RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2025-02-01 05:01:59 +01:00
Code Issues Projects Releases Wiki Activity

AMDGPU: Change internal tracking of wave size

Browse Source 
Store the log2 wave size instead of forcing division and log2
operations when querying either.
This commit is contained in:
Matt Arsenault 2020-06-01 13:16:58 -04:00
parent 76c68e0c0b
commit 3cd292c66e
3 changed files with 19 additions and 19 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

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

@ -29,16 +29,16 @@ def FeatureLocalMemorySize0 : SubtargetFeatureLocalMemorySize<0>;
def FeatureLocalMemorySize32768 : SubtargetFeatureLocalMemorySize<32768>;
def FeatureLocalMemorySize65536 : SubtargetFeatureLocalMemorySize<65536>;
class SubtargetFeatureWavefrontSize <int Value> : SubtargetFeature<
"wavefrontsize"#Value,
"WavefrontSize",
!cast<string>(Value),
class SubtargetFeatureWavefrontSize <int ValueLog2> : SubtargetFeature<
"wavefrontsize"#!shl(1, ValueLog2),
"WavefrontSizeLog2",
!cast<string>(ValueLog2),
"The number of threads per wavefront"
>;
def FeatureWavefrontSize16 : SubtargetFeatureWavefrontSize<16>;
def FeatureWavefrontSize32 : SubtargetFeatureWavefrontSize<32>;
def FeatureWavefrontSize64 : SubtargetFeatureWavefrontSize<64>;
def FeatureWavefrontSize16 : SubtargetFeatureWavefrontSize<4>;
def FeatureWavefrontSize32 : SubtargetFeatureWavefrontSize<5>;
def FeatureWavefrontSize64 : SubtargetFeatureWavefrontSize<6>;
class SubtargetFeatureGeneration <string Value, string FeatureName,
string Subtarget,

6
lib/Target/AMDGPU/AMDGPUSubtarget.cpp
View File

@ -126,8 +126,8 @@ GCNSubtarget::initializeSubtargetDependencies(const Triple &TT,
}
// Don't crash on invalid devices.
if (WavefrontSize == 0)
WavefrontSize = 64;
if (WavefrontSizeLog2 == 0)
WavefrontSizeLog2 = 5;
HasFminFmaxLegacy = getGeneration() < AMDGPUSubtarget::VOLCANIC_ISLANDS;
@ -163,7 +163,7 @@ AMDGPUSubtarget::AMDGPUSubtarget(const Triple &TT) :
HasTrigReducedRange(false),
MaxWavesPerEU(10),
LocalMemorySize(0),
WavefrontSize(0)
WavefrontSizeLog2(0)
{ }
GCNSubtarget::GCNSubtarget(const Triple &TT, StringRef GPU, StringRef FS,

16
lib/Target/AMDGPU/AMDGPUSubtarget.h
View File

@ -77,7 +77,7 @@ protected:
bool HasTrigReducedRange;
unsigned MaxWavesPerEU;
int LocalMemorySize;
unsigned WavefrontSize;
char WavefrontSizeLog2;
public:
AMDGPUSubtarget(const Triple &TT);
@ -181,7 +181,11 @@ public:
}
unsigned getWavefrontSize() const {
return WavefrontSize;
return 1 << WavefrontSizeLog2;
}
unsigned getWavefrontSizeLog2() const {
return WavefrontSizeLog2;
}
int getLocalMemorySize() const {
@ -237,7 +241,7 @@ public:
/// \returns Corresponsing DWARF register number mapping flavour for the
/// \p WavefrontSize.
AMDGPUDwarfFlavour getAMDGPUDwarfFlavour() const {
return WavefrontSize == 32 ? AMDGPUDwarfFlavour::Wave32
return getWavefrontSize() == 32 ? AMDGPUDwarfFlavour::Wave32
: AMDGPUDwarfFlavour::Wave64;
}
@ -454,10 +458,6 @@ public:
return (Generation)Gen;
}
unsigned getWavefrontSizeLog2() const {
return Log2_32(WavefrontSize);
}
/// Return the number of high bits known to be zero fror a frame index.
unsigned getKnownHighZeroBitsForFrameIndex() const {
return countLeadingZeros(MaxWaveScratchSize) + getWavefrontSizeLog2();
@ -1160,7 +1160,7 @@ public:
const override;
bool isWave32() const {
return WavefrontSize == 32;
return getWavefrontSize() == 32;
}
const TargetRegisterClass *getBoolRC() const {