Skip to content

[AMDGPU] Remove wavefrontsize feature from GFX10+ #98400

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 5 commits into from
Jul 16, 2024
Merged

[AMDGPU] Remove wavefrontsize feature from GFX10+ #98400

Show file tree
Hide file tree
Changes from 3 commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
ac6a483
[AMDGPU] Remove wavefrontsize feature from GFX10+
rampitec Jul 10, 2024
55dda8d
Just add FeatureWavefrontSize32 unconditionally
rampitec Jul 11, 2024
d09e31c
Move getIsaVersion to its original place
rampitec Jul 11, 2024
f5ebe7c
Readded braces.
rampitec Jul 11, 2024
51b32be
Updated comment
rampitec Jul 11, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 0 additions & 3 deletions llvm/lib/Target/AMDGPU/AMDGPU.td
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1464,7 +1464,6 @@ def FeatureISAVersion10_Common : FeatureSet<
FeatureLDSBankCount32,
FeatureDLInsts,
FeatureNSAEncoding,
FeatureWavefrontSize32,
FeatureBackOffBarrier]>;

def FeatureISAVersion10_1_Common : FeatureSet<
Expand Down Expand Up @@ -1548,7 +1547,6 @@ def FeatureISAVersion11_Common : FeatureSet<
FeatureDot10Insts,
FeatureNSAEncoding,
FeaturePartialNSAEncoding,
FeatureWavefrontSize32,
FeatureShaderCyclesRegister,
FeatureArchitectedFlatScratch,
FeatureAtomicFaddRtnInsts,
Expand Down Expand Up @@ -1625,7 +1623,6 @@ def FeatureISAVersion12 : FeatureSet<
FeatureDot11Insts,
FeatureNSAEncoding,
FeaturePartialNSAEncoding,
FeatureWavefrontSize32,
FeatureShaderCyclesHiLoRegisters,
FeatureArchitectedFlatScratch,
FeatureArchitectedSGPRs,
Expand Down
7 changes: 7 additions & 0 deletions llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -105,6 +105,13 @@ GCNSubtarget::initializeSubtargetDependencies(const Triple &TT,
: AMDGPUSubtarget::SOUTHERN_ISLANDS;
}

if (!hasFeature(AMDGPU::FeatureWavefrontSize32) &&
!hasFeature(AMDGPU::FeatureWavefrontSize64))
// If there is no default wave size it must be a generation before gfx9,
// these have FeatureWavefrontSize64 in their definition already. For gfx10+
// set wave32 as a default.
ToggleFeature(AMDGPU::FeatureWavefrontSize32);

// We don't support FP64 for EG/NI atm.
assert(!hasFP64() || (getGeneration() >= AMDGPUSubtarget::SOUTHERN_ISLANDS));

Expand Down
8 changes: 8 additions & 0 deletions llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1408,6 +1408,14 @@ class AMDGPUAsmParser : public MCTargetAsmParser {
copySTI().ToggleFeature("southern-islands");
}

FeatureBitset FB = getFeatureBits();
if (!FB[AMDGPU::FeatureWavefrontSize64] &&
!FB[AMDGPU::FeatureWavefrontSize32])
// If there is no default wave size it must be a generation before gfx9,
// these have FeatureWavefrontSize64 in their definition already. For
// gfx10+ set wave32 as a default.
copySTI().ToggleFeature(AMDGPU::FeatureWavefrontSize32);

setAvailableFeatures(ComputeAvailableFeatures(getFeatureBits()));

AMDGPU::IsaVersion ISA = AMDGPU::getIsaVersion(getSTI().getCPU());
Expand Down
20 changes: 18 additions & 2 deletions llvm/lib/Target/AMDGPU/Disassembler/AMDGPUDisassembler.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -45,10 +45,26 @@ using namespace llvm;

using DecodeStatus = llvm::MCDisassembler::DecodeStatus;

static const MCSubtargetInfo &addDefaultWaveSize(const MCSubtargetInfo &STI,
MCContext &Ctx) {
if (!STI.hasFeature(AMDGPU::FeatureWavefrontSize64) &&
!STI.hasFeature(AMDGPU::FeatureWavefrontSize32)) {
MCSubtargetInfo &STICopy = Ctx.getSubtargetCopy(STI);
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Don't really understand this subtargetcopy business

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

STI is const, I cannot just flip the bit. The same is done in the AsmParser with copySTI().

// If there is no default wave size it must be a generation before gfx9,
// these have FeatureWavefrontSize64 in their definition already. For gfx10+
// set wave32 as a default.
STICopy.ToggleFeature(AMDGPU::FeatureWavefrontSize32);
return STICopy;
}

return STI;
}

AMDGPUDisassembler::AMDGPUDisassembler(const MCSubtargetInfo &STI,
MCContext &Ctx, MCInstrInfo const *MCII)
: MCDisassembler(STI, Ctx), MCII(MCII), MRI(*Ctx.getRegisterInfo()),
MAI(*Ctx.getAsmInfo()), TargetMaxInstBytes(MAI.getMaxInstLength(&STI)),
: MCDisassembler(addDefaultWaveSize(STI, Ctx), Ctx), MCII(MCII),
MRI(*Ctx.getRegisterInfo()), MAI(*Ctx.getAsmInfo()),
TargetMaxInstBytes(MAI.getMaxInstLength(&STI)),
CodeObjectVersion(AMDGPU::getDefaultAMDHSACodeObjectVersion()) {
// ToDo: AMDGPUDisassembler supports only VI ISA.
if (!STI.hasFeature(AMDGPU::FeatureGCN3Encoding) && !isGFX10Plus())
Expand Down
2 changes: 0 additions & 2 deletions llvm/test/CodeGen/AMDGPU/check-subtarget-features.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,3 @@
; RUN: not llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,-wavefrontsize64 < %s 2>&1 | FileCheck %s -check-prefix=ERR -implicit-check-not=error:
; RUN: not llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,-wavefrontsize64 < %s 2>&1 | FileCheck %s -check-prefix=ERR -implicit-check-not=error:
Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This diagnostics is now missing, BE will just initialize wavesize to default. If we want to catch this case we would still need to scan the original string.

; RUN: not llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,+wavefrontsize64 < %s 2>&1 | FileCheck %s -check-prefix=ERR -implicit-check-not=error:
; RUN: not llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,+wavefrontsize64 < %s 2>&1 | FileCheck %s -check-prefix=ERR -implicit-check-not=error:

Expand Down
16 changes: 8 additions & 8 deletions llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wavefrontsize.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,19 +1,19 @@
; RUN: llc -mtriple=amdgcn -mcpu=fiji -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,W64 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1010 -mattr=+wavefrontsize32,-wavefrontsize64 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,W32 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1010 -mattr=-wavefrontsize32,+wavefrontsize64 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,W64 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 -verify-machineinstrs -amdgpu-enable-vopd=0 < %s | FileCheck -check-prefixes=GCN,W32 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,W64 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1010 -mattr=+wavefrontsize32 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,W32 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1010 -mattr=+wavefrontsize64 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,W64 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32 -verify-machineinstrs -amdgpu-enable-vopd=0 < %s | FileCheck -check-prefixes=GCN,W32 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize64 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,W64 %s

; RUN: opt -O3 -S < %s | FileCheck -check-prefix=OPT %s
; RUN: opt -mtriple=amdgcn-- -O3 -S < %s | FileCheck -check-prefix=OPT %s
; RUN: opt -mtriple=amdgcn-- -O3 -mattr=+wavefrontsize32 -S < %s | FileCheck -check-prefix=OPT %s
; RUN: opt -mtriple=amdgcn-- -passes='default<O3>' -mattr=+wavefrontsize32 -S < %s | FileCheck -check-prefix=OPT %s
; RUN: opt -mtriple=amdgcn-- -O3 -mattr=+wavefrontsize64 -S < %s | FileCheck -check-prefix=OPT %s
; RUN: opt -mtriple=amdgcn-- -mcpu=tonga -O3 -S < %s | FileCheck -check-prefix=OPT %s
; RUN: opt -mtriple=amdgcn-- -mcpu=gfx1010 -O3 -mattr=+wavefrontsize32,-wavefrontsize64 -S < %s | FileCheck -check-prefix=OPT %s
; RUN: opt -mtriple=amdgcn-- -mcpu=gfx1010 -O3 -mattr=-wavefrontsize32,+wavefrontsize64 -S < %s | FileCheck -check-prefix=OPT %s
; RUN: opt -mtriple=amdgcn-- -mcpu=gfx1100 -O3 -mattr=+wavefrontsize32,-wavefrontsize64 -S < %s | FileCheck -check-prefix=OPT %s
; RUN: opt -mtriple=amdgcn-- -mcpu=gfx1100 -O3 -mattr=-wavefrontsize32,+wavefrontsize64 -S < %s | FileCheck -check-prefix=OPT %s
; RUN: opt -mtriple=amdgcn-- -mcpu=gfx1010 -O3 -mattr=+wavefrontsize32 -S < %s | FileCheck -check-prefix=OPT %s
; RUN: opt -mtriple=amdgcn-- -mcpu=gfx1010 -O3 -mattr=+wavefrontsize64 -S < %s | FileCheck -check-prefix=OPT %s
; RUN: opt -mtriple=amdgcn-- -mcpu=gfx1100 -O3 -mattr=+wavefrontsize32 -S < %s | FileCheck -check-prefix=OPT %s
; RUN: opt -mtriple=amdgcn-- -mcpu=gfx1100 -O3 -mattr=+wavefrontsize64 -S < %s | FileCheck -check-prefix=OPT %s

; GCN-LABEL: {{^}}fold_wavefrontsize:
; OPT-LABEL: define amdgpu_kernel void @fold_wavefrontsize(
Expand Down
2 changes: 1 addition & 1 deletion llvm/test/CodeGen/AMDGPU/unknown-processor.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
; RUN: not llc -mtriple=amdgcn-- -mcpu=unknown -verify-machineinstrs < %s 2>&1 | FileCheck -check-prefix=ERROR -check-prefix=GCN %s
; RUN: llc -mtriple=amdgcn-- -mcpu=unknown -verify-machineinstrs < %s 2>&1 | FileCheck -check-prefix=ERROR -check-prefix=GCN %s
; RUN: llc -mtriple=r600-- -mcpu=unknown -verify-machineinstrs < %s 2>&1 | FileCheck -check-prefix=ERROR -check-prefix=R600 %s
target datalayout = "A5"

Expand Down
Loading
Loading