RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2025-01-31 12:41:49 +01:00
Code Issues Projects Releases Wiki Activity

AMDGPU: Add feature for fast f32 denormals

Browse Source
This commit is contained in:
Matt Arsenault 2018-08-15 22:45:04 +03:00 committed by Matt Arsenault
parent 0e9a095779
commit f2d158d7a0
3 changed files with 12 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
lib/Target/AMDGPU/AMDGPU.td
View File

@ -33,6 +33,12 @@ def FeatureFastFMAF32 : SubtargetFeature<"fast-fmaf",
"Assuming f32 fma is at least as fast as mul + add"
>;
def FeatureFastDenormalF32 : SubtargetFeature<"fast-denormal-f32",
"FastDenormalF32",
"true",
"Enabling denormals does not cause f32 instructions to run at f64 rates"
>;
def FeatureMIMG_R128 : SubtargetFeature<"mimg-r128",
"MIMG_R128",
"true",
@ -632,7 +638,7 @@ def FeatureVolcanicIslands : GCNSubtargetFeatureGeneration<"VOLCANIC_ISLANDS",
FeatureScalarStores, FeatureInv2PiInlineImm,
FeatureSDWA, FeatureSDWAOutModsVOPC, FeatureSDWAMac, FeatureDPP,
FeatureIntClamp, FeatureTrigReducedRange, FeatureDoesNotSupportSRAMECC,
FeatureGFX8Insts, FeatureGFX7GFX8GFX9Insts
FeatureGFX8Insts, FeatureGFX7GFX8GFX9Insts, FeatureFastDenormalF32
]
>;
@ -647,8 +653,8 @@ def FeatureGFX9 : GCNSubtargetFeatureGeneration<"GFX9",
FeatureSDWA, FeatureSDWAOmod, FeatureSDWAScalar, FeatureSDWASdst,
FeatureFlatInstOffsets, FeatureFlatGlobalInsts, FeatureFlatScratchInsts,
FeatureAddNoCarryInsts, FeatureGFX8Insts, FeatureGFX7GFX8GFX9Insts,
FeatureScalarFlatScratchInsts, FeatureScalarAtomics, FeatureR128A16
]
FeatureScalarFlatScratchInsts, FeatureScalarAtomics, FeatureR128A16,
FeatureFastDenormalF32]
>;
def FeatureGFX10 : GCNSubtargetFeatureGeneration<"GFX10",
@ -665,7 +671,7 @@ def FeatureGFX10 : GCNSubtargetFeatureGeneration<"GFX10",
FeatureNoSdstCMPX, FeatureVscnt, FeatureRegisterBanking,
FeatureVOP3Literal, FeatureDPP8,
FeatureNoDataDepHazard, FeaturePkFmacF16Inst, FeatureDoesNotSupportSRAMECC,
FeatureGFX10A16
FeatureGFX10A16, FeatureFastDenormalF32
]
>;

1
lib/Target/AMDGPU/AMDGPUSubtarget.cpp
View File

@ -178,6 +178,7 @@ GCNSubtarget::GCNSubtarget(const Triple &TT, StringRef GPU, StringRef FS,
MaxPrivateElementSize(0),
FastFMAF32(false),
FastDenormalF32(false),
HalfRate64Ops(false),
FlatForGlobal(false),

1
lib/Target/AMDGPU/AMDGPUSubtarget.h
View File

@ -291,6 +291,7 @@ protected:
// Possibly statically set by tablegen, but may want to be overridden.
bool FastFMAF32;
bool FastDenormalF32;
bool HalfRate64Ops;
// Dynamially set bits that enable features.