpd	40 * SIZE(BB), %xmm3
 	addsd	%xmm0, %xmm7
 	movlpd	 8 * SIZE(AA), %xmm0
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	PREFETCH (PREFETCHSIZE  + 8) * SIZE(AA)
 #endif
 	mulsd	%xmm1, %xmm2
@@ -1697,7 +1697,7 @@
 
 .L42:
 	mulpd	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	mulpd	 2 * SIZE(BB), %xmm0
@@ -1727,7 +1727,7 @@
 	addpd	%xmm0, %xmm7
 	movapd	16 * SIZE(AA), %xmm0
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 8) * SIZE(AA)
 #endif
 	mulpd	%xmm1, %xmm2
diff --git a/kernel/x86/trsm_kernel_LN_4x4_sse.S b/kernel/x86/trsm_kernel_LN_4x4_sse.S
index 82bb1d3ec..84da443a8 100644
--- a/kernel/x86/trsm_kernel_LN_4x4_sse.S
+++ b/kernel/x86/trsm_kernel_LN_4x4_sse.S
@@ -64,7 +64,7 @@
 #define BORIG	60(%esp)
 #define BUFFER 128(%esp)
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCH     prefetch
 #define PREFETCHW    prefetchw
 #define PREFETCHSIZE (16 * 10 + 8)
@@ -437,7 +437,7 @@
 .L32:
 	mulss	%xmm0, %xmm2
 	addss	%xmm2, %xmm4
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movss	 4 * SIZE(BB), %xmm2
@@ -833,7 +833,7 @@
 .L22:
 	mulps	%xmm0, %xmm2
 	addps	%xmm2, %xmm4
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movaps	 4 * SIZE(BB), %xmm2
@@ -1848,7 +1848,7 @@
 
 .L72:
 	mulss	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	mulss	 4 * SIZE(BB), %xmm0
@@ -2109,7 +2109,7 @@
 	ALIGN_4
 
 .L62:
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 
@@ -2429,7 +2429,7 @@
 
 .L52:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	mulps	 4 * SIZE(BB), %xmm0
@@ -2459,7 +2459,7 @@
 	addps	%xmm0, %xmm5
 	movaps	32 * SIZE(AA), %xmm0
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 16) * SIZE(AA)
 #endif
 	mulps	%xmm1, %xmm2
@@ -2952,7 +2952,7 @@
 
 .L112:
 	mulss	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movss	 1 * SIZE(AA), %xmm0
@@ -3148,7 +3148,7 @@
 
 .L102:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movsd	 2 * SIZE(AA), %xmm0
@@ -3389,7 +3389,7 @@
 
 .L92:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movaps	 4 * SIZE(AA), %xmm0
@@ -3404,7 +3404,7 @@
 	mulps	12 * SIZE(BB), %xmm0
 	addps	%xmm0, %xmm7
 	movaps	32 * SIZE(AA), %xmm0
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 16) * SIZE(AA)
 #endif
 	mulps	%xmm1, %xmm3
diff --git a/kernel/x86/trsm_kernel_LT_2x4_sse2.S b/kernel/x86/trsm_kernel_LT_2x4_sse2.S
index d81177b7e..0bd924cba 100644
--- a/kernel/x86/trsm_kernel_LT_2x4_sse2.S
+++ b/kernel/x86/trsm_kernel_LT_2x4_sse2.S
@@ -69,7 +69,7 @@
 #define STACK_ALIGN	4096
 #define STACK_OFFSET	1024
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCH     prefetch
 #define PREFETCHSIZE (8 * 10 + 4)
 #endif
@@ -910,7 +910,7 @@
 .L22:
 	mulsd	%xmm0, %xmm2
 	addsd	%xmm2, %xmm4
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	PREFETCH (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movlpd	 2 * SIZE(BB), %xmm2
@@ -959,7 +959,7 @@
 	movlpd	40 * SIZE(BB), %xmm3
 	addsd	%xmm0, %xmm7
 	movlpd	 8 * SIZE(AA), %xmm0
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	PREFETCH (PREFETCHSIZE  + 8) * SIZE(AA)
 #endif
 	mulsd	%xmm1, %xmm2
@@ -1439,7 +1439,7 @@
 
 .L42:
 	mulpd	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	mulpd	 2 * SIZE(BB), %xmm0
@@ -1469,7 +1469,7 @@
 	addpd	%xmm0, %xmm7
 	movapd	16 * SIZE(AA), %xmm0
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 8) * SIZE(AA)
 #endif
 	mulpd	%xmm1, %xmm2
diff --git a/kernel/x86/trsm_kernel_LT_4x4_sse.S b/kernel/x86/trsm_kernel_LT_4x4_sse.S
index 854c44e7a..de7c04593 100644
--- a/kernel/x86/trsm_kernel_LT_4x4_sse.S
+++ b/kernel/x86/trsm_kernel_LT_4x4_sse.S
@@ -64,7 +64,7 @@
 #define BORIG	60(%esp)
 #define BUFFER 128(%esp)
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCH     prefetch
 #define PREFETCHW    prefetchw
 #define PREFETCHSIZE (16 * 10 + 8)
@@ -872,7 +872,7 @@
 .L22:
 	mulps	%xmm0, %xmm2
 	addps	%xmm2, %xmm4
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movaps	 4 * SIZE(BB), %xmm2
@@ -1316,7 +1316,7 @@
 .L32:
 	mulss	%xmm0, %xmm2
 	addss	%xmm2, %xmm4
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movss	 4 * SIZE(BB), %xmm2
@@ -1855,7 +1855,7 @@
 
 .L52:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	mulps	 4 * SIZE(BB), %xmm0
@@ -1885,7 +1885,7 @@
 	addps	%xmm0, %xmm5
 	movaps	32 * SIZE(AA), %xmm0
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 16) * SIZE(AA)
 #endif
 	mulps	%xmm1, %xmm2
@@ -2249,7 +2249,7 @@
 	ALIGN_4
 
 .L62:
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 
@@ -2562,7 +2562,7 @@
 
 .L72:
 	mulss	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	mulss	 4 * SIZE(BB), %xmm0
@@ -2957,7 +2957,7 @@
 
 .L92:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movaps	 4 * SIZE(AA), %xmm0
@@ -2972,7 +2972,7 @@
 	mulps	12 * SIZE(BB), %xmm0
 	addps	%xmm0, %xmm7
 	movaps	32 * SIZE(AA), %xmm0
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 16) * SIZE(AA)
 #endif
 	mulps	%xmm1, %xmm3
@@ -3280,7 +3280,7 @@
 
 .L102:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movsd	 2 * SIZE(AA), %xmm0
@@ -3515,7 +3515,7 @@
 
 .L112:
 	mulss	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movss	 1 * SIZE(AA), %xmm0
diff --git a/kernel/x86/trsm_kernel_RT_2x4_sse2.S b/kernel/x86/trsm_kernel_RT_2x4_sse2.S
index f7a08c699..f5d5ad465 100644
--- a/kernel/x86/trsm_kernel_RT_2x4_sse2.S
+++ b/kernel/x86/trsm_kernel_RT_2x4_sse2.S
@@ -69,7 +69,7 @@
 #define STACK_ALIGN	4096
 #define STACK_OFFSET	1024
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCH     prefetch
 #define PREFETCHSIZE (8 * 10 + 4)
 #endif
@@ -1036,7 +1036,7 @@
 
 .L42:
 	mulpd	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	mulpd	 2 * SIZE(BB), %xmm0
@@ -1066,7 +1066,7 @@
 	addpd	%xmm0, %xmm7
 	movapd	16 * SIZE(AA), %xmm0
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 8) * SIZE(AA)
 #endif
 	mulpd	%xmm1, %xmm2
@@ -2224,7 +2224,7 @@
 .L22:
 	mulsd	%xmm0, %xmm2
 	addsd	%xmm2, %xmm4
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	PREFETCH (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movlpd	 2 * SIZE(BB), %xmm2
@@ -2273,7 +2273,7 @@
 	movlpd	40 * SIZE(BB), %xmm3
 	addsd	%xmm0, %xmm7
 	movlpd	 8 * SIZE(AA), %xmm0
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	PREFETCH (PREFETCHSIZE  + 8) * SIZE(AA)
 #endif
 	mulsd	%xmm1, %xmm2
diff --git a/kernel/x86/trsm_kernel_RT_4x4_sse.S b/kernel/x86/trsm_kernel_RT_4x4_sse.S
index 80dc2451c..5c2dcd0d6 100644
--- a/kernel/x86/trsm_kernel_RT_4x4_sse.S
+++ b/kernel/x86/trsm_kernel_RT_4x4_sse.S
@@ -64,7 +64,7 @@
 #define BORIG	60(%esp)
 #define BUFFER 128(%esp)
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCH     prefetch
 #define PREFETCHW    prefetchw
 #define PREFETCHSIZE (16 * 10 + 8)
@@ -439,7 +439,7 @@
 
 .L92:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movaps	 4 * SIZE(AA), %xmm0
@@ -454,7 +454,7 @@
 	mulps	12 * SIZE(BB), %xmm0
 	addps	%xmm0, %xmm7
 	movaps	32 * SIZE(AA), %xmm0
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 16) * SIZE(AA)
 #endif
 	mulps	%xmm1, %xmm3
@@ -758,7 +758,7 @@
 
 .L102:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movsd	 2 * SIZE(AA), %xmm0
@@ -993,7 +993,7 @@
 
 .L112:
 	mulss	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movss	 1 * SIZE(AA), %xmm0
@@ -1324,7 +1324,7 @@
 
 .L52:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	mulps	 4 * SIZE(BB), %xmm0
@@ -1354,7 +1354,7 @@
 	addps	%xmm0, %xmm5
 	movaps	32 * SIZE(AA), %xmm0
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 16) * SIZE(AA)
 #endif
 	mulps	%xmm1, %xmm2
@@ -1718,7 +1718,7 @@
 	ALIGN_4
 
 .L62:
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 
@@ -2031,7 +2031,7 @@
 
 .L72:
 	mulss	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	mulss	 4 * SIZE(BB), %xmm0
@@ -2859,7 +2859,7 @@
 .L22:
 	mulps	%xmm0, %xmm2
 	addps	%xmm2, %xmm4
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movaps	 4 * SIZE(BB), %xmm2
@@ -3303,7 +3303,7 @@
 .L32:
 	mulss	%xmm0, %xmm2
 	addss	%xmm2, %xmm4
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movss	 4 * SIZE(BB), %xmm2
diff --git a/kernel/x86/zgemm3m_kernel_4x4_barcelona.S b/kernel/x86/zgemm3m_kernel_4x4_barcelona.S
index 29158df25..623f0beec 100644
--- a/kernel/x86/zgemm3m_kernel_4x4_barcelona.S
+++ b/kernel/x86/zgemm3m_kernel_4x4_barcelona.S
@@ -74,7 +74,7 @@
 #define	BB	%ecx
 #define LDC	%ebp
 
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 #define movsd	movlps
 #endif
 
@@ -625,7 +625,7 @@
 .L22:
 	mulps	%xmm0, %xmm2
 	addps	%xmm2, %xmm4
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movsd	 4 * SIZE(BB), %xmm2
@@ -870,7 +870,7 @@
 .L32:
 	mulss	%xmm0, %xmm2
 	addss	%xmm2, %xmm4
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movss	 4 * SIZE(BB), %xmm2
@@ -1173,7 +1173,7 @@
 
 .L52:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	mulps	 4 * SIZE(BB), %xmm0
@@ -1203,7 +1203,7 @@
 	addps	%xmm0, %xmm5
 	movaps	32 * SIZE(AA), %xmm0
 
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 16) * SIZE(AA)
 #endif
 	mulps	%xmm1, %xmm2
@@ -1359,7 +1359,7 @@
 	ALIGN_4
 
 .L62:
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 
@@ -1536,7 +1536,7 @@
 
 .L72:
 	mulss	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	mulss	 4 * SIZE(BB), %xmm0
@@ -1794,7 +1794,7 @@
 
 .L92:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movaps	 4 * SIZE(AA), %xmm0
@@ -1809,7 +1809,7 @@
 	mulps	12 * SIZE(BB), %xmm0
 	addps	%xmm0, %xmm7
 	movaps	32 * SIZE(AA), %xmm0
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 16) * SIZE(AA)
 #endif
 	mulps	%xmm1, %xmm3
@@ -1936,7 +1936,7 @@
 
 .L102:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movsd	 2 * SIZE(AA), %xmm0
@@ -2069,7 +2069,7 @@
 
 .L112:
 	mulss	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movss	 1 * SIZE(AA), %xmm0
diff --git a/kernel/x86/zgemv_n_sse.S b/kernel/x86/zgemv_n_sse.S
index 8e28bb8e6..0087ac6f4 100644
--- a/kernel/x86/zgemv_n_sse.S
+++ b/kernel/x86/zgemv_n_sse.S
@@ -71,7 +71,7 @@
 #define movsd		movlps
 #endif
 
-#ifdef BARCELONA
+#if defined(BARCELONA)  || defined(BULLDOZER)
 #define PREFETCH	prefetchnta
 #define PREFETCHW	prefetchw
 #define PREFETCHSIZE	(16 * 5)
diff --git a/kernel/x86/zgemv_n_sse2.S b/kernel/x86/zgemv_n_sse2.S
index 607c51de0..f0f2dc0ec 100644
--- a/kernel/x86/zgemv_n_sse2.S
+++ b/kernel/x86/zgemv_n_sse2.S
@@ -58,7 +58,7 @@
 #define movsd		movlps
 #endif
 
-#ifdef BARCELONA
+#if defined(BARCELONA) || defined(BULLDOZER)
 #define PREFETCH	prefetchnta
 #define PREFETCHW	prefetchw
 #define PREFETCHSIZE	(8 * 5)
diff --git a/kernel/x86/zgemv_t_sse.S b/kernel/x86/zgemv_t_sse.S
index fb98226ee..c7ad91235 100644
--- a/kernel/x86/zgemv_t_sse.S
+++ b/kernel/x86/zgemv_t_sse.S
@@ -71,7 +71,7 @@
 #define movsd		movlps
 #endif
 
-#ifdef BARCELONA
+#if defined(BARCELONA) || defined(BULLDOZER)
 #define PREFETCH	prefetchnta
 #define PREFETCHW	prefetchw
 #define PREFETCHSIZE	(16 * 5)
diff --git a/kernel/x86/zgemv_t_sse2.S b/kernel/x86/zgemv_t_sse2.S
index e2f391a82..6c4842893 100644
--- a/kernel/x86/zgemv_t_sse2.S
+++ b/kernel/x86/zgemv_t_sse2.S
@@ -58,7 +58,7 @@
 #define movsd		movlps
 #endif
 
-#ifdef BARCELONA
+#if defined(BARCELONA) || defined(BULLDOZER)
 #define PREFETCH	prefetchnta
 #define PREFETCHW	prefetchw
 #define PREFETCHSIZE	(8 * 5)
diff --git a/kernel/x86/ztrsm_kernel_LN_2x2_sse.S b/kernel/x86/ztrsm_kernel_LN_2x2_sse.S
index ee9eb9d25..d32451574 100644
--- a/kernel/x86/ztrsm_kernel_LN_2x2_sse.S
+++ b/kernel/x86/ztrsm_kernel_LN_2x2_sse.S
@@ -75,7 +75,7 @@
 #define STACK_ALIGN	4096
 #define STACK_OFFSET	1024
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCHSIZE (16 * 10 + 8)
 #define WPREFETCHSIZE 112
 #define PREFETCH      prefetch
@@ -533,7 +533,7 @@
 	addps	%xmm0, %xmm7
 	movsd	16 * SIZE(AA), %xmm0
 	mulps	%xmm1, %xmm2
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht1     (PREFETCHSIZE + 16) * SIZE(AA)
 #endif
 	addps	%xmm2, %xmm4
diff --git a/kernel/x86/ztrsm_kernel_LT_2x2_sse.S b/kernel/x86/ztrsm_kernel_LT_2x2_sse.S
index 9ef572470..9f9449852 100644
--- a/kernel/x86/ztrsm_kernel_LT_2x2_sse.S
+++ b/kernel/x86/ztrsm_kernel_LT_2x2_sse.S
@@ -75,7 +75,7 @@
 #define STACK_ALIGN	4096
 #define STACK_OFFSET	1024
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCHSIZE (16 * 10 + 8)
 #define WPREFETCHSIZE 112
 #define PREFETCH      prefetch
@@ -994,7 +994,7 @@
 	addps	%xmm0, %xmm7
 	movsd	16 * SIZE(AA), %xmm0
 	mulps	%xmm1, %xmm2
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht1     (PREFETCHSIZE + 16) * SIZE(AA)
 #endif
 	addps	%xmm2, %xmm4
diff --git a/kernel/x86/ztrsm_kernel_RT_2x2_sse.S b/kernel/x86/ztrsm_kernel_RT_2x2_sse.S
index cd1bf2f53..dd0c5ab21 100644
--- a/kernel/x86/ztrsm_kernel_RT_2x2_sse.S
+++ b/kernel/x86/ztrsm_kernel_RT_2x2_sse.S
@@ -75,7 +75,7 @@
 #define STACK_ALIGN	4096
 #define STACK_OFFSET	1024
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCHSIZE (16 * 10 + 8)
 #define WPREFETCHSIZE 112
 #define PREFETCH      prefetch
@@ -1820,7 +1820,7 @@
 	addps	%xmm0, %xmm7
 	movsd	16 * SIZE(AA), %xmm0
 	mulps	%xmm1, %xmm2
-#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht1     (PREFETCHSIZE + 16) * SIZE(AA)
 #endif
 	addps	%xmm2, %xmm4
diff --git a/kernel/x86_64/KERNEL.BULLDOZER b/kernel/x86_64/KERNEL.BULLDOZER
new file mode 100644
index 000000000..d59668519
--- /dev/null
+++ b/kernel/x86_64/KERNEL.BULLDOZER
@@ -0,0 +1,62 @@
+ZGEMVNKERNEL = zgemv_n_dup.S
+ZGEMVTKERNEL = zgemv_t_dup.S
+
+SGEMMKERNEL    =  gemm_kernel_8x4_barcelona.S
+SGEMMINCOPY    =  ../generic/gemm_ncopy_8.c
+SGEMMITCOPY    =  ../generic/gemm_tcopy_8.c
+SGEMMONCOPY    =  gemm_ncopy_4_opteron.S
+SGEMMOTCOPY    =  gemm_tcopy_4_opteron.S
+SGEMMINCOPYOBJ =  sgemm_incopy$(TSUFFIX).$(SUFFIX) 
+SGEMMITCOPYOBJ =  sgemm_itcopy$(TSUFFIX).$(SUFFIX) 
+SGEMMONCOPYOBJ =  sgemm_oncopy$(TSUFFIX).$(SUFFIX)
+SGEMMOTCOPYOBJ =  sgemm_otcopy$(TSUFFIX).$(SUFFIX)
+DGEMMKERNEL    =  dgemm_kernel_4x4_bulldozer.S
+DGEMMINCOPY    =
+DGEMMITCOPY    =
+DGEMMONCOPY    =  gemm_ncopy_4_opteron.S
+DGEMMOTCOPY    =  gemm_tcopy_4_opteron.S
+DGEMMINCOPYOBJ =
+DGEMMITCOPYOBJ =
+DGEMMONCOPYOBJ =  dgemm_oncopy$(TSUFFIX).$(SUFFIX)
+DGEMMOTCOPYOBJ =  dgemm_otcopy$(TSUFFIX).$(SUFFIX)
+CGEMMKERNEL    =  zgemm_kernel_4x2_barcelona.S
+CGEMMINCOPY    =  ../generic/zgemm_ncopy_4.c
+CGEMMITCOPY    =  ../generic/zgemm_tcopy_4.c
+CGEMMONCOPY    =  zgemm_ncopy_2.S
+CGEMMOTCOPY    =  zgemm_tcopy_2.S
+CGEMMINCOPYOBJ =  cgemm_incopy$(TSUFFIX).$(SUFFIX)
+CGEMMITCOPYOBJ =  cgemm_itcopy$(TSUFFIX).$(SUFFIX)
+CGEMMONCOPYOBJ =  cgemm_oncopy$(TSUFFIX).$(SUFFIX)
+CGEMMOTCOPYOBJ =  cgemm_otcopy$(TSUFFIX).$(SUFFIX)
+ZGEMMKERNEL    =  zgemm_kernel_2x2_barcelona.S
+ZGEMMINCOPY    =
+ZGEMMITCOPY    =
+ZGEMMONCOPY    =  zgemm_ncopy_2.S
+ZGEMMOTCOPY    =  zgemm_tcopy_2.S
+ZGEMMINCOPYOBJ =
+ZGEMMITCOPYOBJ =
+ZGEMMONCOPYOBJ =  zgemm_oncopy$(TSUFFIX).$(SUFFIX)
+ZGEMMOTCOPYOBJ =  zgemm_otcopy$(TSUFFIX).$(SUFFIX)
+
+STRSMKERNEL_LN	=  trsm_kernel_LN_8x4_sse.S
+STRSMKERNEL_LT	=  trsm_kernel_LT_8x4_sse.S
+STRSMKERNEL_RN	=  trsm_kernel_LT_8x4_sse.S
+STRSMKERNEL_RT	=  trsm_kernel_RT_8x4_sse.S
+
+DTRSMKERNEL_LN	=  trsm_kernel_LN_4x4_barcelona.S
+DTRSMKERNEL_LT	=  trsm_kernel_LT_4x4_barcelona.S
+DTRSMKERNEL_RN	=  trsm_kernel_LT_4x4_barcelona.S
+DTRSMKERNEL_RT	=  trsm_kernel_RT_4x4_barcelona.S
+
+CTRSMKERNEL_LN	=  ztrsm_kernel_LN_4x2_sse.S
+CTRSMKERNEL_LT	=  ztrsm_kernel_LT_4x2_sse.S
+CTRSMKERNEL_RN	=  ztrsm_kernel_LT_4x2_sse.S
+CTRSMKERNEL_RT	=  ztrsm_kernel_RT_4x2_sse.S
+
+ZTRSMKERNEL_LN	=  ztrsm_kernel_LN_2x2_sse2.S
+ZTRSMKERNEL_LT	=  ztrsm_kernel_LT_2x2_sse2.S
+ZTRSMKERNEL_RN	=  ztrsm_kernel_LT_2x2_sse2.S
+ZTRSMKERNEL_RT	=  ztrsm_kernel_RT_2x2_sse2.S
+
+CGEMM3MKERNEL    =  zgemm3m_kernel_8x4_barcelona.S
+ZGEMM3MKERNEL    =  zgemm3m_kernel_4x4_barcelona.S
diff --git a/kernel/x86_64/dgemm_kernel_4x4_bulldozer.S b/kernel/x86_64/dgemm_kernel_4x4_bulldozer.S
new file mode 100644
index 000000000..b06b07edf
--- /dev/null
+++ b/kernel/x86_64/dgemm_kernel_4x4_bulldozer.S
@@ -0,0 +1,1860 @@
+/*********************************************************************/
+/* Copyright 2009, 2010 The University of Texas at Austin.           */
+/* All rights reserved.                                              */
+/*                                                                   */
+/* Redistribution and use in source and binary forms, with or        */
+/* without modification, are permitted provided that the following   */
+/* conditions are met:                                               */
+/*                                                                   */
+/*   1. Redistributions of source code must retain the above         */
+/*      copyright notice, this list of conditions and the following  */
+/*      disclaimer.                                                  */
+/*                                                                   */
+/*   2. Redistributions in binary form must reproduce the above      */
+/*      copyright notice, this list of conditions and the following  */
+/*      disclaimer in the documentation and/or other materials       */
+/*      provided with the distribution.                              */
+/*                                                                   */
+/*    THIS  SOFTWARE IS PROVIDED  BY THE  UNIVERSITY OF  TEXAS AT    */
+/*    AUSTIN  ``AS IS''  AND ANY  EXPRESS OR  IMPLIED WARRANTIES,    */
+/*    INCLUDING, BUT  NOT LIMITED  TO, THE IMPLIED  WARRANTIES OF    */
+/*    MERCHANTABILITY  AND FITNESS FOR  A PARTICULAR  PURPOSE ARE    */
+/*    DISCLAIMED.  IN  NO EVENT SHALL THE UNIVERSITY  OF TEXAS AT    */
+/*    AUSTIN OR CONTRIBUTORS BE  LIABLE FOR ANY DIRECT, INDIRECT,    */
+/*    INCIDENTAL,  SPECIAL, EXEMPLARY,  OR  CONSEQUENTIAL DAMAGES    */
+/*    (INCLUDING, BUT  NOT LIMITED TO,  PROCUREMENT OF SUBSTITUTE    */
+/*    GOODS  OR  SERVICES; LOSS  OF  USE,  DATA,  OR PROFITS;  OR    */
+/*    BUSINESS INTERRUPTION) HOWEVER CAUSED  AND ON ANY THEORY OF    */
+/*    LIABILITY, WHETHER  IN CONTRACT, STRICT  LIABILITY, OR TORT    */
+/*    (INCLUDING NEGLIGENCE OR OTHERWISE)  ARISING IN ANY WAY OUT    */
+/*    OF  THE  USE OF  THIS  SOFTWARE,  EVEN  IF ADVISED  OF  THE    */
+/*    POSSIBILITY OF SUCH DAMAGE.                                    */
+/*                                                                   */
+/* The views and conclusions contained in the software and           */
+/* documentation are those of the authors and should not be          */
+/* interpreted as representing official policies, either expressed   */
+/* or implied, of The University of Texas at Austin.                 */
+/*********************************************************************/
+
+#define ASSEMBLER
+#include "common.h"
+ 
+#define OLD_M	%rdi
+#define OLD_N	%rsi
+#define M	%r13
+#define N	%r14
+#define K	%rdx
+
+#define A	%rcx
+#define B	%r8
+#define C	%r9
+#define LDC	%r10
+	
+#define I	%r11
+#define AO	%rdi
+#define BO	%rsi
+#define	CO1	%r15
+#define CO2	%r12
+#define BB	%rbp
+#define	J	%rbx
+
+#ifndef WINDOWS_ABI
+
+#define STACKSIZE 96
+
+#define ALPHA	 48(%rsp)
+#define OFFSET	 56(%rsp)
+#define KK	 64(%rsp)
+#define KKK	 72(%rsp)
+
+#else
+
+#define STACKSIZE 256
+
+#define OLD_A		40 + STACKSIZE(%rsp)
+#define OLD_B		48 + STACKSIZE(%rsp)
+#define OLD_C		56 + STACKSIZE(%rsp)
+#define OLD_LDC		64 + STACKSIZE(%rsp)
+#define OLD_OFFSET	72 + STACKSIZE(%rsp)
+
+#define ALPHA	224(%rsp)
+#define OFFSET	232(%rsp)
+#define KK	240(%rsp)
+#define KKK	248(%rsp)
+
+#endif
+
+#define movapd	movaps
+#define movupd	movups
+
+#define KERNEL1(xx) \
+	vfmaddpd  %xmm8,%xmm1,%xmm0,%xmm8 ;\
+	vmovaps	%xmm2, %xmm0 ;\
+	vmovups -14 * SIZE(AO, %rax, 4),%xmm2 ;\
+	vfmaddpd %xmm9,%xmm3,%xmm0,%xmm9 ;\
+	vfmaddpd %xmm12,%xmm2,%xmm1,%xmm12 ;\
+	vmovddup	-14 * SIZE(BO, %rax, 4), %xmm1 ;\
+	vfmaddpd %xmm13,%xmm2,%xmm3,%xmm13 ;\
+	vmovddup	-13 * SIZE(BO, %rax, 4), %xmm3 ;\
+	vfmaddpd %xmm10,%xmm1,%xmm0,%xmm10 ;\
+	vfmaddpd %xmm11,%xmm3,%xmm0,%xmm11 ;\
+	vfmaddpd %xmm15,%xmm2, %xmm3,%xmm15 ;\
+	vfmaddpd %xmm14,%xmm2, %xmm1,%xmm14 ;\
+	vmovups	-12 * SIZE(AO, %rax, 4), %xmm0 ;\
+	vmovddup	-12 * SIZE(BO, %rax, 4), %xmm1 ;\
+ 	vmovddup	-11 * SIZE(BO, %rax, 4), %xmm3 ;\
+	vmovaps	%xmm0, %xmm2
+
+#define KERNEL2(xx) \
+	vfmaddpd %xmm8,%xmm1, %xmm0,%xmm8 ;\
+	vmovaps	%xmm2, %xmm0 ;\
+	vmovups -10 * SIZE(AO, %rax, 4),%xmm2 ;\
+/*A*/	vmovups	  (AO, %rax, 4), %xmm6 ;\
+	vfmaddpd %xmm12,%xmm2, %xmm1,%xmm12 ;\
+	vfmaddpd %xmm9,%xmm3, %xmm0,%xmm9 ;\
+	vmovddup	-10 * SIZE(BO, %rax, 4), %xmm1 ;\
+	vfmaddpd %xmm13,%xmm2, %xmm3,%xmm13 ;\
+	vmovddup	 -9 * SIZE(BO, %rax, 4), %xmm3 ;\
+	vfmaddpd %xmm10,%xmm1, %xmm0,%xmm10 ;\
+	vfmaddpd %xmm14,%xmm2, %xmm1,%xmm14 ;\
+/**/	vmovddup	  (BO, %rax, 4), %xmm1 ;\
+	vfmaddpd %xmm11,%xmm3, %xmm0,%xmm11 ;\
+	vfmaddpd %xmm15,%xmm2, %xmm3,%xmm15 ;\
+ 	vmovddup	 -7 * SIZE(BO, %rax, 4), %xmm3 ;\
+	vmovaps	%xmm4, %xmm2
+
+#define KERNEL3(xx) \
+	vfmaddpd %xmm8,%xmm5, %xmm4, %xmm8 ;\
+	vmovaps	%xmm2, %xmm4 ;\
+	vmovups -6 * SIZE(AO, %rax, 4),%xmm2 ;\
+	vfmaddpd %xmm12,%xmm2, %xmm5,%xmm12 ;\
+	vfmaddpd %xmm9,%xmm3, %xmm4,%xmm9 ;\
+	vmovddup	 -6 * SIZE(BO, %rax, 4), %xmm5 ;\
+	vfmaddpd %xmm13,%xmm2, %xmm3,%xmm13 ;\
+	vmovddup	 -5 * SIZE(BO, %rax, 4), %xmm3 ;\
+	vfmaddpd %xmm10,%xmm5, %xmm4,%xmm10 ;\
+	vfmaddpd %xmm14,%xmm2, %xmm5,%xmm14 ;\
+	vfmaddpd %xmm11,%xmm3, %xmm4, %xmm11 ;\
+	vmovups	 -4 * SIZE(AO, %rax, 4), %xmm4 ;\
+	vmovddup	 -4 * SIZE(BO, %rax, 4), %xmm5 ;\
+	vfmaddpd %xmm15,%xmm2,%xmm3,%xmm15 ;\
+ 	vmovddup	 -3 * SIZE(BO, %rax, 4), %xmm3 ;\
+	vmovaps	%xmm4, %xmm2
+
+#define KERNEL4(xx) \
+	vfmaddpd %xmm8,%xmm5, %xmm4,%xmm8 ;\
+	vmovaps	%xmm2, %xmm4 ;\
+	vmovups -2 * SIZE(AO, %rax, 4),%xmm2 ;\
+	vfmaddpd %xmm12,%xmm2, %xmm5 ,%xmm12;\
+/*A*/	vmovups	  8 * SIZE(AO, %rax, 4), %xmm7 ;\
+	vmovddup	 -2 * SIZE(BO, %rax, 4), %xmm5 ;\
+	vfmaddpd %xmm9,%xmm3, %xmm4,%xmm9 ;\
+	vfmaddpd %xmm13,%xmm2, %xmm3,%xmm13 ;\
+	vmovddup	 -1 * SIZE(BO, %rax, 4), %xmm3 ;\
+	vfmaddpd %xmm10,%xmm5, %xmm4,%xmm10 ;\
+	vfmaddpd %xmm14,%xmm2, %xmm5,%xmm14 ;\
+/**/	vmovddup	  8 * SIZE(BO, %rax, 4), %xmm5 ;\
+	vfmaddpd %xmm11,%xmm3, %xmm4,%xmm11 ;\
+	vfmaddpd %xmm15,%xmm2, %xmm3,%xmm15 ;\
+ 	vmovddup	  1 * SIZE(BO, %rax, 4), %xmm3 ;\
+	vmovaps	%xmm6, %xmm2
+
+#define KERNEL5(xx) \
+	vfmaddpd %xmm8,%xmm1, %xmm6,%xmm8 ;\
+	vmovaps	%xmm2, %xmm6 ;\
+	vmovups 2 * SIZE(AO, %rax, 4),%xmm2 ;\
+	vfmaddpd %xmm12,%xmm2, %xmm1,%xmm12 ;\
+	vmovddup	  2 * SIZE(BO, %rax, 4), %xmm1 ;\
+	vfmaddpd %xmm9,%xmm3, %xmm6,%xmm9 ;\
+	vfmaddpd %xmm13,%xmm2, %xmm3,%xmm13 ;\
+	vmovddup	  3 * SIZE(BO, %rax, 4), %xmm3 ;\
+	vfmaddpd %xmm10,%xmm1, %xmm6,%xmm10 ;\
+	vfmaddpd %xmm14,%xmm2, %xmm1,%xmm14 ;\
+	vfmaddpd %xmm11,%xmm3, %xmm6,%xmm11 ;\
+	vmovups	  4 * SIZE(AO, %rax, 4), %xmm6 ;\
+	vmovddup	  4 * SIZE(BO, %rax, 4), %xmm1 ;\
+	vfmaddpd %xmm15,%xmm2, %xmm3,%xmm15 ;\
+ 	vmovddup	  5 * SIZE(BO, %rax, 4), %xmm3 ;\
+	vmovaps	%xmm6, %xmm2
+
+#define KERNEL6(xx) \
+	vfmaddpd %xmm8,%xmm1, %xmm6,%xmm8 ;\
+	vmovaps	%xmm2, %xmm6 ;\
+	vmovups	6 * SIZE(AO, %rax, 4),%xmm2 ;\
+	vfmaddpd %xmm12,%xmm2, %xmm1,%xmm12 ;\
+/*A*/	vmovups	 16 * SIZE(AO, %rax, 4), %xmm0 ;\
+	vmovddup	  6 * SIZE(BO, %rax, 4), %xmm1 ;\
+	vfmaddpd %xmm9,%xmm3, %xmm6,%xmm9 ;\
+	vfmaddpd %xmm13,%xmm2, %xmm3,%xmm13 ;\
+	vmovddup	  7 * SIZE(BO, %rax, 4), %xmm3 ;\
+	vfmaddpd %xmm10,%xmm1, %xmm6,%xmm10 ;\
+	vfmaddpd %xmm14,%xmm2, %xmm1,%xmm14 ;\
+/**/	vmovddup	 16 * SIZE(BO, %rax, 4), %xmm1 ;\
+	vfmaddpd %xmm11,%xmm3, %xmm6,%xmm11 ;\
+	vfmaddpd %xmm15,%xmm2,%xmm3,%xmm15 ;\
+ 	vmovddup	  9 * SIZE(BO, %rax, 4), %xmm3 ;\
+	vmovaps	%xmm7, %xmm2
+
+#define KERNEL7(xx) \
+	vfmaddpd %xmm8,%xmm5, %xmm7,%xmm8 ;\
+	vmovaps	%xmm2, %xmm7 ;\
+	vmovups 10 * SIZE(AO, %rax, 4),%xmm2 ;\
+	vfmaddpd %xmm12,%xmm2, %xmm5,%xmm12 ;\
+	vmovddup	 10 * SIZE(BO, %rax, 4), %xmm5 ;\
+	vfmaddpd %xmm9,%xmm3, %xmm7,%xmm9 ;\
+	vfmaddpd %xmm13,%xmm2, %xmm3,%xmm13 ;\
+	vmovddup	 11 * SIZE(BO, %rax, 4), %xmm3 ;\
+	vfmaddpd %xmm10,%xmm5, %xmm7,%xmm10 ;\
+	vfmaddpd %xmm14,%xmm2, %xmm5,%xmm14 ;\
+	vfmaddpd %xmm11,%xmm3, %xmm7,%xmm11 ;\
+	vmovups	 12 * SIZE(AO, %rax, 4), %xmm7 ;\
+	vmovddup	 12 * SIZE(BO, %rax, 4), %xmm5 ;\
+	vfmaddpd %xmm15,%xmm2, %xmm3,%xmm15 ;\
+ 	vmovddup	 13 * SIZE(BO, %rax, 4), %xmm3 ;\
+	vmovaps	%xmm7, %xmm2
+
+#define KERNEL8(xx) \
+	vfmaddpd %xmm8,%xmm5, %xmm7,%xmm8 ;\
+	vmovaps	%xmm2, %xmm7 ;\
+	vmovups 14 * SIZE(AO, %rax, 4),%xmm2 ;\
+/*A*/	vmovups	 24 * SIZE(AO, %rax, 4), %xmm4 ;\
+	vfmaddpd %xmm12,%xmm2, %xmm5,%xmm12 ;\
+	vfmaddpd %xmm9,%xmm3, %xmm7,%xmm9 ;\
+	vmovddup	 14 * SIZE(BO, %rax, 4), %xmm5 ;\
+	vfmaddpd %xmm13, %xmm2, %xmm3,%xmm13 ;\
+	vmovddup	 15 * SIZE(BO, %rax, 4), %xmm3 ;\
+	vfmaddpd %xmm10,%xmm5, %xmm7,%xmm10 ;\
+	vfmaddpd %xmm14,%xmm2, %xmm5,%xmm14 ;\
+/**/	vmovddup	 24 * SIZE(BO, %rax, 4), %xmm5 ;\
+	vfmaddpd %xmm11,%xmm3, %xmm7,%xmm11 ;\
+	vfmaddpd %xmm15,%xmm2, %xmm3,%xmm15 ;\
+ 	vmovddup	 17 * SIZE(BO, %rax, 4), %xmm3 ;\
+	vmovaps	%xmm0, %xmm2 ;\
+	addq	$8 * SIZE, %rax ;\
+
+#define KERNEL_SUB1(xx) \
+        vfmaddpd   %xmm8, %xmm1, %xmm0,%xmm8 ;\
+        vmovapd  %xmm2, %xmm0 ;\
+	vmovups  -14 * SIZE(AO),%xmm2 ;\
+        vfmaddpd   %xmm12, %xmm2, %xmm1,%xmm12 ;\
+        vmovddup -14 * SIZE(BO), %xmm1 ;\
+        vfmaddpd   %xmm9, %xmm3, %xmm0,%xmm9 ;\
+        vfmaddpd   %xmm13, %xmm2, %xmm3,%xmm13 ;\
+        vmovddup -13 * SIZE(BO), %xmm3 ;\
+        vfmaddpd   %xmm10, %xmm1, %xmm0,%xmm10 ;\
+        vfmaddpd   %xmm14,%xmm2, %xmm1 ,%xmm14 ;\
+        vfmaddpd   %xmm11, %xmm3, %xmm0,%xmm11 ;\
+        vfmaddpd   %xmm15, %xmm2, %xmm3,%xmm15 ;\
+        vmovups  -12 * SIZE(AO), %xmm0 ;\
+        vmovddup -12 * SIZE(BO), %xmm1 ;\
+        vmovddup -11 * SIZE(BO), %xmm3 ;\
+        vmovapd  %xmm0, %xmm2
+
+
+#define KERNEL_SUB2(xx) \
+	vfmaddpd %xmm8,%xmm1, %xmm0,%xmm8 ;\
+	vmovaps	%xmm2, %xmm0 ;\
+	vmovups -10 * SIZE(AO),%xmm2 ;\
+	vfmaddpd %xmm12,%xmm2, %xmm1,%xmm12 ;\
+	vfmaddpd %xmm9,%xmm3, %xmm0,%xmm9 ;\
+	vfmaddpd %xmm13,%xmm2, %xmm3,%xmm13 ;\
+	vmovddup	-10 * SIZE(BO), %xmm1 ;\
+	vmovddup	 -9 * SIZE(BO), %xmm3 ;\
+	vfmaddpd %xmm10,%xmm1, %xmm0,%xmm10 ;\
+	vfmaddpd %xmm14,%xmm2, %xmm1,%xmm14 ;\
+	vfmaddpd %xmm11,%xmm3, %xmm0,%xmm11 ;\
+	vfmaddpd %xmm15,%xmm2, %xmm3,%xmm15 ;\
+	vmovups	  (AO), %xmm0 ;\
+	vmovddup	  (BO), %xmm1 ;\
+ 	vmovddup	 -7 * SIZE(BO), %xmm3 ;\
+	vmovaps	%xmm4, %xmm2
+
+#define KERNEL_SUB3(xx) \
+	vfmaddpd %xmm8,%xmm5, %xmm4,%xmm8 ;\
+	vmovaps	%xmm2, %xmm4 ;\
+	vmovups -6 * SIZE(AO),%xmm2 ;\
+	vfmaddpd %xmm12,%xmm2, %xmm5,%xmm12 ;\
+	vfmaddpd %xmm9,%xmm3, %xmm4,%xmm9 ;\
+	vmovddup	 -6 * SIZE(BO), %xmm5 ;\
+	vfmaddpd %xmm13,%xmm2, %xmm3,%xmm13 ;\
+	vmovddup	 -5 * SIZE(BO), %xmm3 ;\
+	vfmaddpd %xmm10,%xmm5, %xmm4,%xmm10 ;\
+	vfmaddpd %xmm14,%xmm2, %xmm5,%xmm14 ;\
+	vfmaddpd %xmm11,%xmm3, %xmm4,%xmm11 ;\
+	vmovups	 -4 * SIZE(AO), %xmm4 ;\
+	vmovddup	 -4 * SIZE(BO), %xmm5 ;\
+	vfmaddpd %xmm15,%xmm2, %xmm3,%xmm15 ;\
+ 	vmovddup	 -3 * SIZE(BO), %xmm3 ;\
+	vmovaps	%xmm4, %xmm2
+
+#define KERNEL_SUB4(xx) \
+	vfmaddpd %xmm8,%xmm5, %xmm4,%xmm8 ;\
+	vmovaps	%xmm2, %xmm4 ;\
+	vmovups -2 * SIZE(AO),%xmm2 ;\
+	vfmaddpd %xmm12,%xmm2, %xmm5,%xmm12 ;\
+	vfmaddpd %xmm13,%xmm2, %xmm3,%xmm13 ;\
+	vmovddup	 -2 * SIZE(BO), %xmm5 ;\
+	vfmaddpd %xmm9,%xmm3, %xmm4,%xmm9 ;\
+	vmovddup	 -1 * SIZE(BO), %xmm3 ;\
+	vfmaddpd %xmm10,%xmm5, %xmm4,%xmm10 ;\
+	vfmaddpd %xmm14,%xmm2, %xmm5,%xmm14 ;\
+	vfmaddpd %xmm11,%xmm3, %xmm4,%xmm11 ;\
+	vfmaddpd %xmm15,%xmm2, %xmm3,%xmm15 ;\
+ 	vmovddup	  1 * SIZE(BO), %xmm3 ;\
+	vmovaps	%xmm0, %xmm2
+
+	PROLOGUE
+	PROFCODE
+	
+	subq	$STACKSIZE, %rsp
+	movq	%rbx,   (%rsp)
+	movq	%rbp,  8(%rsp)
+	movq	%r12, 16(%rsp)
+	movq	%r13, 24(%rsp)
+	movq	%r14, 32(%rsp)
+	movq	%r15, 40(%rsp)
+
+	vzeroupper
+
+#ifdef WINDOWS_ABI
+	movq	%rdi,    48(%rsp)
+	movq	%rsi,    56(%rsp)
+	movups	%xmm6,   64(%rsp)
+	movups	%xmm7,   80(%rsp)
+	movups	%xmm8,   96(%rsp)
+	movups	%xmm9,  112(%rsp)
+	movups	%xmm10, 128(%rsp)
+	movups	%xmm11, 144(%rsp)
+	movups	%xmm12, 160(%rsp)
+	movups	%xmm13, 176(%rsp)
+	movups	%xmm14, 192(%rsp)
+	movups	%xmm15, 208(%rsp)
+
+	movq	ARG1,      OLD_M
+	movq	ARG2,      OLD_N
+	movq	ARG3,      K
+	movq	OLD_A,     A
+	movq	OLD_B,     B
+	movq	OLD_C,     C
+	movq	OLD_LDC,   LDC
+#ifdef TRMMKERNEL
+	movsd	OLD_OFFSET, %xmm12
+#endif
+	vmovaps	%xmm3, %xmm0
+
+#else
+	movq	STACKSIZE +  8(%rsp), LDC
+#ifdef TRMMKERNEL
+	movsd	STACKSIZE + 16(%rsp), %xmm12
+#endif
+
+#endif
+
+	movq	OLD_M, M
+	movq	OLD_N, N
+
+	subq	$-16 * SIZE, A
+	subq	$-16 * SIZE, B
+
+	vmovsd	 %xmm0, ALPHA
+
+	salq	$BASE_SHIFT, LDC
+
+#ifdef TRMMKERNEL
+	vmovsd	%xmm12, OFFSET
+	vmovsd	%xmm12, KK
+#ifndef LEFT
+	negq	KK
+#endif	
+#endif
+	movq	N,  J
+	sarq	$2, J		# j = (n >> 2)
+	jle	.L40
+	ALIGN_4
+
+.L01:
+	movq	C, CO1			# coffset1 = c
+	leaq	(C, LDC, 2), CO2	# coffset2 = c + ldc
+	
+	leaq	(C, LDC, 4), C		# c += 4 * ldc
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+	movq	OFFSET, %rax
+	movq	%rax, KK
+#endif	
+
+	movq	A, AO		# aoffset = a
+
+	movq	K, %rax
+	salq	$BASE_SHIFT + 2, %rax
+	leaq	 (B, %rax), BB
+
+	movq	M,  I
+	sarq	$2, I	# i = (m >> 2)
+	jle	.L20
+	ALIGN_4
+
+	.align 16
+.L11:
+#if !defined(TRMMKERNEL) || \
+	(defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+	(defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+	movq	B, BO
+#else
+	movq	KK, %rax
+	leaq	(, %rax, SIZE), %rax
+	leaq	(AO, %rax, 4), AO
+	leaq	(B,  %rax, 4), BO
+#endif	
+
+	vzeroall
+	prefetcht0 256(CO1)
+	prefetcht0 320(CO1)
+	prefetcht0 256(CO2)
+	prefetcht0 320(CO2)
+	vmovups	-16 * SIZE(AO), %xmm0
+	vmovddup	-16 * SIZE(BO), %xmm1
+ 	vmovddup	-15 * SIZE(BO), %xmm3
+	vmovups	 -8 * SIZE(AO), %xmm4
+	vmovddup	 -8 * SIZE(BO), %xmm5
+
+	vmovaps	%xmm0, %xmm2
+
+
+#ifndef TRMMKERNEL
+	movq	K, %rax
+#elif (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+	movq	K, %rax
+	subq	KK, %rax
+	movq	%rax, KKK	
+#else
+	movq	KK, %rax
+#ifdef LEFT
+	addq	$4, %rax
+#else
+	addq	$4, %rax
+#endif
+	movq	%rax, KKK
+#endif
+
+	andq	$-8, %rax
+	salq	$BASE_SHIFT, %rax
+	leaq	(AO, %rax, 4), AO
+	leaq	(BO, %rax, 4), BO
+	negq	%rax
+	NOBRANCH
+	je	.L15
+	ALIGN_4
+
+	.align 16
+.L12:
+	prefetcht0 (AO,%rax,4)
+	prefetcht0 (BO,%rax,4)
+	KERNEL1(16 *  0)
+	KERNEL2(16 *  0)
+	KERNEL3(16 *  0)
+	KERNEL4(16 *  0)
+	KERNEL5(16 *  0)
+	KERNEL6(16 *  0)
+	KERNEL7(16 *  0)
+	KERNEL8(16 *  0)
+	NOBRANCH
+	je	.L15
+	KERNEL1(16 *  0)
+	KERNEL2(16 *  0)
+	KERNEL3(16 *  0)
+	KERNEL4(16 *  0)
+	KERNEL5(16 *  0)
+	KERNEL6(16 *  0)
+	KERNEL7(16 *  0)
+	KERNEL8(16 *  0)
+	NOBRANCH
+	je	.L15
+	KERNEL1(16 *  0)
+	KERNEL2(16 *  0)
+	KERNEL3(16 *  0)
+	KERNEL4(16 *  0)
+	KERNEL5(16 *  0)
+	KERNEL6(16 *  0)
+	KERNEL7(16 *  0)
+	KERNEL8(16 *  0)
+	NOBRANCH
+	je	.L15
+	KERNEL1(16 *  0)
+	KERNEL2(16 *  0)
+	KERNEL3(16 *  0)
+	KERNEL4(16 *  0)
+	KERNEL5(16 *  0)
+	KERNEL6(16 *  0)
+	KERNEL7(16 *  0)
+	KERNEL8(16 *  0)
+	NOBRANCH
+	je	.L15
+	KERNEL1(16 *  0)
+	KERNEL2(16 *  0)
+	KERNEL3(16 *  0)
+	KERNEL4(16 *  0)
+	KERNEL5(16 *  0)
+	KERNEL6(16 *  0)
+	KERNEL7(16 *  0)
+	KERNEL8(16 *  0)
+	NOBRANCH
+	je	.L15
+	KERNEL1(16 *  0)
+	KERNEL2(16 *  0)
+	KERNEL3(16 *  0)
+	KERNEL4(16 *  0)
+	KERNEL5(16 *  0)
+	KERNEL6(16 *  0)
+	KERNEL7(16 *  0)
+	KERNEL8(16 *  0)
+	NOBRANCH
+	je	.L15
+	KERNEL1(16 *  0)
+	KERNEL2(16 *  0)
+	KERNEL3(16 *  0)
+	KERNEL4(16 *  0)
+	KERNEL5(16 *  0)
+	KERNEL6(16 *  0)
+	KERNEL7(16 *  0)
+	KERNEL8(16 *  0)
+	NOBRANCH
+	je	.L15
+	KERNEL1(16 *  0)
+	KERNEL2(16 *  0)
+	KERNEL3(16 *  0)
+	KERNEL4(16 *  0)
+	KERNEL5(16 *  0)
+	KERNEL6(16 *  0)
+	KERNEL7(16 *  0)
+	KERNEL8(16 *  0)
+	jl	.L12
+	ALIGN_4
+
+.L15:
+	vmovddup	ALPHA, %xmm7
+
+#ifndef TRMMKERNEL
+	movq	K, %rax
+#else
+	movq	KKK, %rax
+#endif
+	testq	$4, %rax
+	je .L16
+	ALIGN_4
+
+	KERNEL_SUB1(16 *  0)
+	KERNEL_SUB2(16 *  0)
+	KERNEL_SUB3(16 *  0)
+	KERNEL_SUB4(16 *  0)
+
+	subq	$-16 * SIZE, BO
+	subq	$-16 * SIZE, AO
+	ALIGN_4
+
+.L16:
+#ifndef TRMMKERNEL
+	movq	K, %rax
+#else
+	movq	KKK, %rax
+#endif
+	andq	$3, %rax		# if (k & 1)
+	je .L19
+
+	leaq	(, %rax, SIZE), %rax
+	leaq	(AO, %rax, 4), AO
+	leaq	(BO, %rax, 4), BO
+	negq	%rax
+	ALIGN_4
+
+.L17:
+	vfmaddpd %xmm8,%xmm1, %xmm0,%xmm8
+	vmovaps	%xmm2, %xmm0
+	vmovups -14 * SIZE(AO, %rax, 4),%xmm2
+	vfmaddpd %xmm12,%xmm2, %xmm1,%xmm12
+	vmovddup	-14 * SIZE(BO, %rax, 4), %xmm1
+	vfmaddpd %xmm9,%xmm3, %xmm0,%xmm9
+	vfmaddpd %xmm13,%xmm2, %xmm3,%xmm13
+	vmovddup	-13 * SIZE(BO, %rax, 4), %xmm3
+	vfmaddpd %xmm10,%xmm1, %xmm0,%xmm10
+	vfmaddpd %xmm14,%xmm2, %xmm1,%xmm14
+	vfmaddpd %xmm11,%xmm3, %xmm0,%xmm11
+	vmovups	-12 * SIZE(AO, %rax, 4), %xmm0
+	vmovddup	-12 * SIZE(BO, %rax, 4), %xmm1
+	vfmaddpd %xmm15,%xmm2, %xmm3,%xmm15
+ 	vmovddup	-11 * SIZE(BO, %rax, 4), %xmm3
+	vmovaps	%xmm0, %xmm2
+
+	addq	$SIZE, %rax
+	jl	.L17
+	ALIGN_4
+
+.L19:
+	// prefetch	   -8 * SIZE(BB)
+	subq		 $-16 * SIZE, BB
+
+#ifndef TRMMKERNEL
+
+	vfmaddpd (CO1),%xmm7, %xmm8,%xmm8
+	vfmaddpd 2 * SIZE(CO1),%xmm7, %xmm12,%xmm12
+	.align 2
+	vfmaddpd (CO1, LDC),%xmm7, %xmm9,%xmm9
+	vfmaddpd 2 * SIZE(CO1, LDC),%xmm7, %xmm13,%xmm13
+	.align 2
+	vfmaddpd (CO2),%xmm7, %xmm10,%xmm10
+	vfmaddpd 2 * SIZE(CO2),%xmm7, %xmm14,%xmm14
+	.align 2
+	vfmaddpd (CO2, LDC),%xmm7, %xmm11,%xmm11
+	vfmaddpd 2 * SIZE(CO2, LDC),%xmm7, %xmm15,%xmm15
+
+#else
+	vmulpd	%xmm7, %xmm8,%xmm8
+	vmulpd	%xmm7, %xmm12,%xmm12
+	vmulpd	%xmm7, %xmm9,%xmm9
+	vmulpd	%xmm7, %xmm13,%xmm13
+	vmulpd	%xmm7, %xmm10,%xmm10
+	vmulpd	%xmm7, %xmm14,%xmm14
+	vmulpd	%xmm7, %xmm11,%xmm11
+	vmulpd	%xmm7, %xmm15,%xmm15
+
+#endif
+
+	.align 2
+	vmovups	%xmm8,  (CO1)
+	vmovups	%xmm12, 2 * SIZE(CO1)
+	.align 2
+	vmovups	%xmm9,  (CO1, LDC)
+	vmovups	%xmm13, 2 * SIZE(CO1, LDC)
+	.align 2
+	vmovups	%xmm10, (CO2)
+	vmovups	%xmm14, 2 * SIZE(CO2)
+	.align 2
+	vmovups	%xmm11, (CO2, LDC)
+	vmovups	%xmm15, 2 * SIZE(CO2, LDC)
+
+#if (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+    (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+	movq	K, %rax
+	subq	KKK, %rax
+	leaq	(,%rax, SIZE), %rax
+	leaq	(AO, %rax, 4), AO
+	leaq	(BO, %rax, 4), BO
+#endif
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+	addq	$4, KK
+#endif
+
+	addq	$4 * SIZE, CO1		# coffset += 4
+	addq	$4 * SIZE, CO2		# coffset += 4
+	decq	I			# i --
+	BRANCH
+	jg	.L11
+	ALIGN_4	
+
+.L20:
+	testq	$3, M
+	je	.L39
+
+	testq	$2, M
+	je	.L30
+	ALIGN_4
+
+.L21:
+#if !defined(TRMMKERNEL) || \
+	(defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+	(defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+	movq	B, BO
+#else
+	movq	KK, %rax
+	leaq	(, %rax, SIZE), %rax
+	leaq	(AO, %rax, 2), AO
+	leaq	(B,  %rax, 4), BO
+#endif	
+
+	vmovups	-16 * SIZE(AO), %xmm0
+	vxorps	%xmm8, %xmm8, %xmm8
+	vmovups	-12 * SIZE(AO), %xmm2
+	vxorps	%xmm9, %xmm9 ,%xmm9
+	vmovddup	-16 * SIZE(BO), %xmm1
+	vxorps	%xmm10, %xmm10, %xmm10
+	vmovddup	-15 * SIZE(BO), %xmm5
+	vxorps	%xmm11, %xmm11, %xmm11
+	vmovddup	 -8 * SIZE(BO), %xmm3
+
+#ifndef TRMMKERNEL
+	movq	K, %rax
+#elif (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+	movq	K, %rax
+	subq	KK, %rax
+	movq	%rax, KKK	
+#else
+	movq	KK, %rax
+#ifdef LEFT
+	addq	$2, %rax
+#else
+	addq	$4, %rax
+#endif
+	movq	%rax, KKK
+#endif
+
+	andq	$-4, %rax
+	leaq	(, %rax, SIZE), %rax
+	leaq	(AO, %rax, 2), AO
+	leaq	(BO, %rax, 4), BO
+	negq	%rax
+	NOBRANCH
+	je	.L26
+	ALIGN_4
+
+.L22:
+	vfmaddpd %xmm8,%xmm0, %xmm1,%xmm8
+	vfmaddpd %xmm9,%xmm0, %xmm5,%xmm9
+	vmovddup	-14 * SIZE(BO, %rax, 4), %xmm1
+	vfmaddpd %xmm10,%xmm0, %xmm1,%xmm10
+	vmovddup	-13 * SIZE(BO, %rax, 4), %xmm5
+	vfmaddpd %xmm11,%xmm0, %xmm5,%xmm11
+	vmovups	-14 * SIZE(AO, %rax, 2), %xmm0
+	vmovddup	-12 * SIZE(BO, %rax, 4), %xmm1
+	vmovddup	-11 * SIZE(BO, %rax, 4), %xmm5
+	vfmaddpd %xmm8,%xmm0, %xmm1,%xmm8
+	vfmaddpd %xmm9,%xmm0, %xmm5,%xmm9
+	vmovddup	-10 * SIZE(BO, %rax, 4), %xmm1
+	vmovddup	 -9 * SIZE(BO, %rax, 4), %xmm5
+	vfmaddpd %xmm10,%xmm0, %xmm1,%xmm10
+	vfmaddpd %xmm11,%xmm0, %xmm5,%xmm11
+	vmovddup	  (BO, %rax, 4), %xmm1
+	vmovddup	 -7 * SIZE(BO, %rax, 4), %xmm5
+	vmovups	 -8 * SIZE(AO, %rax, 2), %xmm0
+	vfmaddpd %xmm8,%xmm2, %xmm3,%xmm8
+	vfmaddpd %xmm9,%xmm2, %xmm5,%xmm9
+	vmovddup	 -6 * SIZE(BO, %rax, 4), %xmm3
+	vmovddup	 -5 * SIZE(BO, %rax, 4), %xmm5
+	vfmaddpd %xmm10,%xmm2, %xmm3,%xmm10
+	vfmaddpd %xmm11,%xmm2, %xmm5,%xmm11
+	vmovups	-10 * SIZE(AO, %rax, 2), %xmm2
+	vmovddup	 -4 * SIZE(BO, %rax, 4), %xmm3
+	vmovddup	 -3 * SIZE(BO, %rax, 4), %xmm5
+	vfmaddpd %xmm8,%xmm2, %xmm3,%xmm8
+	vfmaddpd %xmm9,%xmm2, %xmm5,%xmm9
+	vmovddup	 -2 * SIZE(BO, %rax, 4), %xmm3
+	vmovddup	 -1 * SIZE(BO, %rax, 4), %xmm5
+	vfmaddpd %xmm10,%xmm2, %xmm3,%xmm10
+	vfmaddpd %xmm11,%xmm2, %xmm5,%xmm11
+	vmovddup	  8 * SIZE(BO, %rax, 4), %xmm3
+	vmovups	 -4 * SIZE(AO, %rax, 2), %xmm2
+	vmovddup	  1 * SIZE(BO, %rax, 4), %xmm5
+
+	addq	$4 * SIZE, %rax
+	BRANCH
+	jl	.L22
+	ALIGN_4
+
+.L26:
+	vmovddup	ALPHA, %xmm7
+
+#ifndef TRMMKERNEL
+	movq	K, %rax
+#else
+	movq	KKK, %rax
+#endif
+	andq	$3, %rax		# if (k & 1)
+	je .L29
+
+	leaq	(, %rax, SIZE), %rax
+	leaq	(AO, %rax, 2), AO
+	leaq	(BO, %rax, 4), BO
+	negq	%rax
+	ALIGN_4
+
+.L27:
+	vfmaddpd %xmm8,%xmm0, %xmm1,%xmm8
+	vmovddup	-14 * SIZE(BO, %rax, 4), %xmm1
+	vfmaddpd %xmm9,%xmm0, %xmm5,%xmm9
+	vmovddup	-13 * SIZE(BO, %rax, 4), %xmm5
+	vfmaddpd %xmm10,%xmm0, %xmm1,%xmm10
+	vmovddup	-12 * SIZE(BO, %rax, 4), %xmm1
+	vfmaddpd %xmm11,%xmm0, %xmm5,%xmm11
+	vmovups	-14 * SIZE(AO, %rax, 2), %xmm0
+	vmovddup	-11 * SIZE(BO, %rax, 4), %xmm5
+
+	addq	$SIZE, %rax
+	jl	.L27
+	ALIGN_4
+
+.L29:
+#ifndef TRMMKERNEL
+
+	vfmaddpd (CO1),%xmm7, %xmm8,%xmm8
+	vfmaddpd (CO1, LDC),%xmm7, %xmm9,%xmm9
+	vfmaddpd (CO2),%xmm7, %xmm10,%xmm10
+	vfmaddpd (CO2, LDC),%xmm7, %xmm11,%xmm11
+
+#else
+	vmulpd	%xmm7, %xmm8,%xmm8
+	vmulpd	%xmm7, %xmm9,%xmm9
+	vmulpd	%xmm7, %xmm10,%xmm10
+	vmulpd	%xmm7, %xmm11,%xmm11
+
+#endif
+
+	vmovups	%xmm8,  (CO1)
+	vmovups	%xmm9,  (CO1, LDC)
+
+	vmovups	%xmm10, (CO2)
+	vmovups	%xmm11, (CO2, LDC)
+
+#if (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+    (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+	movq	K, %rax
+	subq	KKK, %rax
+	leaq	(,%rax, SIZE), %rax
+	leaq	(AO, %rax, 2), AO
+	leaq	(BO, %rax, 4), BO
+#endif
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+	addq	$2, KK
+#endif
+
+	addq	$2 * SIZE, CO1
+	addq	$2 * SIZE, CO2
+	ALIGN_4	
+
+.L30:
+	testq	$1, M
+	je	.L39
+
+#if !defined(TRMMKERNEL) || \
+	(defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+	(defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+	movq	B, BO
+#else
+	movq	KK, %rax
+	leaq	(, %rax, SIZE), %rax
+	leaq	(AO, %rax, 1), AO
+	leaq	(B,  %rax, 4), BO
+#endif	
+
+	vmovddup	-16 * SIZE(AO), %xmm0
+	vxorps	%xmm8, %xmm8, %xmm8
+	vmovddup	-14 * SIZE(AO), %xmm2
+	vxorps	%xmm9, %xmm9, %xmm9
+	vmovddup	-15 * SIZE(AO), %xmm4
+	vxorps	%xmm10, %xmm10,%xmm10
+	vmovups	-16 * SIZE(BO), %xmm1
+	vxorps	%xmm11, %xmm11,%xmm11
+	vmovups	 -8 * SIZE(BO), %xmm3
+
+#ifndef TRMMKERNEL
+	movq	K, %rax
+#elif (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+	movq	K, %rax
+	subq	KK, %rax
+	movq	%rax, KKK	
+#else
+	movq	KK, %rax
+#ifdef LEFT
+	addq	$1, %rax
+#else
+	addq	$4, %rax
+#endif
+	movq	%rax, KKK
+#endif
+
+	andq	$-4, %rax
+	leaq	(, %rax, SIZE), %rax
+	leaq	(AO, %rax, 1), AO
+	leaq	(BO, %rax, 4), BO
+	negq	%rax
+	NOBRANCH
+	je	.L36
+	ALIGN_4
+
+.L32:
+	vfmaddpd %xmm8,%xmm0, %xmm1,%xmm8
+	vfmaddpd %xmm9,-14 * SIZE(BO, %rax, 4), %xmm0,%xmm9
+	vmovups	-12 * SIZE(BO, %rax, 4), %xmm1
+	vmovddup -12 * SIZE(AO, %rax, 1), %xmm0
+	vfmaddpd %xmm10,%xmm4, %xmm1,%xmm10
+	vfmaddpd %xmm11,-10 * SIZE(BO, %rax, 4), %xmm4,%xmm11
+	vmovups	  (BO, %rax, 4), %xmm1
+	vmovddup	-11 * SIZE(AO, %rax, 1), %xmm4
+	vfmaddpd %xmm8,%xmm2, %xmm3,%xmm8
+	vfmaddpd %xmm9,-6 * SIZE(BO, %rax, 4), %xmm2,%xmm9
+	vmovups	 -4 * SIZE(BO, %rax, 4), %xmm3
+	vmovddup	-13 * SIZE(AO, %rax, 1), %xmm2
+	vfmaddpd %xmm10,%xmm2, %xmm3,%xmm10
+	vfmaddpd %xmm11,-2 * SIZE(BO, %rax, 4), %xmm2,%xmm11
+	vmovups	  8 * SIZE(BO, %rax, 4), %xmm3
+	vmovddup	-10 * SIZE(AO, %rax, 1), %xmm2
+
+	addq	$4 * SIZE, %rax
+	BRANCH
+	jl	.L32
+	ALIGN_4
+
+.L36:
+	vmovddup	ALPHA, %xmm7
+
+#ifndef TRMMKERNEL
+	movq	K, %rax
+#else
+	movq	KKK, %rax
+#endif
+	andq	$3, %rax		# if (k & 1)
+	je .L38
+
+	leaq	(, %rax, SIZE), %rax
+	leaq	(AO, %rax, 1), AO
+	leaq	(BO, %rax, 4), BO
+	negq	%rax
+	ALIGN_4
+
+.L37:
+	vfmaddpd %xmm8,%xmm0, %xmm1,%xmm8
+	vfmaddpd %xmm9,-14 * SIZE(BO, %rax, 4), %xmm0,%xmm9
+	vmovups	-12 * SIZE(BO, %rax, 4), %xmm1
+	vmovddup -15 * SIZE(AO, %rax, 1), %xmm0
+
+	addq	$SIZE, %rax
+	jl	.L37
+	ALIGN_4
+
+.L38:
+	vaddpd	%xmm10, %xmm8,%xmm8
+	vaddpd	%xmm11, %xmm9,%xmm9
+
+#ifndef TRMMKERNEL
+	vmovsd	(CO1), %xmm0
+	vmovhpd	(CO1, LDC), %xmm0,%xmm0
+	vmovsd	(CO2), %xmm1
+	vmovhpd	(CO2, LDC), %xmm1,%xmm1
+
+
+	vfmaddpd	%xmm0, %xmm7,%xmm8,%xmm8
+	vfmaddpd	%xmm1, %xmm7,%xmm9,%xmm9
+#else
+
+	vmulpd	%xmm7, %xmm8,%xmm8
+	vmulpd	%xmm7, %xmm9,%xmm9
+
+#endif
+
+	vmovsd	%xmm8,  (CO1)
+	vmovhpd	%xmm8,  (CO1, LDC)
+	vmovsd	%xmm9,  (CO2)
+	vmovhpd	%xmm9,  (CO2, LDC)
+
+#if (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+    (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+	movq	K, %rax
+	subq	KKK, %rax
+	leaq	(,%rax, SIZE), %rax
+	leaq	(AO, %rax, 1), AO
+	leaq	(BO, %rax, 4), BO
+#endif
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+	addq	$1, KK
+#endif
+	ALIGN_4	
+	
+.L39:
+#if defined(TRMMKERNEL) && !defined(LEFT)
+	addq	$4, KK
+#endif
+	
+	movq	BO, B
+
+	decq	J			# j --
+	jg	.L01
+	ALIGN_4
+
+.L40:
+	testq	$3, N
+	je	.L999
+
+	testq	$2, N
+	je	.L80
+	ALIGN_4
+
+.L41:
+#if defined(TRMMKERNEL) && defined(LEFT)
+	movq	OFFSET, %rax
+	movq	%rax, KK
+#endif	
+
+	movq	C, CO1			# coffset1 = c
+	leaq	(C, LDC, 1), CO2	# coffset2 = c + ldc
+	movq	A, AO		# aoffset = a
+
+	movq	K, %rax
+	salq	$BASE_SHIFT + 1, %rax
+	leaq	(B, %rax), BB
+
+	movq	M,  I
+	sarq	$2, I	# i = (m >> 2)
+	jle	.L60
+	ALIGN_4
+
+.L51:
+#if !defined(TRMMKERNEL) || \
+	(defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+	(defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+	movq	B, BO
+#else
+	movq	KK, %rax
+	leaq	(, %rax, SIZE), %rax
+	leaq	(AO, %rax, 4), AO
+	leaq	(B,  %rax, 2), BO
+#endif	
+
+	vmovddup	-16 * SIZE(BO), %xmm1
+	vmovddup	-15 * SIZE(BO), %xmm5
+	vmovddup	-12 * SIZE(BO), %xmm3
+	vxorps	%xmm8, %xmm8,%xmm8
+	vxorps	%xmm9, %xmm9,%xmm9
+	vxorps	%xmm12, %xmm12,%xmm12
+	vxorps	%xmm13, %xmm13,%xmm13
+	vmovups	-16 * SIZE(AO), %xmm0
+	vmovups	 -8 * SIZE(AO), %xmm4
+	// prefetcht0 256(CO1)
+	// prefetcht0 320(CO1)
+	// prefetcht0 256(CO2)
+	// prefetcht0 320(CO2)
+	// prefetchnta	  24 * SIZE(CO1)
+	// prefetchnta	  32 * SIZE(CO1)
+	// prefetchw      3 * SIZE(CO1)
+	vmovups	%xmm0, %xmm2
+	// prefetchw      3 * SIZE(CO2)
+	// prefetchnta -16 * SIZE(BB)
+	// prefetch	 -16 * SIZE(BB)
+	subq		 $-8 * SIZE, BB
+
+#ifndef TRMMKERNEL
+	movq	K, %rax
+#elif (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+	movq	K, %rax
+	subq	KK, %rax
+	movq	%rax, KKK	
+#else
+	movq	KK, %rax
+#ifdef LEFT
+	addq	$4, %rax
+#else
+	addq	$2, %rax
+#endif
+	movq	%rax, KKK
+#endif
+
+
+	andq	$-4, %rax
+	leaq	(, %rax, SIZE), %rax
+	leaq	(AO, %rax, 4), AO
+	leaq	(BO, %rax, 2), BO
+	negq	%rax
+	NOBRANCH
+	je	.L56
+	ALIGN_4
+
+.L52:
+	vfmaddpd %xmm8,%xmm1, %xmm0,%xmm8
+	vfmaddpd %xmm9,%xmm5, %xmm2,%xmm9
+	vmovups  -14 * SIZE(AO, %rax, 4),%xmm2
+	vfmaddpd %xmm12,%xmm2, %xmm1,%xmm12
+	vmovups	-12 * SIZE(AO, %rax, 4), %xmm0
+	vmovddup	-14 * SIZE(BO, %rax, 2), %xmm1
+	vfmaddpd %xmm13,%xmm2, %xmm5,%xmm13
+	vmovddup	-13 * SIZE(BO, %rax, 2), %xmm5
+	vmovups 	-10 * SIZE(AO, %rax, 4), %xmm2
+	vfmaddpd %xmm8,%xmm1, %xmm0,%xmm8
+	vfmaddpd %xmm12,%xmm2, %xmm1,%xmm12
+	vfmaddpd %xmm9,%xmm5, %xmm0,%xmm9
+	vmovups	  (AO, %rax, 4), %xmm0
+	vmovddup	 -8 * SIZE(BO, %rax, 2), %xmm1
+	vfmaddpd %xmm13,%xmm2, %xmm5,%xmm13
+	vmovddup	-11 * SIZE(BO, %rax, 2), %xmm5
+	vmovups	-6 * SIZE(AO, %rax, 4), %xmm2
+	vfmaddpd %xmm8,%xmm3, %xmm4,%xmm8
+	vfmaddpd %xmm12,%xmm2, %xmm3,%xmm12
+	vfmaddpd %xmm9,%xmm5, %xmm4,%xmm9
+	vmovups	 -4 * SIZE(AO, %rax, 4), %xmm4
+	vmovddup	-10 * SIZE(BO, %rax, 2), %xmm3
+	vfmaddpd %xmm13,%xmm2, %xmm5,%xmm13
+	vmovddup	 -9 * SIZE(BO, %rax, 2), %xmm5
+	vmovups		-2 * SIZE(AO, %rax, 4), %xmm2
+	vfmaddpd %xmm8,%xmm3, %xmm4,%xmm8
+	vfmaddpd %xmm12,%xmm2, %xmm3,%xmm12
+	vfmaddpd %xmm9,%xmm5, %xmm4,%xmm9
+	vfmaddpd %xmm13,%xmm2, %xmm5,%xmm13
+	vmovups	  8 * SIZE(AO, %rax, 4), %xmm4
+	vmovddup	 -4 * SIZE(BO, %rax, 2), %xmm3
+	vmovddup	 -7 * SIZE(BO, %rax, 2), %xmm5
+	vmovaps	%xmm0, %xmm2
+
+	addq	$4 * SIZE, %rax
+	BRANCH
+	jl	.L52
+	ALIGN_4
+
+.L56:
+	vmovddup	ALPHA, %xmm7
+
+#ifndef TRMMKERNEL
+	movq	K, %rax
+#else
+	movq	KKK, %rax
+#endif
+	andq	$3, %rax		# if (k & 1)
+	je .L59
+
+	leaq	(, %rax, SIZE), %rax
+	leaq	(AO, %rax, 4), AO
+	leaq	(BO, %rax, 2), BO
+	negq	%rax
+	ALIGN_4
+
+.L57:
+	vfmaddpd %xmm8,%xmm1, %xmm0,%xmm8
+	vfmaddpd %xmm9,%xmm5, %xmm2,%xmm9
+	vmovups -14 * SIZE(AO, %rax, 4),%xmm2
+	vfmaddpd %xmm12,%xmm2, %xmm1,%xmm12
+	vmovups	-12 * SIZE(AO, %rax, 4), %xmm0
+	vmovddup	-14 * SIZE(BO, %rax, 2), %xmm1
+	vfmaddpd %xmm13,%xmm2, %xmm5,%xmm13
+	vmovddup	-13 * SIZE(BO, %rax, 2), %xmm5
+	vmovaps	%xmm0, %xmm2
+
+	addq	$SIZE, %rax
+	jl	.L57
+	ALIGN_4
+
+.L59:
+#ifndef TRMMKERNEL
+	vfmaddpd	(CO1),%xmm7, %xmm8, %xmm8
+	vfmaddpd	2 * SIZE(CO1),%xmm7, %xmm12, %xmm12
+	vfmaddpd	(CO2),%xmm7, %xmm9, %xmm9
+	vfmaddpd	2 * SIZE(CO2),%xmm7, %xmm13, %xmm13
+
+#else
+	vmulpd	%xmm7, %xmm8,%xmm8
+	vmulpd	%xmm7, %xmm9,%xmm9
+	vmulpd	%xmm7, %xmm12,%xmm12
+	vmulpd	%xmm7, %xmm13,%xmm13
+
+#endif
+
+	vmovups	%xmm8,  (CO1)
+	vmovups	%xmm12, 2 * SIZE(CO1)
+
+	vmovups	%xmm9,  (CO2)
+	vmovups	%xmm13, 2 * SIZE(CO2)
+
+#if (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+    (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+	movq	K, %rax
+	subq	KKK, %rax
+	leaq	(,%rax, SIZE), %rax
+	leaq	(AO, %rax, 4), AO
+	leaq	(BO, %rax, 2), BO
+#endif
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+	addq	$4, KK
+#endif
+
+	addq	$4 * SIZE, CO1		# coffset += 4
+	addq	$4 * SIZE, CO2		# coffset += 4
+	decq	I			# i --
+	jg	.L51
+	ALIGN_4	
+
+.L60:
+	testq	$2, M
+	je	.L70
+	ALIGN_4
+
+.L61:
+#if !defined(TRMMKERNEL) || \
+	(defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+	(defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+	movq	B, BO
+#else
+	movq	KK, %rax
+	leaq	(, %rax, SIZE), %rax
+	leaq	(AO, %rax, 2), AO
+	leaq	(B,  %rax, 2), BO
+#endif	
+
+	vmovups	-16 * SIZE(AO), %xmm0
+	vxorps	%xmm8, %xmm8,%xmm8
+	vmovups	-12 * SIZE(AO), %xmm2
+	vxorps	%xmm9, %xmm9,%xmm9
+	vmovddup	-16 * SIZE(BO), %xmm1
+	vxorps	%xmm10, %xmm10,%xmm10
+	vmovddup	-15 * SIZE(BO), %xmm3
+	vxorps	%xmm11, %xmm11,%xmm11
+
+#ifndef TRMMKERNEL
+	movq	K, %rax
+#elif (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+	movq	K, %rax
+	subq	KK, %rax
+	movq	%rax, KKK	
+#else
+	movq	KK, %rax
+#ifdef LEFT
+	addq	$2, %rax
+#else
+	addq	$2, %rax
+#endif
+	movq	%rax, KKK
+#endif
+
+	andq	$-4, %rax
+	leaq	(, %rax, SIZE), %rax
+	leaq	(AO, %rax, 2), AO
+	leaq	(BO, %rax, 2), BO
+	negq	%rax
+	NOBRANCH
+	je	.L66
+	ALIGN_4
+
+.L62:
+	vfmaddpd %xmm8,%xmm0, %xmm1,%xmm8
+	vmovddup	-14 * SIZE(BO, %rax, 2), %xmm1
+	vfmaddpd %xmm9,%xmm0, %xmm3,%xmm9
+	vmovups	-14 * SIZE(AO, %rax, 2), %xmm0
+	vmovddup	-13 * SIZE(BO, %rax, 2), %xmm3
+	vfmaddpd %xmm10,%xmm0, %xmm1,%xmm10
+	vmovddup	-12 * SIZE(BO, %rax, 2), %xmm1
+	vfmaddpd %xmm11,%xmm0, %xmm3,%xmm11
+	vmovups	 -8 * SIZE(AO, %rax, 2), %xmm0
+	vmovddup	-11 * SIZE(BO, %rax, 2), %xmm3
+	vfmaddpd %xmm8,%xmm2, %xmm1,%xmm8
+	vmovddup	-10 * SIZE(BO, %rax, 2), %xmm1
+	vfmaddpd %xmm9,%xmm2, %xmm3,%xmm9
+	vmovups	-10 * SIZE(AO, %rax, 2), %xmm2
+	vmovddup	 -9 * SIZE(BO, %rax, 2), %xmm3
+	vfmaddpd %xmm10,%xmm2, %xmm1,%xmm10
+	vmovddup	 -8 * SIZE(BO, %rax, 2), %xmm1
+	vfmaddpd %xmm11,%xmm2, %xmm3,%xmm11
+	vmovups	 -4 * SIZE(AO, %rax, 2), %xmm2
+	vmovddup	 -7 * SIZE(BO, %rax, 2), %xmm3
+
+	addq	$4 * SIZE, %rax
+	BRANCH
+	jl	.L62
+	ALIGN_4
+
+.L66:
+	vmovddup	ALPHA, %xmm7
+
+#ifndef TRMMKERNEL
+	movq	K, %rax
+#else
+	movq	KKK, %rax
+#endif
+	andq	$3, %rax		# if (k & 1)
+	je .L69
+
+	leaq	(, %rax, SIZE), %rax
+	leaq	(AO, %rax, 2), AO
+	leaq	(BO, %rax, 2), BO
+	negq	%rax
+	ALIGN_4
+
+.L67:
+	vfmaddpd %xmm8,%xmm0, %xmm1,%xmm8
+	vmovddup	-14 * SIZE(BO, %rax, 2), %xmm1
+	vfmaddpd %xmm9,%xmm0, %xmm3,%xmm9
+	vmovups	-14 * SIZE(AO, %rax, 2), %xmm0
+	vmovddup	-13 * SIZE(BO, %rax, 2), %xmm3
+
+	addq	$SIZE, %rax
+	jl	.L67
+	ALIGN_4
+
+.L69:
+	vaddpd	%xmm10, %xmm8,%xmm8
+	vaddpd	%xmm11, %xmm9,%xmm9
+	
+#ifndef TRMMKERNEL
+
+	vfmaddpd	(CO1),%xmm7, %xmm8,%xmm8
+	vfmaddpd	(CO2),%xmm7, %xmm9,%xmm9
+
+#else
+
+	vmulpd	%xmm7, %xmm8,%xmm8
+	vmulpd	%xmm7, %xmm9,%xmm9
+
+#endif
+
+	vmovups	%xmm8,  (CO1)
+	vmovups	%xmm9,  (CO2)
+
+#if (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+    (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+	movq	K, %rax
+	subq	KKK, %rax
+	leaq	(,%rax, SIZE), %rax
+	leaq	(AO, %rax, 2), AO
+	leaq	(BO, %rax, 2), BO
+#endif
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+	addq	$2, KK
+#endif
+
+	addq	$2 * SIZE, CO1		# coffset += 4
+	addq	$2 * SIZE, CO2		# coffset += 4
+	ALIGN_4	
+
+.L70:
+	testq	$1, M
+	je	.L79
+	ALIGN_4
+
+.L71:
+#if !defined(TRMMKERNEL) || \
+	(defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+	(defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+	movq	B, BO
+#else
+	movq	KK, %rax
+	leaq	(, %rax, SIZE), %rax
+	leaq	(AO, %rax, 1), AO
+	leaq	(B,  %rax, 2), BO
+#endif	
+
+	vmovddup	-16 * SIZE(AO), %xmm0
+	vxorps	%xmm8, %xmm8,%xmm8
+	vmovddup	-15 * SIZE(AO), %xmm1
+	vxorps	%xmm9, %xmm9,%xmm9
+	vmovddup	-14 * SIZE(AO), %xmm2
+	vxorps	%xmm10, %xmm10,%xmm10
+	vmovddup	-13 * SIZE(AO), %xmm3
+	vxorps	%xmm11, %xmm11,%xmm11
+
+#ifndef TRMMKERNEL
+	movq	K, %rax
+#elif (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+	movq	K, %rax
+	subq	KK, %rax
+	movq	%rax, KKK	
+#else
+	movq	KK, %rax
+#ifdef LEFT
+	addq	$1, %rax
+#else
+	addq	$2, %rax
+#endif
+	movq	%rax, KKK
+#endif
+
+	andq	$-4, %rax
+	leaq	(, %rax, SIZE), %rax
+	leaq	(AO, %rax, 1), AO
+	leaq	(BO, %rax, 2), BO
+	negq	%rax
+	NOBRANCH
+	je	.L76
+	ALIGN_4
+
+.L72:
+	vfmaddpd %xmm8,-16 * SIZE(BO, %rax, 2), %xmm0,%xmm8
+	vmovddup	-12 * SIZE(AO, %rax, 1), %xmm0
+
+	vfmaddpd %xmm9,-14 * SIZE(BO, %rax, 2), %xmm1,%xmm9
+	vmovddup	-11 * SIZE(AO, %rax, 1), %xmm1
+
+	vfmaddpd %xmm10,-12 * SIZE(BO, %rax, 2), %xmm2,%xmm10
+	vmovddup	-10 * SIZE(AO, %rax, 1), %xmm2
+
+	vfmaddpd %xmm11,-10 * SIZE(BO, %rax, 2), %xmm3,%xmm11
+	vmovddup	 -9 * SIZE(AO, %rax, 1), %xmm3
+
+	addq	$4 * SIZE, %rax
+	BRANCH
+	jl	.L72
+	ALIGN_4
+
+.L76:
+	vmovddup	ALPHA, %xmm7
+
+#ifndef TRMMKERNEL
+	movq	K, %rax
+#else
+	movq	KKK, %rax
+#endif
+	andq	$3, %rax		# if (k & 1)
+	je .L78
+
+	leaq	(, %rax, SIZE), %rax
+	leaq	(AO, %rax, 1), AO
+	leaq	(BO, %rax, 2), BO
+	negq	%rax
+	ALIGN_4
+
+.L77:
+	vfmaddpd %xmm8,-16 * SIZE(BO, %rax, 2), %xmm0,%xmm8
+	vmovddup	-15 * SIZE(AO, %rax, 1), %xmm0
+
+	addq	$SIZE, %rax
+	jl	.L77
+	ALIGN_4
+
+.L78:
+	vaddpd	%xmm9,  %xmm8,%xmm8
+	vaddpd	%xmm11, %xmm10,%xmm10
+	vaddpd	%xmm10, %xmm8,%xmm8
+
+#ifndef TRMMKERNEL
+	vmovsd	(CO1), %xmm0
+	vmovhpd	(CO2), %xmm0,%xmm0
+#endif
+
+	vmulpd	%xmm7, %xmm8,%xmm8
+
+#ifndef TRMMKERNEL
+	vaddpd	%xmm0, %xmm8,%xmm8
+#endif
+
+	vmovsd	%xmm8,  (CO1)
+	vmovhpd	%xmm8,  (CO2)
+
+#if (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+    (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+	movq	K, %rax
+	subq	KKK, %rax
+	leaq	(,%rax, SIZE), %rax
+	leaq	(AO, %rax, 1), AO
+	leaq	(BO, %rax, 2), BO
+#endif
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+	addq	$1, KK
+#endif
+	ALIGN_4	
+	
+.L79:
+#if defined(TRMMKERNEL) && !defined(LEFT)
+	addq	$2, KK
+#endif
+
+	movq	BO, B
+
+	leaq	(C, LDC, 2), C
+	ALIGN_4
+
+.L80:
+	testq	$1, N
+	je	.L999
+	ALIGN_4
+
+.L81:
+#if defined(TRMMKERNEL) && defined(LEFT)
+	movq	OFFSET, %rax
+	movq	%rax, KK
+#endif	
+
+	movq	C, CO1			# coffset1 = c
+	movq	A, AO		# aoffset = a
+
+	movq	M,  I
+	sarq	$2, I	# i = (m >> 2)
+	jle	.L100
+	ALIGN_4
+
+.L91:
+#if !defined(TRMMKERNEL) || \
+	(defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+	(defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+	movq	B, BO
+#else
+	movq	KK, %rax
+	leaq	(, %rax, SIZE), %rax
+	leaq	(AO, %rax, 4), AO
+	leaq	(B,  %rax, 1), BO
+#endif	
+
+	vmovups	 -8 * SIZE(AO), %xmm2
+	vxorps	%xmm8, %xmm8,%xmm8
+	vmovups	-16 * SIZE(AO), %xmm0
+	vxorps	%xmm9, %xmm9,%xmm9
+	vmovddup	-16 * SIZE(BO), %xmm1
+	vxorps	%xmm12, %xmm12,%xmm12
+	vmovddup	-14 * SIZE(BO), %xmm3
+	vxorps	%xmm13, %xmm13,%xmm13
+	vmovddup	-15 * SIZE(BO), %xmm5
+
+	// prefetchw      3 * SIZE(CO1)
+
+#ifndef TRMMKERNEL
+	movq	K, %rax
+#elif (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+	movq	K, %rax
+	subq	KK, %rax
+	movq	%rax, KKK	
+#else
+	movq	KK, %rax
+#ifdef LEFT
+	addq	$4, %rax
+#else
+	addq	$1, %rax
+#endif
+	movq	%rax, KKK
+#endif
+
+	andq	$-4, %rax
+	leaq	(, %rax, SIZE), %rax
+	leaq	(AO, %rax, 4), AO
+	leaq	(BO, %rax, 1), BO
+	negq	%rax
+	NOBRANCH
+	je	.L96
+	ALIGN_4
+
+.L92:
+	vfmaddpd %xmm8,%xmm1, %xmm0,%xmm8
+	vfmaddpd %xmm12,-14 * SIZE(AO, %rax, 4), %xmm1,%xmm12
+	vmovapd	-12 * SIZE(AO, %rax, 4), %xmm0
+	vmovddup	-12 * SIZE(BO, %rax, 1), %xmm1
+	vfmaddpd %xmm9,%xmm5, %xmm0,%xmm9
+	vfmaddpd %xmm13,-10 * SIZE(AO, %rax, 4), %xmm5,%xmm13
+	vmovapd	  (AO, %rax, 4), %xmm0
+	vmovddup	-13 * SIZE(BO, %rax, 1), %xmm5
+	vfmaddpd %xmm8,%xmm3, %xmm2,%xmm8
+	vfmaddpd %xmm12,-6 * SIZE(AO, %rax, 4), %xmm3,%xmm12
+	vmovapd	 -4 * SIZE(AO, %rax, 4), %xmm2
+	vmovddup	-10 * SIZE(BO, %rax, 1), %xmm3
+	vfmaddpd %xmm9,%xmm5, %xmm2,%xmm9
+	vfmaddpd %xmm13,-2 * SIZE(AO, %rax, 4), %xmm5,%xmm13
+	vmovapd	  8 * SIZE(AO, %rax, 4), %xmm2
+	vmovddup	-11 * SIZE(BO, %rax, 1), %xmm5
+
+	addq	$4 * SIZE, %rax
+	BRANCH
+	jl	.L92
+	ALIGN_4
+
+.L96:
+	vmovddup	ALPHA, %xmm7
+
+#ifndef TRMMKERNEL
+	movq	K, %rax
+#else
+	movq	KKK, %rax
+#endif
+	andq	$3, %rax		# if (k & 1)
+	je .L99
+
+	leaq	(, %rax, SIZE), %rax
+	leaq	(AO, %rax, 4), AO
+	leaq	(BO, %rax, 1), BO
+	negq	%rax
+	ALIGN_4
+
+.L97:
+	vfmaddpd %xmm8,%xmm1, %xmm0,%xmm8
+	vfmaddpd %xmm12,-14 * SIZE(AO, %rax, 4), %xmm1,%xmm12
+	vmovups	-12 * SIZE(AO, %rax, 4), %xmm0
+	vmovddup	-15 * SIZE(BO, %rax, 1), %xmm1
+
+	addq	$SIZE, %rax
+	jl	.L97
+	ALIGN_4
+
+.L99:
+	vaddpd	%xmm9,  %xmm8,%xmm8
+	vaddpd	%xmm13, %xmm12,%xmm12
+
+#ifndef TRMMKERNEL
+
+	vfmaddpd		(CO1),%xmm7, %xmm8,%xmm8
+	vfmaddpd	2 * SIZE(CO1),%xmm7,%xmm12,%xmm12
+
+#else
+	vmulpd	%xmm7, %xmm8,%xmm8
+	vmulpd	%xmm7, %xmm12,%xmm12
+
+#endif
+
+	vmovups	%xmm8,  (CO1)
+	vmovups	%xmm12, 2 * SIZE(CO1)
+
+#if (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+    (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+	movq	K, %rax
+	subq	KKK, %rax
+	leaq	(,%rax, SIZE), %rax
+	leaq	(AO, %rax, 4), AO
+	leaq	(BO, %rax, 1), BO
+#endif
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+	addq	$4, KK
+#endif
+
+	addq	$4 * SIZE, CO1		# coffset += 4
+	decq	I			# i --
+	jg	.L91
+	ALIGN_4	
+
+.L100:
+	testq	$2, M
+	je	.L110
+	ALIGN_4
+
+.L101:
+#if !defined(TRMMKERNEL) || \
+	(defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+	(defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+	movq	B, BO
+#else
+	movq	KK, %rax
+	leaq	(, %rax, SIZE), %rax
+	leaq	(AO, %rax, 2), AO
+	leaq	(B,  %rax, 1), BO
+#endif	
+
+	vmovddup	-16 * SIZE(BO), %xmm0
+	vxorps	%xmm8, %xmm8,%xmm8
+	vmovddup	-15 * SIZE(BO), %xmm1
+	vxorps	%xmm9, %xmm9,%xmm9
+	vmovddup	-14 * SIZE(BO), %xmm2
+	vxorps	%xmm10, %xmm10,%xmm10
+	vmovddup	-13 * SIZE(BO), %xmm3
+	vxorps	%xmm11, %xmm11,%xmm11
+
+#ifndef TRMMKERNEL
+	movq	K, %rax
+#elif (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+	movq	K, %rax
+	subq	KK, %rax
+	movq	%rax, KKK	
+#else
+	movq	KK, %rax
+#ifdef LEFT
+	addq	$2, %rax
+#else
+	addq	$1, %rax
+#endif
+	movq	%rax, KKK
+#endif
+
+	andq	$-4, %rax
+	leaq	(, %rax, SIZE), %rax
+	leaq	(AO, %rax, 2), AO
+	leaq	(BO, %rax, 1), BO
+	negq	%rax
+	NOBRANCH
+	je	.L106
+	ALIGN_4
+
+.L102:
+	vfmaddpd %xmm8,-16 * SIZE(AO, %rax, 2), %xmm0,%xmm8
+	vmovddup	-12 * SIZE(BO, %rax, 1), %xmm0
+
+	vfmaddpd %xmm9,-14 * SIZE(AO, %rax, 2), %xmm1,%xmm9
+	vmovddup	-11 * SIZE(BO, %rax, 1), %xmm1
+
+	vfmaddpd %xmm10,-12 * SIZE(AO, %rax, 2), %xmm2,%xmm10
+	vmovddup	-10 * SIZE(BO, %rax, 1), %xmm2
+
+	vfmaddpd %xmm11,-10 * SIZE(AO, %rax, 2), %xmm3,%xmm11
+	vmovddup	 -9 * SIZE(BO, %rax, 1), %xmm3
+
+	addq	$4 * SIZE, %rax
+	BRANCH
+	jl	.L102
+	ALIGN_4
+
+.L106:
+	vmovddup	ALPHA, %xmm7
+
+#ifndef TRMMKERNEL
+	movq	K, %rax
+#else
+	movq	KKK, %rax
+#endif
+	andq	$3, %rax		# if (k & 1)
+	je .L109
+
+	leaq	(, %rax, SIZE), %rax
+	leaq	(AO, %rax, 2), AO
+	leaq	(BO, %rax, 1), BO
+	negq	%rax
+	ALIGN_4
+
+.L107:
+	vmovddup	-16 * SIZE(BO, %rax, 1), %xmm0
+	vfmaddpd %xmm8,-16 * SIZE(AO, %rax, 2), %xmm0,%xmm8
+
+	addq	$SIZE, %rax
+	jl	.L107
+	ALIGN_4
+
+.L109:
+	vaddpd	%xmm9,  %xmm8,%xmm8
+	vaddpd	%xmm11, %xmm10,%xmm10
+	vaddpd	%xmm10, %xmm8,%xmm8
+
+#ifndef TRMMKERNEL
+
+	vfmaddpd	(CO1),%xmm7, %xmm8,%xmm8
+#else
+	vmulpd	%xmm7, %xmm8,%xmm8
+
+#endif
+
+	vmovups	%xmm8,  (CO1)
+
+#if (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+    (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+	movq	K, %rax
+	subq	KKK, %rax
+	leaq	(,%rax, SIZE), %rax
+	leaq	(AO, %rax, 2), AO
+	leaq	(BO, %rax, 1), BO
+#endif
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+	addq	$2, KK
+#endif
+
+	addq	$2 * SIZE, CO1		# coffset += 4
+
+	ALIGN_4	
+
+.L110:
+	testq	$1, M
+	je	.L999
+	ALIGN_4
+
+.L111:
+#if !defined(TRMMKERNEL) || \
+	(defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+	(defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+	movq	B, BO
+#else
+	movq	KK, %rax
+	leaq	(, %rax, SIZE), %rax
+	leaq	(AO, %rax, 1), AO
+	leaq	(B,  %rax, 1), BO
+#endif	
+
+	vmovups	-16 * SIZE(AO), %xmm0
+	vxorps	%xmm8, %xmm8,%xmm8
+	movups	-14 * SIZE(AO), %xmm1
+	vxorps	%xmm9, %xmm9,%xmm9
+
+#ifndef TRMMKERNEL
+	movq	K, %rax
+#elif (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+	movq	K, %rax
+	subq	KK, %rax
+	movq	%rax, KKK	
+#else
+	movq	KK, %rax
+#ifdef LEFT
+	addq	$1, %rax
+#else
+	addq	$1, %rax
+#endif
+	movq	%rax, KKK
+#endif
+
+	andq	$-4, %rax
+	leaq	(, %rax, SIZE), %rax
+	leaq	(AO, %rax, 1), AO
+	leaq	(BO, %rax, 1), BO
+	negq	%rax
+	NOBRANCH
+	je	.L116
+	ALIGN_4
+
+.L112:
+	vfmaddpd %xmm8,-16 * SIZE(BO, %rax, 1), %xmm0,%xmm8
+	vmovups	-12 * SIZE(AO, %rax, 1), %xmm0
+
+	vfmaddpd %xmm9,-14 * SIZE(BO, %rax, 1), %xmm1,%xmm9
+	vmovups	-10 * SIZE(AO, %rax, 1), %xmm1
+
+	addq	$4 * SIZE, %rax
+	BRANCH
+	jl	.L112
+	ALIGN_4
+
+.L116:
+	vmovddup	ALPHA, %xmm7
+
+#ifndef TRMMKERNEL
+	movq	K, %rax
+#else
+	movq	KKK, %rax
+#endif
+	andq	$3, %rax		# if (k & 1)
+	je .L118
+
+	leaq	(, %rax, SIZE), %rax
+	leaq	(AO, %rax, 1), AO
+	leaq	(BO, %rax, 1), BO
+	negq	%rax
+	ALIGN_4
+
+.L117:
+	vmulsd	-16 * SIZE(BO, %rax, 1), %xmm0,%xmm0
+	vaddsd	%xmm0, %xmm8,%xmm8
+	vmovsd	-15 * SIZE(AO, %rax, 1), %xmm0
+
+	addq	$SIZE, %rax
+	jl	.L117
+	ALIGN_4
+
+.L118:
+	vaddpd	%xmm9, %xmm8,%xmm8
+	vhaddpd	%xmm8, %xmm8,%xmm8
+
+#ifndef TRMMKERNEL
+	vmovsd	(CO1), %xmm0
+#endif
+
+	vmulsd	%xmm7, %xmm8,%xmm8
+
+#ifndef TRMMKERNEL
+	vaddsd	%xmm0, %xmm8,%xmm8
+#endif
+
+	vmovsd	%xmm8,  (CO1)
+	ALIGN_4
+	
+.L999:
+	movq	   (%rsp), %rbx
+	movq	  8(%rsp), %rbp
+	movq	 16(%rsp), %r12
+	movq	 24(%rsp), %r13
+	movq	 32(%rsp), %r14
+	movq	 40(%rsp), %r15
+
+#ifdef WINDOWS_ABI
+	movq	 48(%rsp), %rdi
+	movq	 56(%rsp), %rsi
+	movups	 64(%rsp), %xmm6
+	movups	 80(%rsp), %xmm7
+	movups	 96(%rsp), %xmm8
+	movups	112(%rsp), %xmm9
+	movups	128(%rsp), %xmm10
+	movups	144(%rsp), %xmm11
+	movups	160(%rsp), %xmm12
+	movups	176(%rsp), %xmm13
+	movups	192(%rsp), %xmm14
+	movups	208(%rsp), %xmm15
+#endif
+
+	addq	$STACKSIZE, %rsp
+	ret
+
+	EPILOGUE
diff --git a/kernel/x86_64/gemm_kernel_8x4_barcelona.S b/kernel/x86_64/gemm_kernel_8x4_barcelona.S
index b40c8bac7..becd19544 100644
--- a/kernel/x86_64/gemm_kernel_8x4_barcelona.S
+++ b/kernel/x86_64/gemm_kernel_8x4_barcelona.S
@@ -930,7 +930,7 @@
 .L22:
 	mulps	%xmm8, %xmm9
 	addps	%xmm9, %xmm0
-#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI)
+#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 	PREFETCH	(PREFETCHSIZE +  0) * SIZE(AO)
 #endif
 	movaps	 4 * SIZE(BO), %xmm9
@@ -983,7 +983,7 @@
 	addps	%xmm8, %xmm3
 	movaps	 0 * SIZE(AO), %xmm8
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI)
+#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 	PREFETCH	(PREFETCHSIZE + 16) * SIZE(AO)
 #endif
 	mulps	%xmm10, %xmm9
@@ -1178,7 +1178,7 @@
 .L32:
 	mulps	%xmm8, %xmm9
 	addps	%xmm9, %xmm0
-#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI)
+#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 	PREFETCH	(PREFETCHSIZE +  0) * SIZE(AO)
 #endif
 	movsd	 4 * SIZE(BO), %xmm9
@@ -1423,7 +1423,7 @@
 .L42:
 	mulss	%xmm8, %xmm9
 	addss	%xmm9, %xmm0
-#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI)
+#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 	PREFETCH	(PREFETCHSIZE +  0) * SIZE(AO)
 #endif
 	movss	 4 * SIZE(BO), %xmm9
@@ -1765,7 +1765,7 @@
 
 .L62:
 	mulps	%xmm8, %xmm9
-#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI)
+#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 	PREFETCH	(PREFETCHSIZE +  0) * SIZE(AO)
 #endif
 	mulps	 4 * SIZE(BO), %xmm8
@@ -1793,7 +1793,7 @@
 	addps	%xmm8, %xmm5
 	movaps	32 * SIZE(AO), %xmm8
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI)
+#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 	PREFETCH	(PREFETCHSIZE + 16) * SIZE(AO)
 #endif
 	mulps	%xmm10, %xmm11
@@ -1822,7 +1822,7 @@
 	addps	%xmm10, %xmm5
 	movaps	48 * SIZE(AO), %xmm10
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI)
+#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 	PREFETCH	(PREFETCHSIZE + 32) * SIZE(AO)
 #endif
 	mulps	%xmm12, %xmm13
@@ -1851,7 +1851,7 @@
 	addps	%xmm12, %xmm5
 	movaps	64 * SIZE(AO), %xmm12
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI)
+#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 	PREFETCH	(PREFETCHSIZE + 48) * SIZE(AO)
 #endif
 	mulps	%xmm14, %xmm15
@@ -2024,7 +2024,7 @@
 
 .L72:
 	mulps	%xmm8, %xmm9
-#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI)
+#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 	PREFETCH	(PREFETCHSIZE +  0) * SIZE(AO)
 #endif
 
@@ -2208,7 +2208,7 @@
 .L82:
 	mulps	%xmm8, %xmm9
 	addps	%xmm9, %xmm0
-#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI)
+#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 	PREFETCH	(PREFETCHSIZE +  0) * SIZE(AO)
 #endif
 	movsd	 4 * SIZE(BO), %xmm9
@@ -2395,7 +2395,7 @@
 .L92:
 	mulps	%xmm8, %xmm9
 	addps	%xmm9, %xmm0
-#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI)
+#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 	PREFETCH	(PREFETCHSIZE +  0) * SIZE(AO)
 #endif
 	movss	 4 * SIZE(BO), %xmm9
@@ -2670,7 +2670,7 @@
 
 .L112:
 	mulps	%xmm9, %xmm8
-#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI)
+#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 	PREFETCH	(PREFETCHSIZE +  0) * SIZE(AO)
 #endif
 
@@ -2687,7 +2687,7 @@
 	addps	%xmm9, %xmm4
 	movaps	 8 * SIZE(BO), %xmm9
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI)
+#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 	PREFETCH	(PREFETCHSIZE + 16) * SIZE(AO)
 #endif
 	mulps	%xmm9, %xmm10
@@ -2704,7 +2704,7 @@
 	addps	%xmm9, %xmm4
 	movaps	32 * SIZE(BO), %xmm9
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI)
+#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 	PREFETCH	(PREFETCHSIZE + 32) * SIZE(AO)
 #endif
 	mulps	%xmm11, %xmm12
@@ -2721,7 +2721,7 @@
 	addps	%xmm11, %xmm4
 	movaps	24 * SIZE(BO), %xmm11
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI)
+#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 	PREFETCH	(PREFETCHSIZE + 48) * SIZE(AO)
 #endif
 	mulps	%xmm11, %xmm14
@@ -2857,7 +2857,7 @@
 
 .L122:
 	mulps	%xmm8, %xmm9
-#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI)
+#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 	PREFETCH	(PREFETCHSIZE +  0) * SIZE(AO)
 #endif
 	movaps	-28 * SIZE(AO), %xmm8
@@ -2873,7 +2873,7 @@
 	addps	%xmm8, %xmm3
 	movaps	  0 * SIZE(AO), %xmm8
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI)
+#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 	PREFETCH	(PREFETCHSIZE + 16) * SIZE(AO)
 #endif
 	mulps	%xmm10, %xmm11
@@ -3003,7 +3003,7 @@
 
 .L132:
 	mulps	%xmm8, %xmm9
-#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI)
+#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 	PREFETCH	(PREFETCHSIZE +  0) * SIZE(AO)
 #endif
 	movsd	-30 * SIZE(AO), %xmm8
@@ -3150,7 +3150,7 @@
 
 .L142:
 	mulss	%xmm8, %xmm9
-#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI)
+#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 	PREFETCH	(PREFETCHSIZE +  0) * SIZE(AO)
 #endif
 	movss	-31 * SIZE(AO), %xmm8
diff --git a/kernel/x86_64/gemm_ncopy_4_opteron.S b/kernel/x86_64/gemm_ncopy_4_opteron.S
index edde7e2c1..e5cbd62eb 100644
--- a/kernel/x86_64/gemm_ncopy_4_opteron.S
+++ b/kernel/x86_64/gemm_ncopy_4_opteron.S
@@ -39,7 +39,7 @@
 #define ASSEMBLER
 #include "common.h"
 
-#if defined(BARCELONA) || defined(SHANGHAI)
+#if defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 #define RPREFETCHSIZE (12 + 4)
 #define WPREFETCHSIZE (48 + 4)
 #define MOVNTQ	 MOVQ
@@ -79,7 +79,7 @@
 #define AO3	%r13
 #define AO4	%rax
 
-#if defined(BARCELONA) || defined(SHANGHAI)
+#if defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 #define RPREFETCH prefetch
 #else
 #define RPREFETCH prefetch
diff --git a/kernel/x86_64/gemm_tcopy_4_opteron.S b/kernel/x86_64/gemm_tcopy_4_opteron.S
index 459eeb8c5..105fe3b47 100644
--- a/kernel/x86_64/gemm_tcopy_4_opteron.S
+++ b/kernel/x86_64/gemm_tcopy_4_opteron.S
@@ -39,7 +39,7 @@
 #define ASSEMBLER
 #include "common.h"
 
-#if defined(BARCELONA) || defined(SHANGHAI)
+#if defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 #define RPREFETCHSIZE (12 + 4)
 #define WPREFETCHSIZE (12 + 4)
 #define MOVNTQ	 MOVQ
@@ -96,7 +96,7 @@
 
 #endif
 
-#if defined(BARCELONA) || defined(SHANGHAI)
+#if defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 #define RPREFETCH prefetch
 #else
 #define RPREFETCH prefetch
diff --git a/kernel/x86_64/izamax_sse2.S b/kernel/x86_64/izamax_sse2.S
index 4e66e5338..404608256 100644
--- a/kernel/x86_64/izamax_sse2.S
+++ b/kernel/x86_64/izamax_sse2.S
@@ -469,7 +469,7 @@
 	ALIGN_4
 
 .L71:
-#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI)
+#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 	prefetch	PREFETCHSIZE * SIZE(X)
 #endif
 
diff --git a/kernel/x86_64/scal_sse.S b/kernel/x86_64/scal_sse.S
index 323e8b9dd..9c8dd9dc2 100644
--- a/kernel/x86_64/scal_sse.S
+++ b/kernel/x86_64/scal_sse.S
@@ -266,7 +266,7 @@
 	sarq	$5, I
 	jle	.L113
 
-#if defined(BARCELONA) || defined(SHANGHAI)
+#if defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 
 	movaps	%xmm0, %xmm1
 	mulps	-32 * SIZE(X), %xmm1
diff --git a/kernel/x86_64/scal_sse2.S b/kernel/x86_64/scal_sse2.S
index 8f5612081..3823b1fc9 100644
--- a/kernel/x86_64/scal_sse2.S
+++ b/kernel/x86_64/scal_sse2.S
@@ -251,7 +251,7 @@
 	sarq	$4, I
 	jle	.L113
 
-#if defined(BARCELONA) || defined(SHANGHAI)
+#if defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 
 	movaps  %xmm0, %xmm1
 	mulpd	-16 * SIZE(X), %xmm1
diff --git a/kernel/x86_64/symv_L_sse.S b/kernel/x86_64/symv_L_sse.S
index 5a123d7f6..0f1ebd564 100644
--- a/kernel/x86_64/symv_L_sse.S
+++ b/kernel/x86_64/symv_L_sse.S
@@ -76,7 +76,7 @@
 #define movsd		movlps
 #endif
 
-#if defined(BARCELONA) || defined(SHANGHAI) || defined(BOBCAT)
+#if defined(BARCELONA) || defined(SHANGHAI) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCH	prefetch
 #define PREFETCHW	prefetchw
 #define PREFETCHSIZE	(16 * 16)
diff --git a/kernel/x86_64/symv_L_sse2.S b/kernel/x86_64/symv_L_sse2.S
index 8afdc87db..9dd123c52 100644
--- a/kernel/x86_64/symv_L_sse2.S
+++ b/kernel/x86_64/symv_L_sse2.S
@@ -76,7 +76,7 @@
 #define movsd		movlpd
 #endif
 
-#if defined(BARCELONA) || defined(SHANGHAI) || defined(BOBCAT)
+#if defined(BARCELONA) || defined(SHANGHAI) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCH	prefetch
 #define PREFETCHW	prefetchw
 #define PREFETCHSIZE	(16 * 16)
diff --git a/kernel/x86_64/symv_U_sse.S b/kernel/x86_64/symv_U_sse.S
index 5aef6b461..93a66aaa7 100644
--- a/kernel/x86_64/symv_U_sse.S
+++ b/kernel/x86_64/symv_U_sse.S
@@ -76,7 +76,7 @@
 #define movsd		movlps
 #endif
 
-#if defined(BARCELONA) || defined(SHANGHAI) || defined(BOBCAT)
+#if defined(BARCELONA) || defined(SHANGHAI) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCH	prefetch
 #define PREFETCHW	prefetchw
 #define PREFETCHSIZE	(16 * 16)
diff --git a/kernel/x86_64/symv_U_sse2.S b/kernel/x86_64/symv_U_sse2.S
index fa1bfba85..f412b3e2f 100644
--- a/kernel/x86_64/symv_U_sse2.S
+++ b/kernel/x86_64/symv_U_sse2.S
@@ -76,7 +76,7 @@
 #define movsd		movlpd
 #endif
 
-#if defined(BARCELONA) || defined(SHANGHAI) || defined(BOBCAT)
+#if defined(BARCELONA) || defined(SHANGHAI) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCH	prefetch
 #define PREFETCHW	prefetchw
 #define PREFETCHSIZE	(16 * 16)
diff --git a/kernel/x86_64/trsm_kernel_LN_8x4_sse.S b/kernel/x86_64/trsm_kernel_LN_8x4_sse.S
index 513572ee9..552dbacdc 100644
--- a/kernel/x86_64/trsm_kernel_LN_8x4_sse.S
+++ b/kernel/x86_64/trsm_kernel_LN_8x4_sse.S
@@ -86,7 +86,7 @@
 #define PREFETCHW    prefetcht0
 #endif
 
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 #define PREFETCH     prefetch
 #define PREFETCHW    prefetchw
 #define movsd movlps
diff --git a/kernel/x86_64/trsm_kernel_LT_8x4_sse.S b/kernel/x86_64/trsm_kernel_LT_8x4_sse.S
index 526a78c57..7727fd591 100644
--- a/kernel/x86_64/trsm_kernel_LT_8x4_sse.S
+++ b/kernel/x86_64/trsm_kernel_LT_8x4_sse.S
@@ -86,7 +86,7 @@
 #define PREFETCHW    prefetcht0
 #endif
 
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 #define PREFETCH     prefetch
 #define PREFETCHW    prefetchw
 #define movsd movlps
diff --git a/kernel/x86_64/trsm_kernel_RT_8x4_sse.S b/kernel/x86_64/trsm_kernel_RT_8x4_sse.S
index e96496fd6..699364941 100644
--- a/kernel/x86_64/trsm_kernel_RT_8x4_sse.S
+++ b/kernel/x86_64/trsm_kernel_RT_8x4_sse.S
@@ -86,7 +86,7 @@
 #define PREFETCHW    prefetcht0
 #endif
 
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 #define PREFETCH     prefetch
 #define PREFETCHW    prefetchw
 #define movsd movlps
diff --git a/kernel/x86_64/zgemm_ncopy_2.S b/kernel/x86_64/zgemm_ncopy_2.S
index bf318b7ff..8876b61ff 100644
--- a/kernel/x86_64/zgemm_ncopy_2.S
+++ b/kernel/x86_64/zgemm_ncopy_2.S
@@ -85,7 +85,7 @@
 #define movsd movlpd
 #endif
 
-#if defined(BARCELONA) || defined(SHANGHAI)
+#if defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 #define RPREFETCHSIZE 32
 #define WPREFETCHSIZE 48
 #endif
diff --git a/kernel/x86_64/zsymv_L_sse.S b/kernel/x86_64/zsymv_L_sse.S
index 6af65a4ba..dcfe83189 100644
--- a/kernel/x86_64/zsymv_L_sse.S
+++ b/kernel/x86_64/zsymv_L_sse.S
@@ -160,7 +160,7 @@
 #define a3     %xmm14
 #define	xt1    %xmm15
 
-#if (defined(HAVE_SSE3) && !defined(CORE_OPTERON)) || defined(BARCELONA) || defined(SHANGHAI) || defined(BOBCAT)
+#if (defined(HAVE_SSE3) && !defined(CORE_OPTERON)) || defined(BARCELONA) || defined(SHANGHAI) || defined(BOBCAT) || defined(BULLDOZER)
 #define MOVDDUP(a, b, c)	movddup	a(b), c
 #define MOVDDUP2(a, b, c)	movddup	a##b, c
 #else
diff --git a/kernel/x86_64/zsymv_L_sse2.S b/kernel/x86_64/zsymv_L_sse2.S
index 71aca0198..04605e3cb 100644
--- a/kernel/x86_64/zsymv_L_sse2.S
+++ b/kernel/x86_64/zsymv_L_sse2.S
@@ -76,7 +76,7 @@
 #define movsd		movlpd
 #endif
 
-#if defined(BARCELONA)  || defined(SHANGHAI) || defined(BOBCAT)
+#if defined(BARCELONA)  || defined(SHANGHAI) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCH	prefetch
 #define PREFETCHW	prefetchw
 #define PREFETCHSIZE	(16 * 16)
@@ -167,7 +167,7 @@
 #define a3     %xmm14
 #define	xt1    %xmm15
 
-#if (defined(HAVE_SSE3) && !defined(CORE_OPTERON)) || defined(BARCELONA) || defined(SHANGHAI)
+#if (defined(HAVE_SSE3) && !defined(CORE_OPTERON)) || defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 #define MOVDDUP(a, b, c)	movddup	a(b), c
 #define MOVDDUP2(a, b, c)	movddup	a##b, c
 #else
diff --git a/kernel/x86_64/zsymv_U_sse.S b/kernel/x86_64/zsymv_U_sse.S
index 4b8422d82..e8b01ad7a 100644
--- a/kernel/x86_64/zsymv_U_sse.S
+++ b/kernel/x86_64/zsymv_U_sse.S
@@ -76,7 +76,7 @@
 #define movsd		movlpd
 #endif
 
-#if defined(BARCELONA)  || defined(SHANGHAI) || defined(BOBCAT)
+#if defined(BARCELONA)  || defined(SHANGHAI) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCH	prefetch
 #define PREFETCHW	prefetchw
 #define PREFETCHSIZE	(16 * 16)
@@ -166,7 +166,7 @@
 #define	xt1    %xmm14
 #define	xt2    %xmm15
 
-#if (defined(HAVE_SSE3) && !defined(CORE_OPTERON)) || defined(BARCELONA) || defined(SHANGHAI)
+#if (defined(HAVE_SSE3) && !defined(CORE_OPTERON)) || defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 #define MOVDDUP(a, b, c)	movddup	a(b), c
 #define MOVDDUP2(a, b, c)	movddup	a##b, c
 #else
diff --git a/kernel/x86_64/zsymv_U_sse2.S b/kernel/x86_64/zsymv_U_sse2.S
index 33667f79e..40246e52e 100644
--- a/kernel/x86_64/zsymv_U_sse2.S
+++ b/kernel/x86_64/zsymv_U_sse2.S
@@ -76,7 +76,7 @@
 #define movsd		movlpd
 #endif
 
-#if defined(BARCELONA)  || defined(SHANGHAI) || defined(BOBCAT)
+#if defined(BARCELONA)  || defined(SHANGHAI) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCH	prefetch
 #define PREFETCHW	prefetchw
 #define PREFETCHSIZE	(16 * 16)
@@ -166,7 +166,7 @@
 #define a3     %xmm14
 #define	xt1    %xmm15
 
-#if (defined(HAVE_SSE3) && !defined(CORE_OPTERON)) || defined(BARCELONA) || defined(SHANGHAI)
+#if (defined(HAVE_SSE3) && !defined(CORE_OPTERON)) || defined(BARCELONA) || defined(SHANGHAI) || defined(BULLDOZER)
 #define MOVDDUP(a, b, c)	movddup	a(b), c
 #define MOVDDUP2(a, b, c)	movddup	a##b, c
 #else
diff --git a/kernel/x86_64/ztrsm_kernel_LN_2x2_sse2.S b/kernel/x86_64/ztrsm_kernel_LN_2x2_sse2.S
index b8caa9a44..79f20b641 100644
--- a/kernel/x86_64/ztrsm_kernel_LN_2x2_sse2.S
+++ b/kernel/x86_64/ztrsm_kernel_LN_2x2_sse2.S
@@ -86,7 +86,7 @@
 #define BORIG	 72(%rsp)
 #define BUFFER	128(%rsp)
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCH     prefetch
 #define PREFETCHW    prefetchw
 #define PREFETCHNTA  prefetchnta
diff --git a/kernel/x86_64/ztrsm_kernel_LN_4x2_sse.S b/kernel/x86_64/ztrsm_kernel_LN_4x2_sse.S
index 2db8cbc5d..f5c100ec1 100644
--- a/kernel/x86_64/ztrsm_kernel_LN_4x2_sse.S
+++ b/kernel/x86_64/ztrsm_kernel_LN_4x2_sse.S
@@ -95,7 +95,7 @@
 #define PREFETCHSIZE (8 * 6 + 4)
 #endif
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCH     prefetch
 #define PREFETCHW    prefetchw
 #define PREFETCHNTA  prefetchnta
diff --git a/kernel/x86_64/ztrsm_kernel_LT_2x2_sse2.S b/kernel/x86_64/ztrsm_kernel_LT_2x2_sse2.S
index 16c9ca828..18edeed57 100644
--- a/kernel/x86_64/ztrsm_kernel_LT_2x2_sse2.S
+++ b/kernel/x86_64/ztrsm_kernel_LT_2x2_sse2.S
@@ -86,7 +86,7 @@
 #define BORIG	 72(%rsp)
 #define BUFFER	128(%rsp)
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCH     prefetch
 #define PREFETCHW    prefetchw
 #define PREFETCHNTA  prefetchnta
diff --git a/kernel/x86_64/ztrsm_kernel_LT_4x2_sse.S b/kernel/x86_64/ztrsm_kernel_LT_4x2_sse.S
index dbdbfe2e1..f58cecdf5 100644
--- a/kernel/x86_64/ztrsm_kernel_LT_4x2_sse.S
+++ b/kernel/x86_64/ztrsm_kernel_LT_4x2_sse.S
@@ -95,7 +95,7 @@
 #define PREFETCHSIZE (8 * 6 + 4)
 #endif
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCH     prefetch
 #define PREFETCHW    prefetchw
 #define PREFETCHNTA  prefetchnta
diff --git a/kernel/x86_64/ztrsm_kernel_RT_2x2_sse2.S b/kernel/x86_64/ztrsm_kernel_RT_2x2_sse2.S
index 181cdd29c..1b589e0cf 100644
--- a/kernel/x86_64/ztrsm_kernel_RT_2x2_sse2.S
+++ b/kernel/x86_64/ztrsm_kernel_RT_2x2_sse2.S
@@ -86,7 +86,7 @@
 #define BORIG	 72(%rsp)
 #define BUFFER	128(%rsp)
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCH     prefetch
 #define PREFETCHW    prefetchw
 #define PREFETCHNTA  prefetchnta
diff --git a/kernel/x86_64/ztrsm_kernel_RT_4x2_sse.S b/kernel/x86_64/ztrsm_kernel_RT_4x2_sse.S
index c28d02927..2c47ce3fd 100644
--- a/kernel/x86_64/ztrsm_kernel_RT_4x2_sse.S
+++ b/kernel/x86_64/ztrsm_kernel_RT_4x2_sse.S
@@ -95,7 +95,7 @@
 #define PREFETCHSIZE (8 * 6 + 4)
 #endif
 
-#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BOBCAT)
+#if defined(OPTERON) || defined(BARCELONA) || defined(SHANGHAI) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCH     prefetch
 #define PREFETCHW    prefetchw
 #define PREFETCHNTA  prefetchnta
diff --git a/l1param.h b/l1param.h
index 6fe756f17..0b216c7c5 100644
--- a/l1param.h
+++ b/l1param.h
@@ -74,6 +74,13 @@
 #define ALIGNED_ACCESS
 #endif
 
+#ifdef BULLDOZER
+#define PREFETCH	prefetch
+#define PREFETCHW	prefetchw
+#define PREFETCHSIZE (128 *   5)
+#define ALIGNED_ACCESS
+#endif
+
 #ifdef NANO
 #define PREFETCH        prefetcht0
 #define PREFETCHW       prefetcht0
diff --git a/l2param.h b/l2param.h
index cdbd8805e..01fe7943d 100644
--- a/l2param.h
+++ b/l2param.h
@@ -85,7 +85,7 @@
 #define movsd		movlps
 #endif
 
-#if defined(BARCELONA) || defined(SHANGHAI) || defined(BOBCAT)
+#if defined(BARCELONA) || defined(SHANGHAI) || defined(BOBCAT) || defined(BULLDOZER)
 #define ALIGNED_ACCESS
 #define MOVUPS_A	movaps
 #define MOVUPS_XL	movaps
diff --git a/param.h b/param.h
index 11c1a269e..5b6a19ad5 100644
--- a/param.h
+++ b/param.h
@@ -143,7 +143,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #endif
 
-#if defined(BARCELONA) || defined(SHANGHAI) || defined(BOBCAT)
+#if defined(BARCELONA) || defined(SHANGHAI) || defined(BOBCAT) || defined(BULLDOZER)
 
 #define SNUMOPT		8
 #define DNUMOPT		4