Merge pull request #99 from xianyi/develop

rebase

Author: martin-frbg

Makefile.x86_64

@@ -12,6 +12,10 @@ ifdef HAVE_SSE3
 ifndef DYNAMIC_ARCH
 CCOMMON_OPT += -msse3
 FCOMMON_OPT += -msse3
+ifdef HAVE_SSSE3
+CCOMMON_OPT += -mssse3
+FCOMMON_OPT += -mssse3
+endif
 endif
 endif
 
@@ -60,7 +64,7 @@ endif
 endif
 endif
 
-ifeq ($(CORE), HASWELL)
+ifeq ($(CORE), $(filter $(CORE), HASWELL ZEN SKYLAKEX COOPERLAKE))
 ifndef DYNAMIC_ARCH
 ifndef NO_AVX2
 ifeq ($(C_COMPILER), GCC)

cmake/cc.cmake

@@ -109,10 +109,25 @@ if (${CORE} STREQUAL "COOPERLAKE")
     if (NOT NO_AVX512)
       execute_process(COMMAND ${CMAKE_C_COMPILER} -dumpversion OUTPUT_VARIABLE GCC_VERSION)
       if (${GCC_VERSION} VERSION_GREATER 10.1 OR ${GCC_VERSION} VERSION_EQUAL 10.1)
-        set (CCOMMON_OPT = "${CCOMMON_OPT} -march=cooperlake")
+        set (CCOMMON_OPT  "${CCOMMON_OPT} -march=cooperlake")
       else ()
         set (CCOMMON_OPT "${CCOMMON_OPT} -march=skylake-avx512")
       endif()  
     endif ()
   endif ()
 endif ()
+
+if (NOT DYNAMIC_ARCH)
+	if (HAVE_AVX2)
+        set (CCOMMON_OPT  "${CCOMMON_OPT} -mavx2")
+	endif ()
+	if (HAVE_AVX)
+        set (CCOMMON_OPT  "${CCOMMON_OPT} -mavx")
+	endif ()
+	if (HAVE_SSE3)
+        set (CCOMMON_OPT  "${CCOMMON_OPT} -msse3")
+	endif ()
+	if (HAVE_SSSE3)
+        set (CCOMMON_OPT  "${CCOMMON_OPT} -mssse3")
+	endif ()
+endif()

ctest/CMakeLists.txt

@@ -12,7 +12,7 @@ FILE(WRITE ${CMAKE_CURRENT_BINARY_DIR}/test_cblas_helper.sh
 foreach(float_type ${FLOAT_TYPES})
   string(SUBSTRING ${float_type} 0 1 float_char_upper)
   string(TOLOWER ${float_char_upper} float_char)
-  if (${float_char} STREQUAL "h")
+  if (${float_char} STREQUAL "b")
    continue()
   endif()
   #level1

kernel/Makefile

@@ -8,6 +8,9 @@ include $(TOPDIR)/Makefile.system
 ifdef HAVE_SSE3
 CFLAGS += -msse3
 endif
+ifdef HAVE_SSSE3
+CFLAGS += -mssse3
+endif
 
 ifeq ($(C_COMPILER), GCC)
 GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
@@ -41,8 +44,8 @@ ifdef NO_AVX2
 endif
 
 ifdef TARGET_CORE
-	ifeq ($(TARGET_CORE), $(filter $(TARGET_CORE),PRESCOTT CORE2 PENRYN DUNNINGTON ATOM NANO NEHALEM BARCELONA BOBCAT BULLDOZER PILEDRIVER EXCAVATOR STEAMROLLER OPTERON_SSE3))
-	override CFLAGS += -msse3
+	ifeq ($(TARGET_CORE), $(filter $(TARGET_CORE),PRESCOTT CORE2 PENRYN DUNNINGTON ATOM NANO SANDYBRIDGE HASWELL NEHALEM ZEN BARCELONA BOBCAT BULLDOZER PILEDRIVER EXCAVATOR STEAMROLLER OPTERON_SSE3))
+	override CFLAGS += -msse3 -mssse3
 endif
 ifeq ($(TARGET_CORE), COOPERLAKE)
  override CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE)

kernel/power/KERNEL.POWER10

@@ -9,10 +9,10 @@ else
 
 SBGEMM_BETA = ../generic/gemm_beta.c
 SBGEMMKERNEL    = sbgemm_kernel_power10.c
-SBGEMMINCOPY    = ../generic/gemm_ncopy_16.c
-SBGEMMITCOPY    = ../generic/gemm_tcopy_16.c
-SBGEMMONCOPY    = ../generic/gemm_ncopy_8.c
-SBGEMMOTCOPY    = ../generic/gemm_tcopy_8.c
+SBGEMMINCOPY    = sbgemm_ncopy_16_power10.c
+SBGEMMITCOPY    = sbgemm_tcopy_16_power10.c
+SBGEMMONCOPY    = sbgemm_ncopy_8_power10.c
+SBGEMMOTCOPY    = sbgemm_tcopy_8_power10.c
 SBGEMMINCOPYOBJ =  sbgemm_incopy$(TSUFFIX).$(SUFFIX)
 SBGEMMITCOPYOBJ =  sbgemm_itcopy$(TSUFFIX).$(SUFFIX)
 SBGEMMONCOPYOBJ =  sbgemm_oncopy$(TSUFFIX).$(SUFFIX)