[AMDGPU] Remove wavefrontsize feature from GFX10+ (#98400)

Processor definition shall not include a default feature which may be switched off by a different wave size. This allows not to write -mattr=-wavefrontsize32,+wavefrontsize64 in tests.
llvm · Jul 16, 2024 · b132dd4 · b132dd4
1 parent 3a0e015
commit b132dd4
Show file tree

Hide file tree

Showing 10 changed files with 377 additions and 349 deletions.
diff --git a/llvm/lib/Target/AMDGPU/AMDGPU.td b/llvm/lib/Target/AMDGPU/AMDGPU.td
@@ -1464,7 +1464,6 @@ def FeatureISAVersion10_Common : FeatureSet<
    FeatureLDSBankCount32,
    FeatureDLInsts,
    FeatureNSAEncoding,
-   FeatureWavefrontSize32,
    FeatureBackOffBarrier]>;
 
 def FeatureISAVersion10_1_Common : FeatureSet<
@@ -1548,7 +1547,6 @@ def FeatureISAVersion11_Common : FeatureSet<
    FeatureDot10Insts,
    FeatureNSAEncoding,
    FeaturePartialNSAEncoding,
-   FeatureWavefrontSize32,
    FeatureShaderCyclesRegister,
    FeatureArchitectedFlatScratch,
    FeatureAtomicFaddRtnInsts,
@@ -1625,7 +1623,6 @@ def FeatureISAVersion12 : FeatureSet<
    FeatureDot11Insts,
    FeatureNSAEncoding,
    FeaturePartialNSAEncoding,
-   FeatureWavefrontSize32,
    FeatureShaderCyclesHiLoRegisters,
    FeatureArchitectedFlatScratch,
    FeatureArchitectedSGPRs,

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp b/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp
@@ -105,6 +105,14 @@ 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 gfx10,
+    // 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));
 

diff --git a/llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp b/llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp
@@ -1408,6 +1408,15 @@ 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 gfx10,
+      // 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());

diff --git a/llvm/lib/Target/AMDGPU/Disassembler/AMDGPUDisassembler.cpp b/llvm/lib/Target/AMDGPU/Disassembler/AMDGPUDisassembler.cpp
@@ -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);
+    // If there is no default wave size it must be a generation before gfx10,
+    // 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())

diff --git a/llvm/test/CodeGen/AMDGPU/check-subtarget-features.ll b/llvm/test/CodeGen/AMDGPU/check-subtarget-features.ll
@@ -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:
 ; 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:
 

diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wavefrontsize.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wavefrontsize.ll
@@ -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(

diff --git a/llvm/test/CodeGen/AMDGPU/unknown-processor.ll b/llvm/test/CodeGen/AMDGPU/unknown-processor.ll
@@ -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"