aboutsummaryrefslogtreecommitdiffstats
path: root/lib/Target/R600/AMDGPUSubtarget.cpp
diff options
context:
space:
mode:
authorStephen Hines <srhines@google.com>2015-03-23 12:10:34 -0700
committerStephen Hines <srhines@google.com>2015-03-23 12:10:34 -0700
commitebe69fe11e48d322045d5949c83283927a0d790b (patch)
treec92f1907a6b8006628a4b01615f38264d29834ea /lib/Target/R600/AMDGPUSubtarget.cpp
parentb7d2e72b02a4cb8034f32f8247a2558d2434e121 (diff)
downloadexternal_llvm-ebe69fe11e48d322045d5949c83283927a0d790b.zip
external_llvm-ebe69fe11e48d322045d5949c83283927a0d790b.tar.gz
external_llvm-ebe69fe11e48d322045d5949c83283927a0d790b.tar.bz2
Update aosp/master LLVM for rebase to r230699.
Change-Id: I2b5be30509658cb8266be782de0ab24f9099f9b9
Diffstat (limited to 'lib/Target/R600/AMDGPUSubtarget.cpp')
-rw-r--r--lib/Target/R600/AMDGPUSubtarget.cpp75
1 files changed, 49 insertions, 26 deletions
diff --git a/lib/Target/R600/AMDGPUSubtarget.cpp b/lib/Target/R600/AMDGPUSubtarget.cpp
index 9d09a19..70c8525 100644
--- a/lib/Target/R600/AMDGPUSubtarget.cpp
+++ b/lib/Target/R600/AMDGPUSubtarget.cpp
@@ -16,11 +16,11 @@
#include "R600ISelLowering.h"
#include "R600InstrInfo.h"
#include "R600MachineScheduler.h"
-#include "SIInstrInfo.h"
#include "SIISelLowering.h"
+#include "SIInstrInfo.h"
+#include "SIMachineFunctionInfo.h"
#include "llvm/ADT/SmallString.h"
-
-#include "llvm/ADT/SmallString.h"
+#include "llvm/CodeGen/MachineScheduler.h"
using namespace llvm;
@@ -31,22 +31,9 @@ using namespace llvm;
#define GET_SUBTARGETINFO_CTOR
#include "AMDGPUGenSubtargetInfo.inc"
-static std::string computeDataLayout(const AMDGPUSubtarget &ST) {
- std::string Ret = "e-p:32:32";
-
- if (ST.is64bit()) {
- // 32-bit private, local, and region pointers. 64-bit global and constant.
- Ret += "-p1:64:64-p2:64:64-p3:32:32-p4:64:64-p5:32:32-p24:64:64";
- }
-
- Ret += "-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256"
- "-v512:512-v1024:1024-v2048:2048-n32:64";
-
- return Ret;
-}
-
AMDGPUSubtarget &
-AMDGPUSubtarget::initializeSubtargetDependencies(StringRef GPU, StringRef FS) {
+AMDGPUSubtarget::initializeSubtargetDependencies(StringRef TT, StringRef GPU,
+ StringRef FS) {
// Determine default and user-specified characteristics
// On SI+, we want FP64 denormals to be on by default. FP32 denormals can be
// enabled, but some instructions do not respect them and they run at the
@@ -59,6 +46,9 @@ AMDGPUSubtarget::initializeSubtargetDependencies(StringRef GPU, StringRef FS) {
SmallString<256> FullFS("+promote-alloca,+fp64-denormals,");
FullFS += FS;
+ if (GPU == "" && Triple(TT).getArch() == Triple::amdgcn)
+ GPU = "SI";
+
ParseSubtargetFeatures(GPU, FullFS);
// FIXME: I don't think think Evergreen has any useful support for
@@ -76,21 +66,24 @@ AMDGPUSubtarget::AMDGPUSubtarget(StringRef TT, StringRef GPU, StringRef FS,
: AMDGPUGenSubtargetInfo(TT, GPU, FS), DevName(GPU), Is64bit(false),
DumpCode(false), R600ALUInst(false), HasVertexCache(false),
TexVTXClauseSize(0), Gen(AMDGPUSubtarget::R600), FP64(false),
- FP64Denormals(false), FP32Denormals(false), CaymanISA(false),
- FlatAddressSpace(false), EnableIRStructurizer(true),
- EnablePromoteAlloca(false), EnableIfCvt(true),
- EnableLoadStoreOpt(false), WavefrontSize(0), CFALUBug(false), LocalMemorySize(0),
- DL(computeDataLayout(initializeSubtargetDependencies(GPU, FS))),
+ FP64Denormals(false), FP32Denormals(false), FastFMAF32(false),
+ CaymanISA(false), FlatAddressSpace(false), EnableIRStructurizer(true),
+ EnablePromoteAlloca(false), EnableIfCvt(true), EnableLoadStoreOpt(false),
+ WavefrontSize(0), CFALUBug(false), LocalMemorySize(0),
+ EnableVGPRSpilling(false),
FrameLowering(TargetFrameLowering::StackGrowsUp,
64 * 16, // Maximum stack alignment (long16)
0),
- InstrItins(getInstrItineraryForCPU(GPU)) {
+ InstrItins(getInstrItineraryForCPU(GPU)), TargetTriple(TT) {
+
+ initializeSubtargetDependencies(TT, GPU, FS);
+
if (getGeneration() <= AMDGPUSubtarget::NORTHERN_ISLANDS) {
InstrInfo.reset(new R600InstrInfo(*this));
- TLInfo.reset(new R600TargetLowering(TM));
+ TLInfo.reset(new R600TargetLowering(TM, *this));
} else {
InstrInfo.reset(new SIInstrInfo(*this));
- TLInfo.reset(new SITargetLowering(TM));
+ TLInfo.reset(new SITargetLowering(TM, *this));
}
}
@@ -107,3 +100,33 @@ unsigned AMDGPUSubtarget::getStackEntrySize() const {
llvm_unreachable("Illegal wavefront size.");
}
}
+
+unsigned AMDGPUSubtarget::getAmdKernelCodeChipID() const {
+ switch(getGeneration()) {
+ default: llvm_unreachable("ChipID unknown");
+ case SEA_ISLANDS: return 12;
+ }
+}
+
+bool AMDGPUSubtarget::isVGPRSpillingEnabled(
+ const SIMachineFunctionInfo *MFI) const {
+ return MFI->getShaderType() == ShaderType::COMPUTE || EnableVGPRSpilling;
+}
+
+void AMDGPUSubtarget::overrideSchedPolicy(MachineSchedPolicy &Policy,
+ MachineInstr *begin,
+ MachineInstr *end,
+ unsigned NumRegionInstrs) const {
+ if (getGeneration() >= SOUTHERN_ISLANDS) {
+
+ // Track register pressure so the scheduler can try to decrease
+ // pressure once register usage is above the threshold defined by
+ // SIRegisterInfo::getRegPressureSetLimit()
+ Policy.ShouldTrackPressure = true;
+
+ // Enabling both top down and bottom up scheduling seems to give us less
+ // register spills than just using one of these approaches on its own.
+ Policy.OnlyTopDown = false;
+ Policy.OnlyBottomUp = false;
+ }
+}