Merge pull request #1607 from martin-frbg/dynarch

Move some x86_64 DYNAMIC_ARCH targets to new DYNAMIC_OLDER option

Makefile

@@ -153,6 +153,9 @@ ifeq ($(DYNAMIC_ARCH), 1)
 	do  $(MAKE) GOTOBLAS_MAKEFILE= -C kernel TARGET_CORE=$$d kernel || exit 1 ;\
 	done
 	@echo DYNAMIC_ARCH=1 >> Makefile.conf_last
+ifeq ($(DYNAMIC_OLDER), 1)
+	@echo DYNAMIC_OLDER=1 >> Makefile.conf_last
+endif	
 endif
 ifdef USE_THREAD
 	@echo USE_THREAD=$(USE_THREAD) >>  Makefile.conf_last

Makefile.install

@@ -98,7 +98,7 @@ endif
 	@echo Generating openblas.pc in "$(DESTDIR)$(OPENBLAS_PKGCONFIG_DIR)"
 	@echo 'libdir='$(OPENBLAS_LIBRARY_DIR) > "$(DESTDIR)$(OPENBLAS_PKGCONFIG_DIR)/openblas.pc"
 	@echo 'includedir='$(OPENBLAS_INCLUDE_DIR) >> "$(DESTDIR)$(OPENBLAS_PKGCONFIG_DIR)/openblas.pc"
-	@echo 'openblas_config= USE_64BITINT='$(USE_64BITINT) 'DYNAMIC_ARCH='$(DYNAMIC_ARCH) 'NO_CBLAS='$(NO_CBLAS) 'NO_LAPACK='$(NO_LAPACK) 'NO_LAPACKE='$(NO_LAPACKE) 'NO_AFFINITY='$(NO_AFFINITY) 'USE_OPENMP='$(USE_OPENMP) $(CORE) 'MAX_THREADS='$(NUM_THREADS)>> "$(DESTDIR)$(OPENBLAS_PKGCONFIG_DIR)/openblas.pc"
+	@echo 'openblas_config= USE_64BITINT='$(USE_64BITINT) 'DYNAMIC_ARCH='$(DYNAMIC_ARCH) 'DYNAMIC_OLDER='$(DYNAMIC_OLDER) 'NO_CBLAS='$(NO_CBLAS) 'NO_LAPACK='$(NO_LAPACK) 'NO_LAPACKE='$(NO_LAPACKE) 'NO_AFFINITY='$(NO_AFFINITY) 'USE_OPENMP='$(USE_OPENMP) $(CORE) 'MAX_THREADS='$(NUM_THREADS)>> "$(DESTDIR)$(OPENBLAS_PKGCONFIG_DIR)/openblas.pc"
 	@echo 'version='$(VERSION) >> "$(DESTDIR)$(OPENBLAS_PKGCONFIG_DIR)/openblas.pc"
 	@echo 'extralib='$(EXTRALIB) >> "$(DESTDIR)$(OPENBLAS_PKGCONFIG_DIR)/openblas.pc"
 	@cat openblas.pc.in >> "$(DESTDIR)$(OPENBLAS_PKGCONFIG_DIR)/openblas.pc"

Makefile.rule

@@ -17,6 +17,11 @@ VERSION = 0.3.1.dev
 # If you want to support multiple architecture in one binary
 # DYNAMIC_ARCH = 1
 
+# If you want the full list of x86_64 architectures supported in DYNAMIC_ARCH
+# mode (including individual optimizied codes for PENRYN, DUNNINGTON, OPTERON,
+# OPTERON_SSE3, ATOM and NANO rather than fallbacks to older architectures)
+# DYNAMIC_OLDER = 1
+
 # C compiler including binary type(32bit / 64bit). Default is gcc.
 # Don't use Intel Compiler or PGI, it won't generate right codes as I expect.
 # CC = gcc

Makefile.system

@@ -472,7 +472,18 @@ DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
 endif
 
 ifeq ($(ARCH), x86_64)
-DYNAMIC_CORE = PRESCOTT CORE2 PENRYN DUNNINGTON NEHALEM OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
+DYNAMIC_CORE = PRESCOTT CORE2 
+ifeq ($(DYNAMIC_OLDER), 1)
+DYNAMIC_CORE += PENRYN DUNNINGTON 
+endif
+DYNAMIC_CORE += NEHALEM 
+ifeq ($(DYNAMIC_OLDER), 1)
+DYNAMIC_CORE += OPTERON OPTERON_SSE3 
+endif
+DYNAMIC_CORE += BARCELONA 
+ifeq ($(DYNAMIC_OLDER), 1)
+DYNAMIC_CORE += BOBCAT ATOM NANO
+endif
 ifneq ($(NO_AVX), 1)
 DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
 endif
@@ -917,6 +928,10 @@ ifeq ($(DYNAMIC_ARCH), 1)
 CCOMMON_OPT	+= -DDYNAMIC_ARCH
 endif
 
+ifeq ($(DYNAMIC_OLDER), 1)
+CCOMMON_OPT	+= -DDYNAMIC_OLDER
+endif
+
 ifeq ($(NO_LAPACK), 1)
 CCOMMON_OPT	+= -DNO_LAPACK
 #Disable LAPACK C interface

cmake/arch.cmake

@@ -49,7 +49,18 @@ if (DYNAMIC_ARCH)
   endif ()
 
   if (X86_64)
-    set(DYNAMIC_CORE PRESCOTT CORE2 PENRYN DUNNINGTON NEHALEM OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO)
+    set(DYNAMIC_CORE PRESCOTT CORE2)
+    if (DYNAMIC_OLDER)
+	set (DYNAMIC_CORE ${DYNAMIC_CORE} PENRYN DUNNINGTON)
+    endif ()
+    set (DYNAMIC_CORE ${DYNAMIC_CORE} NEHALEM)
+    if (DYNAMIC_OLDER)
+	set (DYNAMIC_CORE ${DYNAMIC_CORE} OPTERON OPTERON_SSE3)
+    endif ()
+    set (DYNAMIC_CORE ${DYNAMIC_CORE} BARCELONA) 
+    if (DYNAMIC_OLDER)
+	set (DYNAMIC_CORE ${DYNAMIC_CORE} BOBCAT ATOM NANO)
+    endif ()
     if (NOT NO_AVX)
       set(DYNAMIC_CORE ${DYNAMIC_CORE} SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR)
     endif ()

cmake/openblas.pc.in

@@ -1,7 +1,7 @@
 libdir=@CMAKE_INSTALL_FULL_LIBDIR@
 includedir=@CMAKE_INSTALL_FULL_INCLUDEDIR@
 
-openblas_config=USE_64BITINT=@USE_64BITINT@ NO_CBLAS=@NO_CBLAS@ NO_LAPACK=@NO_LAPACK@ NO_LAPACKE=@NO_LAPACKE@ DYNAMIC_ARCH=@DYNAMIC_ARCH@ NO_AFFINITY=@NO_AFFINITY@ USE_OPENMP=@USE_OPENMP@ @CORE@ MAX_THREADS=@NUM_THREADS@ 
+openblas_config=USE_64BITINT=@USE_64BITINT@ NO_CBLAS=@NO_CBLAS@ NO_LAPACK=@NO_LAPACK@ NO_LAPACKE=@NO_LAPACKE@ DYNAMIC_ARCH=@DYNAMIC_ARCH@ DYNAMIC_OLDER=@DYNAMIC_OLDER@ NO_AFFINITY=@NO_AFFINITY@ USE_OPENMP=@USE_OPENMP@ @CORE@ MAX_THREADS=@NUM_THREADS@ 
 Name: OpenBLAS
 Description: OpenBLAS is an optimized BLAS library based on GotoBLAS2 1.13 BSD version
 Version: @OPENBLAS_VERSION@

cmake/system.cmake

@@ -163,6 +163,9 @@ endif ()
 
 if (DYNAMIC_ARCH)
   set(CCOMMON_OPT "${CCOMMON_OPT} -DDYNAMIC_ARCH")
+  if (DYNAMIC_OLDER)
+    set(CCOMMON_OPT "${CCOMMON_OPT} -DDYNAMIC_OLDER")
+  endif ()
 endif ()
 
 if (NO_LAPACK)

driver/others/dynamic.c

@@ -56,16 +56,27 @@ EXTERN gotoblas_t  gotoblas_BANIAS;
 EXTERN gotoblas_t  gotoblas_ATHLON;
 
 extern gotoblas_t  gotoblas_PRESCOTT;
+extern gotoblas_t  gotoblas_CORE2;
+extern gotoblas_t  gotoblas_NEHALEM;
+extern gotoblas_t  gotoblas_BARCELONA;
+#ifdef DYNAMIC_OLDER
 extern gotoblas_t  gotoblas_ATOM;
 extern gotoblas_t  gotoblas_NANO;
-extern gotoblas_t  gotoblas_CORE2;
 extern gotoblas_t  gotoblas_PENRYN;
 extern gotoblas_t  gotoblas_DUNNINGTON;
-extern gotoblas_t  gotoblas_NEHALEM;
 extern gotoblas_t  gotoblas_OPTERON;
 extern gotoblas_t  gotoblas_OPTERON_SSE3;
-extern gotoblas_t  gotoblas_BARCELONA;
 extern gotoblas_t  gotoblas_BOBCAT;
+#else
+#define gotoblas_ATOM gotoblas_NEHALEM
+#define gotoblas_NANO gotoblas_NEHALEM
+#define gotoblas_PENRYN gotoblas_CORE2
+#define gotoblas_DUNNINGTON gotoblas_CORE2
+#define gotoblas_OPTERON gotoblas_CORE2
+#define gotoblas_OPTERON_SSE3 gotoblas_CORE2
+#define gotoblas_BOBCAT gotoblas_CORE2
+#endif
+
 #ifndef NO_AVX
 extern gotoblas_t  gotoblas_SANDYBRIDGE;
 extern gotoblas_t  gotoblas_BULLDOZER;