diff options
Diffstat (limited to 'lib/Target/X86/X86Subtarget.cpp')
-rw-r--r-- | lib/Target/X86/X86Subtarget.cpp | 48 |
1 files changed, 45 insertions, 3 deletions
diff --git a/lib/Target/X86/X86Subtarget.cpp b/lib/Target/X86/X86Subtarget.cpp index f26b0f9583..c6b5008f66 100644 --- a/lib/Target/X86/X86Subtarget.cpp +++ b/lib/Target/X86/X86Subtarget.cpp @@ -37,8 +37,7 @@ using namespace llvm; /// ClassifyBlockAddressReference - Classify a blockaddress reference for the /// current subtarget according to how we should reference it in a non-pcrel /// context. -unsigned char X86Subtarget:: -ClassifyBlockAddressReference() const { +unsigned char X86Subtarget::ClassifyBlockAddressReference() const { if (isPICStyleGOT()) // 32-bit ELF targets. return X86II::MO_GOTOFF; @@ -179,6 +178,26 @@ bool X86Subtarget::IsLegalToCallImmediateAddr(const TargetMachine &TM) const { // @LOCALMOD-END } +static bool OSHasAVXSupport() { +#if defined(i386) || defined(__i386__) || defined(__x86__) || defined(_M_IX86)\ + || defined(__x86_64__) || defined(_M_AMD64) || defined (_M_X64) +#if defined(__GNUC__) + // Check xgetbv; this uses a .byte sequence instead of the instruction + // directly because older assemblers do not include support for xgetbv and + // there is no easy way to conditionally compile based on the assembler used. + int rEAX, rEDX; + __asm__ (".byte 0x0f, 0x01, 0xd0" : "=a" (rEAX), "=d" (rEDX) : "c" (0)); +#elif defined(_MSC_FULL_VER) && defined(_XCR_XFEATURE_ENABLED_MASK) + unsigned long long rEAX = _xgetbv(_XCR_XFEATURE_ENABLED_MASK); +#else + int rEAX = 0; // Ensures we return false +#endif + return (rEAX & 6) == 6; +#else + return false; +#endif +} + void X86Subtarget::AutoDetectSubtargetFeatures() { unsigned EAX = 0, EBX = 0, ECX = 0, EDX = 0; unsigned MaxLevel; @@ -201,7 +220,9 @@ void X86Subtarget::AutoDetectSubtargetFeatures() { if ((ECX >> 9) & 1) { X86SSELevel = SSSE3; ToggleFeature(X86::FeatureSSSE3);} if ((ECX >> 19) & 1) { X86SSELevel = SSE41; ToggleFeature(X86::FeatureSSE41);} if ((ECX >> 20) & 1) { X86SSELevel = SSE42; ToggleFeature(X86::FeatureSSE42);} - if ((ECX >> 28) & 1) { X86SSELevel = AVX; ToggleFeature(X86::FeatureAVX); } + if (((ECX >> 27) & 1) && ((ECX >> 28) & 1) && OSHasAVXSupport()) { + X86SSELevel = AVX; ToggleFeature(X86::FeatureAVX); + } bool IsIntel = memcmp(text.c, "GenuineIntel", 12) == 0; bool IsAMD = !IsIntel && memcmp(text.c, "AuthenticAMD", 12) == 0; @@ -291,6 +312,10 @@ void X86Subtarget::AutoDetectSubtargetFeatures() { HasLZCNT = true; ToggleFeature(X86::FeatureLZCNT); } + if (IsIntel && ((ECX >> 8) & 0x1)) { + HasPRFCHW = true; + ToggleFeature(X86::FeaturePRFCHW); + } if (IsAMD) { if ((ECX >> 6) & 0x1) { HasSSE4A = true; @@ -318,6 +343,10 @@ void X86Subtarget::AutoDetectSubtargetFeatures() { HasBMI = true; ToggleFeature(X86::FeatureBMI); } + if ((EBX >> 4) & 0x1) { + HasHLE = true; + ToggleFeature(X86::FeatureHLE); + } if (IsIntel && ((EBX >> 5) & 0x1)) { X86SSELevel = AVX2; ToggleFeature(X86::FeatureAVX2); @@ -330,6 +359,14 @@ void X86Subtarget::AutoDetectSubtargetFeatures() { HasRTM = true; ToggleFeature(X86::FeatureRTM); } + if (IsIntel && ((EBX >> 19) & 0x1)) { + HasADX = true; + ToggleFeature(X86::FeatureADX); + } + if (IsIntel && ((EBX >> 18) & 0x1)) { + HasRDSEED = true; + ToggleFeature(X86::FeatureRDSEED); + } } } } @@ -448,7 +485,10 @@ void X86Subtarget::initializeEnvironment() { HasBMI = false; HasBMI2 = false; HasRTM = false; + HasHLE = false; HasADX = false; + HasPRFCHW = false; + HasRDSEED = false; IsBTMemSlow = false; IsUAMemFast = false; HasVectorUAMem = false; @@ -457,6 +497,8 @@ void X86Subtarget::initializeEnvironment() { HasSlowDivide = false; PostRAScheduler = false; PadShortFunctions = false; + CallRegIndirect = false; + LEAUsesAG = false; stackAlignment = 4; // FIXME: this is a known good value for Yonah. How about others? MaxInlineSizeThreshold = 128; |