summaryrefslogtreecommitdiffstats
path: root/contrib/llvm/lib/Target/X86/X86Subtarget.cpp
diff options
context:
space:
mode:
Diffstat (limited to 'contrib/llvm/lib/Target/X86/X86Subtarget.cpp')
-rw-r--r--contrib/llvm/lib/Target/X86/X86Subtarget.cpp131
1 files changed, 94 insertions, 37 deletions
diff --git a/contrib/llvm/lib/Target/X86/X86Subtarget.cpp b/contrib/llvm/lib/Target/X86/X86Subtarget.cpp
index 7064dd0..452dd7e 100644
--- a/contrib/llvm/lib/Target/X86/X86Subtarget.cpp
+++ b/contrib/llvm/lib/Target/X86/X86Subtarget.cpp
@@ -21,7 +21,6 @@
#include "llvm/Support/Host.h"
#include "llvm/Target/TargetMachine.h"
#include "llvm/Target/TargetOptions.h"
-#include "llvm/ADT/SmallVector.h"
#define GET_SUBTARGETINFO_TARGET_DESC
#define GET_SUBTARGETINFO_CTOR
@@ -177,16 +176,18 @@ unsigned X86Subtarget::getSpecialAddressLatency() const {
void X86Subtarget::AutoDetectSubtargetFeatures() {
unsigned EAX = 0, EBX = 0, ECX = 0, EDX = 0;
+ unsigned MaxLevel;
union {
unsigned u[3];
char c[12];
} text;
-
- if (X86_MC::GetCpuIDAndInfo(0, &EAX, text.u+0, text.u+2, text.u+1))
+
+ if (X86_MC::GetCpuIDAndInfo(0, &MaxLevel, text.u+0, text.u+2, text.u+1) ||
+ MaxLevel < 1)
return;
X86_MC::GetCpuIDAndInfo(0x1, &EAX, &EBX, &ECX, &EDX);
-
+
if ((EDX >> 15) & 1) { HasCMov = true; ToggleFeature(X86::FeatureCMOV); }
if ((EDX >> 23) & 1) { X86SSELevel = MMX; ToggleFeature(X86::FeatureMMX); }
if ((EDX >> 25) & 1) { X86SSELevel = SSE1; ToggleFeature(X86::FeatureSSE1); }
@@ -196,7 +197,7 @@ void X86Subtarget::AutoDetectSubtargetFeatures() {
if ((ECX >> 19) & 1) { X86SSELevel = SSE41; ToggleFeature(X86::FeatureSSE41);}
if ((ECX >> 20) & 1) { X86SSELevel = SSE42; ToggleFeature(X86::FeatureSSE42);}
// FIXME: AVX codegen support is not ready.
- //if ((ECX >> 28) & 1) { HasAVX = true; ToggleFeature(X86::FeatureAVX); }
+ //if ((ECX >> 28) & 1) { X86SSELevel = AVX; ToggleFeature(X86::FeatureAVX); }
bool IsIntel = memcmp(text.c, "GenuineIntel", 12) == 0;
bool IsAMD = !IsIntel && memcmp(text.c, "AuthenticAMD", 12) == 0;
@@ -244,28 +245,69 @@ void X86Subtarget::AutoDetectSubtargetFeatures() {
IsBTMemSlow = true;
ToggleFeature(X86::FeatureSlowBTMem);
}
+
// If it's Nehalem, unaligned memory access is fast.
+ // FIXME: Nehalem is family 6. Also include Westmere and later processors?
if (Family == 15 && Model == 26) {
IsUAMemFast = true;
ToggleFeature(X86::FeatureFastUAMem);
}
- X86_MC::GetCpuIDAndInfo(0x80000001, &EAX, &EBX, &ECX, &EDX);
- if ((EDX >> 29) & 0x1) {
- HasX86_64 = true;
- ToggleFeature(X86::Feature64Bit);
- }
- if ((ECX >> 5) & 0x1) {
- HasLZCNT = true;
- ToggleFeature(X86::FeatureLZCNT);
+ // Set processor type. Currently only Atom is detected.
+ if (Family == 6 && Model == 28) {
+ X86ProcFamily = IntelAtom;
+ ToggleFeature(X86::FeatureLeaForSP);
}
- if (IsAMD && ((ECX >> 6) & 0x1)) {
- HasSSE4A = true;
- ToggleFeature(X86::FeatureSSE4A);
+
+ unsigned MaxExtLevel;
+ X86_MC::GetCpuIDAndInfo(0x80000000, &MaxExtLevel, &EBX, &ECX, &EDX);
+
+ if (MaxExtLevel >= 0x80000001) {
+ X86_MC::GetCpuIDAndInfo(0x80000001, &EAX, &EBX, &ECX, &EDX);
+ if ((EDX >> 29) & 0x1) {
+ HasX86_64 = true;
+ ToggleFeature(X86::Feature64Bit);
+ }
+ if ((ECX >> 5) & 0x1) {
+ HasLZCNT = true;
+ ToggleFeature(X86::FeatureLZCNT);
+ }
+ if (IsAMD) {
+ if ((ECX >> 6) & 0x1) {
+ HasSSE4A = true;
+ ToggleFeature(X86::FeatureSSE4A);
+ }
+ if ((ECX >> 11) & 0x1) {
+ HasXOP = true;
+ ToggleFeature(X86::FeatureXOP);
+ }
+ if ((ECX >> 16) & 0x1) {
+ HasFMA4 = true;
+ ToggleFeature(X86::FeatureFMA4);
+ }
+ }
}
- if (IsAMD && ((ECX >> 16) & 0x1)) {
- HasFMA4 = true;
- ToggleFeature(X86::FeatureFMA4);
+ }
+
+ if (IsIntel && MaxLevel >= 7) {
+ if (!X86_MC::GetCpuIDAndInfoEx(0x7, 0x0, &EAX, &EBX, &ECX, &EDX)) {
+ if (EBX & 0x1) {
+ HasFSGSBase = true;
+ ToggleFeature(X86::FeatureFSGSBase);
+ }
+ if ((EBX >> 3) & 0x1) {
+ HasBMI = true;
+ ToggleFeature(X86::FeatureBMI);
+ }
+ // FIXME: AVX2 codegen support is not ready.
+ //if ((EBX >> 5) & 0x1) {
+ // X86SSELevel = AVX2;
+ // ToggleFeature(X86::FeatureAVX2);
+ //}
+ if ((EBX >> 8) & 0x1) {
+ HasBMI2 = true;
+ ToggleFeature(X86::FeatureBMI2);
+ }
}
}
}
@@ -274,6 +316,7 @@ X86Subtarget::X86Subtarget(const std::string &TT, const std::string &CPU,
const std::string &FS,
unsigned StackAlignOverride, bool is64Bit)
: X86GenSubtargetInfo(TT, CPU, FS)
+ , X86ProcFamily(Others)
, PICStyle(PICStyles::None)
, X86SSELevel(NoMMXSSE)
, X863DNowLevel(NoThreeDNow)
@@ -281,31 +324,35 @@ X86Subtarget::X86Subtarget(const std::string &TT, const std::string &CPU,
, HasX86_64(false)
, HasPOPCNT(false)
, HasSSE4A(false)
- , HasAVX(false)
, HasAES(false)
, HasCLMUL(false)
, HasFMA3(false)
, HasFMA4(false)
+ , HasXOP(false)
, HasMOVBE(false)
, HasRDRAND(false)
, HasF16C(false)
+ , HasFSGSBase(false)
, HasLZCNT(false)
, HasBMI(false)
+ , HasBMI2(false)
, IsBTMemSlow(false)
, IsUAMemFast(false)
, HasVectorUAMem(false)
, HasCmpxchg16b(false)
- , stackAlignment(8)
+ , UseLeaForSP(false)
+ , PostRAScheduler(false)
+ , stackAlignment(4)
// FIXME: this is a known good value for Yonah. How about others?
, MaxInlineSizeThreshold(128)
, TargetTriple(TT)
- , In64BitMode(is64Bit)
- , InNaClMode(false) {
+ , In64BitMode(is64Bit) {
// Determine default and user specified characteristics
+ std::string CPUName = CPU;
if (!FS.empty() || !CPU.empty()) {
- std::string CPUName = CPU;
if (CPUName.empty()) {
-#if defined (__x86_64__) || defined(__i386__)
+#if defined(i386) || defined(__i386__) || defined(__x86__) || defined(_M_IX86)\
+ || defined(__x86_64__) || defined(_M_AMD64) || defined (_M_X64)
CPUName = sys::getHostCPUName();
#else
CPUName = "generic";
@@ -325,6 +372,13 @@ X86Subtarget::X86Subtarget(const std::string &TT, const std::string &CPU,
// If feature string is not empty, parse features string.
ParseSubtargetFeatures(CPUName, FullFS);
} else {
+ if (CPUName.empty()) {
+#if defined (__x86_64__) || defined(__i386__)
+ CPUName = sys::getHostCPUName();
+#else
+ CPUName = "generic";
+#endif
+ }
// Otherwise, use CPUID to auto-detect feature set.
AutoDetectSubtargetFeatures();
@@ -333,7 +387,7 @@ X86Subtarget::X86Subtarget(const std::string &TT, const std::string &CPU,
HasX86_64 = true; ToggleFeature(X86::Feature64Bit);
HasCMov = true; ToggleFeature(X86::FeatureCMOV);
- if (!HasAVX && X86SSELevel < SSE2) {
+ if (X86SSELevel < SSE2) {
X86SSELevel = SSE2;
ToggleFeature(X86::FeatureSSE1);
ToggleFeature(X86::FeatureSSE2);
@@ -341,28 +395,22 @@ X86Subtarget::X86Subtarget(const std::string &TT, const std::string &CPU,
}
}
+ if (X86ProcFamily == IntelAtom) {
+ PostRAScheduler = true;
+ InstrItins = getInstrItineraryForCPU(CPUName);
+ }
+
// It's important to keep the MCSubtargetInfo feature bits in sync with
// target data structure which is shared with MC code emitter, etc.
if (In64BitMode)
ToggleFeature(X86::Mode64Bit);
- if (isTargetNaCl()) {
- InNaClMode = true;
- ToggleFeature(X86::ModeNaCl);
- }
-
- if (HasAVX)
- X86SSELevel = NoMMXSSE;
-
DEBUG(dbgs() << "Subtarget features: SSELevel " << X86SSELevel
<< ", 3DNowLevel " << X863DNowLevel
<< ", 64bit " << HasX86_64 << "\n");
assert((!In64BitMode || HasX86_64) &&
"64-bit code requested on a subtarget that doesn't support it!");
- if(EnableSegmentedStacks && !isTargetELF())
- report_fatal_error("Segmented stacks are only implemented on ELF.");
-
// Stack alignment is 16 bytes on Darwin, FreeBSD, Linux and Solaris (both
// 32 and 64 bit) and for all 64-bit targets.
if (StackAlignOverride)
@@ -371,3 +419,12 @@ X86Subtarget::X86Subtarget(const std::string &TT, const std::string &CPU,
isTargetSolaris() || In64BitMode)
stackAlignment = 16;
}
+
+bool X86Subtarget::enablePostRAScheduler(
+ CodeGenOpt::Level OptLevel,
+ TargetSubtargetInfo::AntiDepBreakMode& Mode,
+ RegClassVector& CriticalPathRCs) const {
+ Mode = TargetSubtargetInfo::ANTIDEP_CRITICAL;
+ CriticalPathRCs.clear();
+ return PostRAScheduler && OptLevel >= CodeGenOpt::Default;
+}
OpenPOWER on IntegriCloud