Merge pull request #7 from xianyi/develop

update

.travis.yml

@@ -17,22 +17,22 @@ matrix:
         - COMMON_FLAGS="DYNAMIC_ARCH=1 TARGET=NEHALEM NUM_THREADS=32"
       script:
         - set -e
-        - travis_wait 45 make QUIET_MAKE=1 $COMMON_FLAGS $BTYPE
+        - make QUIET_MAKE=1 $COMMON_FLAGS $BTYPE
         - make -C test $COMMON_FLAGS $BTYPE
         - make -C ctest $COMMON_FLAGS $BTYPE
         - make -C utest $COMMON_FLAGS $BTYPE
       env:
         - TARGET_BOX=LINUX64
         - BTYPE="BINARY=64"
 
-        #    - <<: *test-ubuntu
-        #      os: linux-ppc64le
-        #      before_script:
-        #        - COMMON_FLAGS="DYNAMIC_ARCH=1 TARGET=POWER8 NUM_THREADS=32"
-        #      env:
-        #        # for matrix annotation only
-        #        - TARGET_BOX=PPC64LE_LINUX
-        #        - BTYPE="BINARY=64 USE_OPENMP=1"
+    - <<: *test-ubuntu
+      os: linux-ppc64le
+      before_script:
+        - COMMON_FLAGS="DYNAMIC_ARCH=1 TARGET=POWER8 NUM_THREADS=32"
+      env:
+        # for matrix annotation only
+        - TARGET_BOX=PPC64LE_LINUX
+        - BTYPE="BINARY=64 USE_OPENMP=1"
 
     - <<: *test-ubuntu
       env:

Makefile.system

@@ -322,12 +322,13 @@ CCOMMON_OPT	+= -DMS_ABI
 endif
 
 ifeq ($(C_COMPILER), GCC)
-#Test for supporting MS_ABI
+#Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
 GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
 GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
+GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
 GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f2 -d.` \>= 7)
 ifeq ($(GCCVERSIONGT4), 1)
-# GCC Majar version > 4
+# GCC Major version > 4
 # It is compatible with MSVC ABI.
 CCOMMON_OPT	+= -DMS_ABI
 endif
@@ -554,8 +555,17 @@ endif
 ifeq ($(ARCH), power)
 DYNAMIC_CORE = POWER6
 DYNAMIC_CORE += POWER8
+ifneq ($(C_COMPILER), GCC)
 DYNAMIC_CORE += POWER9
 endif
+ifeq ($(C_COMPILER), GCC)
+ifeq ($(GCCVERSIONGT5), 1)
+DYNAMIC_CORE += POWER9
+else
+$(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
+endif
+endif
+endif
 
 # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
 ifndef DYNAMIC_CORE

cpuid_arm64.c

@@ -206,6 +206,33 @@ void get_subdirname(void)
 	printf("arm64");
 }
 
+void get_cpucount(void)
+{
+int n=0;
+
+#ifdef linux
+	FILE *infile;
+  	char buffer[2048], *p,*t;
+  	p = (char *) NULL ;
+
+  	infile = fopen("/proc/cpuinfo", "r");
+
+	while (fgets(buffer, sizeof(buffer), infile))
+	{
+
+		if (!strncmp("processor", buffer, 9))
+		n++;
+  	}
+
+  	fclose(infile);
+
+	printf("#define NUM_CORES %d\n",n);
+#endif
+
+}
+
+
+
 void get_cpuconfig(void)
 {
 
@@ -309,6 +336,7 @@ void get_cpuconfig(void)
 			printf("#define DTB_SIZE             4096     \n");
 			break;	
 	}
+	get_cpucount();
 }
 
 
@@ -351,5 +379,3 @@ void get_features(void)
 #endif
 	return;
 }
-
-

driver/others/dynamic_power.c

@@ -3,7 +3,9 @@
 
 extern gotoblas_t gotoblas_POWER6;
 extern gotoblas_t gotoblas_POWER8;
+#if (!defined C_GCC) || (GCC_VERSION >= 60000)
 extern gotoblas_t gotoblas_POWER9;
+#endif
 
 extern void openblas_warning(int verbose, const char *msg);
 
@@ -19,7 +21,9 @@ static char *corename[] = {
 char *gotoblas_corename(void) {
 	if (gotoblas == &gotoblas_POWER6)	return corename[1];
 	if (gotoblas == &gotoblas_POWER8)	return corename[2];
+#if (!defined C_GCC) || (GCC_VERSION >= 60000)
 	if (gotoblas == &gotoblas_POWER9)	return corename[3];
+#endif
 	return corename[0];
 }
 
@@ -29,8 +33,10 @@ static gotoblas_t *get_coretype(void) {
 		return &gotoblas_POWER6;
 	if (__builtin_cpu_is("power8"))
 		return &gotoblas_POWER8;
+#if (!defined C_GCC) || (GCC_VERSION >= 60000)
 	if (__builtin_cpu_is("power9"))
 		return &gotoblas_POWER9;
+#endif
 	return NULL;
 }
 
@@ -53,7 +59,9 @@ static gotoblas_t *force_coretype(char * coretype) {
 	{
 	case  1: return (&gotoblas_POWER6);
 	case  2: return (&gotoblas_POWER8);
+#if (!defined C_GCC) || (GCC_VERSION >= 60000)
 	case  3: return (&gotoblas_POWER9);
+#endif
 	default: return NULL;
 	}
 	snprintf(message, 128, "Core not found: %s\n", coretype);

kernel/Makefile.L3

@@ -24,9 +24,11 @@ ifeq ($(TARGET), LOONGSON3B)
 USE_TRMM = 1
 endif
 
-ifeq ($(CORE), GENERIC)
+ifneq ($(DYNAMIC_ARCH), 1)
+ifeq ($(TARGET), GENERIC)
 USE_TRMM = 1
 endif
+endif
 
 ifeq ($(CORE), HASWELL)
 USE_TRMM = 1

kernel/power/KERNEL.POWER8

@@ -89,14 +89,14 @@ ZTRSMKERNEL_RT	= ../generic/trsm_kernel_RT.c
 #SMINKERNEL   = ../arm/min.c
 #DMINKERNEL   = ../arm/min.c
 #
-ISAMAXKERNEL = isamax.c
+ISAMAXKERNEL = isamax_power8.S
 IDAMAXKERNEL = idamax.c
-ICAMAXKERNEL = icamax.c
+ICAMAXKERNEL = icamax_power8.S
 IZAMAXKERNEL = izamax.c
 #
-ISAMINKERNEL = isamin.c
+ISAMINKERNEL = isamin_power8.S
 IDAMINKERNEL = idamin.c
-ICAMINKERNEL = icamin.c
+ICAMINKERNEL = icamin_power8.S
 IZAMINKERNEL = izamin.c
 #
 #ISMAXKERNEL  = ../arm/imax.c
@@ -112,7 +112,7 @@ ZASUMKERNEL  = zasum.c
 #
 SAXPYKERNEL  = saxpy.c
 DAXPYKERNEL  = daxpy.c
-CAXPYKERNEL  = caxpy.c
+CAXPYKERNEL  = caxpy_power8.S
 ZAXPYKERNEL  = zaxpy.c
 #
 SCOPYKERNEL  = scopy.c

kernel/power/KERNEL.POWER9

@@ -89,14 +89,14 @@ ZTRSMKERNEL_RT	= ../generic/trsm_kernel_RT.c
 #SMINKERNEL   = ../arm/min.c
 #DMINKERNEL   = ../arm/min.c
 #
-ISAMAXKERNEL = isamax.c
+ISAMAXKERNEL = isamax_power9.S
 IDAMAXKERNEL = idamax.c
-ICAMAXKERNEL = icamax.c
+ICAMAXKERNEL = icamax_power9.S
 IZAMAXKERNEL = izamax.c
 #
-ISAMINKERNEL = isamin.c
+ISAMINKERNEL = isamin_power9.S
 IDAMINKERNEL = idamin.c
-ICAMINKERNEL = icamin.c
+ICAMINKERNEL = icamin_power9.S
 IZAMINKERNEL = izamin.c
 #
 #ISMAXKERNEL  = ../arm/imax.c
@@ -112,7 +112,7 @@ ZASUMKERNEL  = zasum.c
 #
 SAXPYKERNEL  = saxpy.c
 DAXPYKERNEL  = daxpy.c
-CAXPYKERNEL  = caxpy.c
+CAXPYKERNEL  = caxpy_power9.S
 ZAXPYKERNEL  = zaxpy.c
 #
 SCOPYKERNEL  = scopy.c
@@ -123,7 +123,7 @@ ZCOPYKERNEL  = zcopy.c
 SDOTKERNEL   =  sdot.c
 DDOTKERNEL   =  ddot.c
 DSDOTKERNEL  =  sdot.c
-CDOTKERNEL   =  cdot.c
+CDOTKERNEL   =  cdot_power9.S
 ZDOTKERNEL   =  zdot.c
 #
 SNRM2KERNEL  = ../arm/nrm2.c