Merge pull request #22 from xianyi/develop

rebase

.travis.yml

@@ -224,12 +224,21 @@ matrix:
       before_script:
         - COMMON_FLAGS="DYNAMIC_ARCH=1 NUM_THREADS=32"
         - brew update
-        - brew install gcc@10
       script:
         - travis_wait 45 make QUIET_MAKE=1 $COMMON_FLAGS $BTYPE
       env:
-        - BTYPE="TARGET=NEHALEM BINARY=64 INTERFACE64=1 FC=gfortran-10"
-        
+        - BTYPE="TARGET=HASWELL USE_OPENMP=1 BINARY=64 INTERFACE64=1 CC=gcc-10 FC=gfortran-10"
+
+    - <<: *test-macos
+      osx_image: xcode12
+      before_script:
+        - COMMON_FLAGS="DYNAMIC_ARCH=1 NUM_THREADS=32"
+        - brew update
+      script:
+        - travis_wait 45 make QUIET_MAKE=1 $COMMON_FLAGS $BTYPE
+      env:
+        - BTYPE="TARGET=NEHALEM BINARY=64 INTERFACE64=1 FC=gfortran-10"      
+
   #  - <<: *test-macos
   #    osx_image: xcode10
   #    env:

azure-pipelines.yml

@@ -68,4 +68,13 @@ jobs:
       dir
       openblas_utest.exe
   
+- job: OSX_OpenMP
+  pool:
+     vmImage: 'macOS-10.15'
+  steps:   
+  - script: |
+      brew update
+      make TARGET=CORE2 DYNAMIC_ARCH=1 USE_OPENMP=1 CC=gcc-10 FC=gfortran-10
+      
       
+       

benchmark/bench.h

@@ -3,6 +3,8 @@
 #include <time.h>
 #ifdef __CYGWIN32__
 #include <sys/time.h>
+#elif defined(__APPLE__)
+#include <mach/mach_time.h>
 #endif
 #include "common.h"
 

kernel/arm64/sgemm_tcopy_16.S

@@ -270,11 +270,6 @@ All rights reserved.
 	ldr	s1, [A02]
 	ldr	s2, [A03]
 	ldr	s3, [A04]
-	
-	add	A01, A01, #4
-	add	A02, A02, #4
-	add	A03, A03, #4
-	add	A04, A04, #4
 
 	stp	s0, s1, [B04]
 	add	B04, B04, #8
@@ -285,11 +280,6 @@ All rights reserved.
 	ldr	s5, [A06]
 	ldr	s6, [A07]
 	ldr	s7, [A08]
-	
-	ldr	d4, [A05], #8
-	ldr	d5, [A06], #8
-	ldr	d6, [A07], #8
-	ldr	d7, [A08], #8
 
 	stp	s4, s5, [B04]
 	add	B04, B04, #8

kernel/power/cdot.c

@@ -28,7 +28,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #else
 
 #include "common.h"
-#if defined(POWER10)
+#if defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 #include "cdot_microk_power10.c"
 #else
 #ifndef HAVE_KERNEL_8
@@ -120,7 +120,7 @@ OPENBLAS_COMPLEX_FLOAT CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLA
 
     if ((inc_x == 1) && (inc_y == 1)) {
 
-#if defined(POWER10)
+#if defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
         BLASLONG n1 = n & -16;
 #else
         BLASLONG n1 = n & -8;

kernel/power/cswap.c

@@ -39,8 +39,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #if defined(__VEC__) || defined(__ALTIVEC__)
 #if defined(POWER8)  || defined(POWER9)
 #include "cswap_microk_power8.c"
-#elif defined(POWER10)
+#elif defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 #include "cswap_microk_power10.c"
+#elif defined(POWER10)
+#include "cswap_microk_power8.c"
 #endif
 #endif
 

kernel/power/dasum.c

@@ -49,8 +49,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #if defined(__VEC__) || defined(__ALTIVEC__)
 #if defined(POWER8) || defined(POWER9)
 #include "dasum_microk_power8.c"
-#elif defined(POWER10)
+#elif defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 #include "dasum_microk_power10.c"
+#elif defined(POWER10)
+#include "dasum_microk_power8.c"
 #endif
 #endif
 
@@ -112,7 +114,7 @@ FLOAT CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x)
 	if ( inc_x == 1 )
 	{
 
-#if defined(POWER10)
+#if defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 		if ( n >= 16 )
 		{
 			BLASLONG align = ((32 - ((uintptr_t)x & (uintptr_t)0x1F)) >> 3) & 0x3;

kernel/power/drot.c

@@ -42,8 +42,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #if defined(__VEC__) || defined(__ALTIVEC__)
 #if defined(POWER8) || defined(POWER9)
 #include "drot_microk_power8.c"
-#elif defined(POWER10)
+#elif defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 #include "drot_microk_power10.c"
+#elif defined(POWER10)
+#include "drot_microk_power8.c"
 #endif
 #endif
 
@@ -117,7 +119,7 @@ int CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG inc_y, FLOAT
 	if ( (inc_x == 1) && (inc_y == 1) )
 	{
 
-#if defined(POWER10)
+#if defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 		if ( n >= 16 )
 		{
 			BLASLONG align = ((32 - ((uintptr_t)y & (uintptr_t)0x1F)) >> 3) & 0x3;

kernel/power/dscal.c

@@ -38,8 +38,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #if defined(__VEC__) || defined(__ALTIVEC__)
 #if defined(POWER8) || defined(POWER9)
 #include "dscal_microk_power8.c"
-#elif defined(POWER10)
+#elif defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 #include "dscal_microk_power10.c"
+#elif defined(POWER10)
+#include "dscal_microk_power8.c"
 #endif
 #endif
 
@@ -102,7 +104,7 @@ int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT da, FLOAT *x, BLAS
 		if ( da == 0.0 )
 		{		
 
-#if defined(POWER10)
+#if defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 			if ( n >= 16 )
 			{
 				BLASLONG align = ((32 - ((uintptr_t)x & (uintptr_t)0x1F)) >> 3) & 0x3;
@@ -136,7 +138,7 @@ int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT da, FLOAT *x, BLAS
 		else
 		{
 
-#if defined(POWER10)
+#if defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 			if ( n >= 16 )
 			{
 				BLASLONG align = ((32 - ((uintptr_t)x & (uintptr_t)0x1F)) >> 3) & 0x3;

kernel/power/dswap.c

@@ -38,8 +38,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #if defined(__VEC__) || defined(__ALTIVEC__)
 #if defined(POWER8) || defined(POWER9)
 #include "dswap_microk_power8.c"
-#elif defined(POWER10)
+#elif defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 #include "swap_microk_power10.c"
+#elif defined(POWER10)
+#include "dswap_microk_power8.c"
 #endif
 #endif
 
@@ -117,7 +119,7 @@ int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT dummy3, FLOAT *x,
 	if ( (inc_x == 1) && (inc_y == 1 ))
 	{
 
-#if defined(POWER10)
+#if defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 		if ( n >= 32 )
 		{
 			BLASLONG align = ((32 - ((uintptr_t)y & (uintptr_t)0x1F)) >> 3) & 0x3;

kernel/power/sasum.c

@@ -49,8 +49,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #if defined(__VEC__) || defined(__ALTIVEC__)
 #if defined(POWER8) || defined(POWER9)
 #include "sasum_microk_power8.c"
-#elif defined(POWER10)
+#elif defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 #include "sasum_microk_power10.c"
+#elif defined(POWER10)
+#include "sasum_microk_power8.c"
 #endif
 #endif
 
@@ -112,7 +114,7 @@ FLOAT CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x)
 	if ( inc_x == 1 )
 	{
 
-#if defined(POWER10)
+#if defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 		if ( n >= 32 )
 		{
 			BLASLONG align = ((32 - ((uintptr_t)x & (uintptr_t)0x1F)) >> 2) & 0x7;

kernel/power/srot.c

@@ -42,8 +42,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #if defined(__VEC__) || defined(__ALTIVEC__)
 #if defined(POWER8) || defined(POWER9)
 #include "srot_microk_power8.c"
-#elif defined(POWER10)
+#elif defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 #include "srot_microk_power10.c"
+#elif defined(POWER10)
+#include "srot_microk_power8.c"
 #endif
 #endif
 
@@ -117,7 +119,7 @@ int CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG inc_y, FLOAT
 	if ( (inc_x == 1) && (inc_y == 1) )
 	{
 
-#if defined(POWER10)
+#if defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 		if ( n >= 16 )
 		{
 			BLASLONG align = ((32 - ((uintptr_t)y & (uintptr_t)0x1F)) >> 2) & 0x7;

kernel/power/sscal.c

@@ -38,8 +38,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #if defined(__VEC__) || defined(__ALTIVEC__)
 #if defined(POWER8) || defined(POWER9)
 #include "sscal_microk_power8.c"
-#elif defined(POWER10)
+#elif defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 #include "sscal_microk_power10.c"
+#elif defined(POWER10)
+#include "sscal_microk_power8.c"
 #endif
 #endif
 
@@ -104,7 +106,7 @@ int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT da, FLOAT *x, BLAS
 		if ( da == 0.0 )
 		{		
 
-#if defined(POWER10)
+#if defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 			if ( n >= 32 )
 			{
 				BLASLONG align = ((32 - ((uintptr_t)x & (uintptr_t)0x1F)) >> 2) & 0x7;
@@ -138,7 +140,7 @@ int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT da, FLOAT *x, BLAS
 		else
 		{
 
-#if defined(POWER10)
+#if defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 			if ( n >= 32 )
 			{
 				BLASLONG align = ((32 - ((uintptr_t)x & (uintptr_t)0x1F)) >> 2) & 0x7;

kernel/power/sswap.c

@@ -38,8 +38,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #if defined(__VEC__) || defined(__ALTIVEC__)
 #if defined(POWER8) || defined(POWER9)
 #include "sswap_microk_power8.c"
-#elif defined(POWER10)
+#elif defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 #include "swap_microk_power10.c"
+#elif defined(POWER10)
+#include "sswap_microk_power8.c"
 #endif
 #endif
 
@@ -117,7 +119,7 @@ int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT dummy3, FLOAT *x,
 	if ( (inc_x == 1) && (inc_y == 1 ))
 	{
 
-#if defined(POWER10)
+#if defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 		if ( n >= 64 )
 		{
 			BLASLONG align = ((32 - ((uintptr_t)y & (uintptr_t)0x1F)) >> 2) & 0x7;

kernel/power/zscal.c

@@ -43,12 +43,16 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #if defined(DOUBLE)
 #include "zscal_microk_power8.c"
 #endif
-#elif defined(POWER10)
+#elif defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 #if defined(DOUBLE)
 #include "zscal_microk_power10.c"
 #else
 #include "cscal_microk_power10.c"
 #endif
+#elif defined(POWER10)
+#if defined(DOUBLE)
+#include "zscal_microk_power8.c"
+#endif
 #endif
 #endif
 

kernel/power/zswap.c

@@ -39,8 +39,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #if defined(__VEC__) || defined(__ALTIVEC__)
 #if defined(POWER8) || defined(POWER9)
 #include "zswap_microk_power8.c"
-#elif defined(POWER10)
+#elif defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 #include "cswap_microk_power10.c"
+#elif defined(POWER10)
+#include "zswap_microk_power8.c"
 #endif
 #endif