floraachy
/
OpenBLAS
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add -mavx2 on Haswell only if the compiler supports it

This commit is contained in:
Martin Kroeker 2018-12-16 13:09:19 +01:00 committed by GitHub
parent 69d206440a
commit 545c2b1bbb
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 23 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

25
kernel/Makefile
View File

@ -5,6 +5,27 @@ endif
TOPDIR = .. TOPDIR = ..
include $(TOPDIR)/Makefile.system include $(TOPDIR)/Makefile.system
AVX2OPT =
ifeq ($(C_COMPILER), GCC)
# AVX2 support was added in 4.7.0
GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f2 -d.` \>= 7)
ifeq ($(GCCVERSIONGTEQ4)$(GCCMINORVERSIONGTEQ7), 11)
AVX2OPT = -mavx2
endif
endif
ifeq ($(C_COMPILER), CLANG)
# Any clang posing as gcc 4.2 should be new enough (3.4 or later)
GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) -dumpversion | cut -f2 -d.` \>= 2)
ifeq ($(GCCVERSIONGTEQ4)$(GCCMINORVERSIONGTEQ2), 11)
AVX2OPT -mavx2
endif
endif
ifdef NO_AVX2
AVX2OPT=
endif
ifdef TARGET_CORE ifdef TARGET_CORE
ifeq ($(TARGET_CORE), SKYLAKEX) ifeq ($(TARGET_CORE), SKYLAKEX)
override CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE) -march=skylake-avx512 override CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE) -march=skylake-avx512
@ -17,9 +38,9 @@ ifeq ($(TARGET_CORE), SKYLAKEX)
endif endif
endif endif
else ifeq($(TARGET_CORE), HASWELL) else ifeq($(TARGET_CORE), HASWELL)
override CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE) -mavx2 override CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE) $(AVX2OPT)
else else
override CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE) override CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE)
endif endif
BUILD_KERNEL = 1 BUILD_KERNEL = 1
KDIR = KDIR =