Formalize the notion that AVX and SSE are non-overlapping extensions from the compiler's point of view. Per email discussion, we either want to always use VEX-prefixed instructions or never use them, and are taking "HasAVX" to mean "Always use VEX". Passing -mattr=-avx,+sse42 should serve to restore legacy SSE support when desirable.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@121439 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/lib/Target/X86/X86Subtarget.cpp b/lib/Target/X86/X86Subtarget.cpp
index 5607c13..c251984 100644
--- a/lib/Target/X86/X86Subtarget.cpp
+++ b/lib/Target/X86/X86Subtarget.cpp
@@ -256,13 +256,13 @@
if ((ECX >> 9) & 1) X86SSELevel = SSSE3;
if ((ECX >> 19) & 1) X86SSELevel = SSE41;
if ((ECX >> 20) & 1) X86SSELevel = SSE42;
+ if ((ECX >> 28) & 1) { HasAVX = true; X86SSELevel = NoMMXSSE; }
bool IsIntel = memcmp(text.c, "GenuineIntel", 12) == 0;
bool IsAMD = !IsIntel && memcmp(text.c, "AuthenticAMD", 12) == 0;
HasCLMUL = IsIntel && ((ECX >> 1) & 0x1);
HasFMA3 = IsIntel && ((ECX >> 12) & 0x1);
- HasAVX = ((ECX >> 28) & 0x1);
HasAES = IsIntel && ((ECX >> 25) & 0x1);
if (IsIntel || IsAMD) {
@@ -316,11 +316,13 @@
ParseSubtargetFeatures(FS, CPU);
// All X86-64 CPUs also have SSE2, however user might request no SSE via
// -mattr, so don't force SSELevel here.
+ if (HasAVX)
+ X86SSELevel = NoMMXSSE;
} else {
// Otherwise, use CPUID to auto-detect feature set.
AutoDetectSubtargetFeatures();
// Make sure SSE2 is enabled; it is available on all X86-64 CPUs.
- if (Is64Bit && X86SSELevel < SSE2)
+ if (Is64Bit && !HasAVX && X86SSELevel < SSE2)
X86SSELevel = SSE2;
}