| Message ID | 20241031045333.1209195-1-seanjc@google.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | KVM: selftests: Don't force -march=x86-64-v2 if it's unsupported | expand |
Sean Christopherson <seanjc@google.com> writes: > Force -march=x86-64-v2 to avoid SSE/AVX instructions if and only if the > uarch definition is supported by the compiler, e.g. gcc 7.5 only supports > x86-64. > > Fixes: 9a400068a158 ("KVM: selftests: x86: Avoid using SSE/AVX instructions") > Cc: Vitaly Kuznetsov <vkuznets@redhat.com> > Signed-off-by: Sean Christopherson <seanjc@google.com> > --- > > My big ol' AVX enabling series ain't looking so bad now ;-) Smaller the change, bigger the troubles :-) > > tools/testing/selftests/kvm/Makefile | 2 ++ > 1 file changed, 2 insertions(+) > > diff --git a/tools/testing/selftests/kvm/Makefile b/tools/testing/selftests/kvm/Makefile > index 156fbfae940f..5fa282643cff 100644 > --- a/tools/testing/selftests/kvm/Makefile > +++ b/tools/testing/selftests/kvm/Makefile > @@ -249,8 +249,10 @@ ifeq ($(ARCH),s390) > CFLAGS += -march=z10 > endif > ifeq ($(ARCH),x86) > +ifeq ($(shell echo "void foo(void) { }" | $(CC) -march=x86-64-v2 -x c - -c -o /dev/null 2>/dev/null; echo "$$?"),0) With this test, the outer "ifeq ($(ARCH),x86)" check is not really needed anymore I guess? > CFLAGS += -march=x86-64-v2 > endif > +endif > ifeq ($(ARCH),arm64) > tools_dir := $(top_srcdir)/tools > arm64_tools_dir := $(tools_dir)/arch/arm64/tools/ > > base-commit: 81983758430957d9a5cb3333fe324fd70cf63e7e Reviewed-and-tested-by: Vitaly Kuznetsov <vkuznets@redhat.com>
On Thu, Oct 31, 2024, Vitaly Kuznetsov wrote: > > diff --git a/tools/testing/selftests/kvm/Makefile b/tools/testing/selftests/kvm/Makefile > > index 156fbfae940f..5fa282643cff 100644 > > --- a/tools/testing/selftests/kvm/Makefile > > +++ b/tools/testing/selftests/kvm/Makefile > > @@ -249,8 +249,10 @@ ifeq ($(ARCH),s390) > > CFLAGS += -march=z10 > > endif > > ifeq ($(ARCH),x86) > > +ifeq ($(shell echo "void foo(void) { }" | $(CC) -march=x86-64-v2 -x c - -c -o /dev/null 2>/dev/null; echo "$$?"),0) > > With this test, the outer "ifeq ($(ARCH),x86)" check is not really > needed anymore I guess? Yeah, but at this point in the v6.12 cycle, I want to do the absolute bare minimum. In 6.13, after the AVX support lands, I think we can/should revert this restriction, i.e. it's a short-term wart.
diff --git a/tools/testing/selftests/kvm/Makefile b/tools/testing/selftests/kvm/Makefile index 156fbfae940f..5fa282643cff 100644 --- a/tools/testing/selftests/kvm/Makefile +++ b/tools/testing/selftests/kvm/Makefile @@ -249,8 +249,10 @@ ifeq ($(ARCH),s390) CFLAGS += -march=z10 endif ifeq ($(ARCH),x86) +ifeq ($(shell echo "void foo(void) { }" | $(CC) -march=x86-64-v2 -x c - -c -o /dev/null 2>/dev/null; echo "$$?"),0) CFLAGS += -march=x86-64-v2 endif +endif ifeq ($(ARCH),arm64) tools_dir := $(top_srcdir)/tools arm64_tools_dir := $(tools_dir)/arch/arm64/tools/
Force -march=x86-64-v2 to avoid SSE/AVX instructions if and only if the uarch definition is supported by the compiler, e.g. gcc 7.5 only supports x86-64. Fixes: 9a400068a158 ("KVM: selftests: x86: Avoid using SSE/AVX instructions") Cc: Vitaly Kuznetsov <vkuznets@redhat.com> Signed-off-by: Sean Christopherson <seanjc@google.com> --- My big ol' AVX enabling series ain't looking so bad now ;-) tools/testing/selftests/kvm/Makefile | 2 ++ 1 file changed, 2 insertions(+) base-commit: 81983758430957d9a5cb3333fe324fd70cf63e7e