[DirectX backend] remove string function attribute and unused module flags

[python3kgae]

Summary

Remove string function attribute and unused module flags to match DXIL which is based on llvm 3.7.

Also move DXILPrepare pass after DXILTranslateMetadata pass because things like "hlsl.numthreads" will be removed by DXILPrepare pass now.

[beanz]

llvm/lib/Target/DirectX/DXILPrepare.cpp

110

I'm not sure I agree with the approach of having a list of keys to delete. That seems maintenance-intensive and error prone.

Why not instead only maintain a list of valid string attributes for DXIL, and strip everything else?

llvm/lib/Target/DirectX/DirectXTargetMachine.cpp

80

I think this makes sense, but I also think the DXILTranslateMetadataPass should be deleting the function attributes that it is translating.

Once we run the DXILTranslateMetadataPass any later passes that need information from that gets captured in the metadata should be reading it out of the preserved analysis. Having DXILTranslateMetadataPass delete the sources of where it read from will help us keep that working as expected.

[python3kgae]

Only keep debug related module flags which are used by DXIL.
Also let DXILPrepare pass preserve DXILResourceWrapper which is used in asm dump pass for test.

[python3kgae]

llvm/lib/Target/DirectX/DXILPrepare.cpp

110

Fixed.

llvm/lib/Target/DirectX/DirectXTargetMachine.cpp

80

That is a good idea.
Do you mind do that in a separate PR?

llvm/lib/Target/DirectX/DXILPrepare.cpp

 //===- DXILPrepare.cpp - Prepare LLVM Module for DXIL encoding ------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 ///
 /// \file This file contains pases and utilities to convert a modern LLVM
 /// module into a module compatible with the LLVM 3.7-based DirectX Intermediate
 /// Language (DXIL).
 //===----------------------------------------------------------------------===//
 
+#include "DXILResourceAnalysis.h"
 #include "DirectX.h"
 #include "DirectXIRPasses/PointerTypeAnalysis.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/SmallVector.h"
+#include "llvm/ADT/StringSet.h"
 #include "llvm/CodeGen/Passes.h"
 #include "llvm/IR/IRBuilder.h"
 #include "llvm/IR/Instruction.h"
 #include "llvm/IR/Module.h"
 #include "llvm/InitializePasses.h"
 #include "llvm/Pass.h"
 #include "llvm/Support/Compiler.h"
 
return is_contained({Attribute::Alignment,
                        Attribute::SanitizeAddress,
                        Attribute::SanitizeThread,
                        Attribute::SanitizeMemory,
                        Attribute::UWTable,
                        Attribute::ZExt},
                       Attr);
 }
 
+static void collectDeadStringAttrs(AttributeMask &DeadAttrs, AttributeSet &&AS,
+                                   StringSet<> LiveKeys) {
+  for (auto &Attr : AS) {
+    if (!Attr.isStringAttribute())
+      continue;
+    StringRef Key = Attr.getKindAsString();
+    if (LiveKeys.contains(Key))
+      continue;
+    DeadAttrs.addAttribute(Key);
+  }
+}
+
+static void removeStringFunctionAttributes(Function &F) {
+  AttributeList Attrs = F.getAttributes();
+  StringSet<> LiveKeys = {"waveops-include-helper-lanes"
+                          "fp32-denorm-mode"};
+  // Collect DeadKeys in FnAttrs.
+  AttributeMask DeadAttrs;
+  collectDeadStringAttrs(DeadAttrs, Attrs.getFnAttrs(), LiveKeys);
+  collectDeadStringAttrs(DeadAttrs, Attrs.getRetAttrs(), LiveKeys);
+
+  F.removeFnAttrs(DeadAttrs);
+  F.removeRetAttrs(DeadAttrs);
+}
+
+static void cleanModuleFlags(Module &M) {
+  NamedMDNode *MDFlags = M.getModuleFlagsMetadata();
+  if (!MDFlags)
+    return;
+
+  StringSet<> LiveKeys = {"Dwarf Version", "Debug Info Version"};
+
+  SmallVector<llvm::Module::ModuleFlagEntry> FlagEntries;
+  M.getModuleFlagsMetadata(FlagEntries);
+  MDFlags->eraseFromParent();
+
+  for (auto &Flag : FlagEntries) {
+    if (!LiveKeys.count(Flag.Key->getString()))
+      continue;
+    M.addModuleFlag(Flag.Behavior, Flag.Key->getString(), Flag.Val);
+  }
+}
+
 class DXILPrepareModule : public ModulePass {
 
   static Value *maybeGenerateBitcast(IRBuilder<> &Builder,
                                      PointerTypeMap &PointerTypes,
                                      Instruction &Inst, Value *Operand,
                                      Type *Ty) {
     // Omit bitcasts if the incoming value matches the instruction type.
     auto It = PointerTypes.find(Operand);
bool runOnModule(Module &M) override {
     for (Attribute::AttrKind I = Attribute::None; I != Attribute::EndAttrKinds;
          I = Attribute::AttrKind(I + 1)) {
       if (!isValidForDXIL(I))
         AttrMask.addAttribute(I);
     }
     for (auto &F : M.functions()) {
       F.removeFnAttrs(AttrMask);
       F.removeRetAttrs(AttrMask);
+      removeStringFunctionAttributes(F);
       for (size_t Idx = 0, End = F.arg_size(); Idx < End; ++Idx)
         F.removeParamAttrs(Idx, AttrMask);
 
       for (auto &BB : F) {
         IRBuilder<> Builder(&BB);
         for (auto &I : make_early_inc_range(BB)) {
           if (I.getOpcode() == Instruction::FNeg) {
             Builder.SetInsertPoint(&I);
for (auto &F : M.functions()) {
             CB->removeRetAttrs(AttrMask);
             for (size_t Idx = 0, End = CB->arg_size(); Idx < End; ++Idx)
               CB->removeParamAttrs(Idx, AttrMask);
             continue;
           }
         }
       }
     }
+    // Remove flags not in llvm3.7.
+    cleanModuleFlags(M);
     return true;
   }
 
   DXILPrepareModule() : ModulePass(ID) {}
 
+  void getAnalysisUsage(AnalysisUsage &AU) const override {
+    AU.addPreserved<DXILResourceWrapper>();
+  }
+
   static char ID; // Pass identification.
 };
 char DXILPrepareModule::ID = 0;
 
 } // end anonymous namespace
 
 INITIALIZE_PASS_BEGIN(DXILPrepareModule, DEBUG_TYPE, "DXIL Prepare Module",
                       false, false)
 INITIALIZE_PASS_END(DXILPrepareModule, DEBUG_TYPE, "DXIL Prepare Module", false,
                     false)
 
 ModulePass *llvm::createDXILPrepareModulePass() {
   return new DXILPrepareModule();
 }

llvm/lib/Target/DirectX/DirectXTargetMachine.cpp

public:
 
   DirectXTargetMachine &getDirectXTargetMachine() const {
     return getTM<DirectXTargetMachine>();
   }
 
   FunctionPass *createTargetRegisterAllocator(bool) override { return nullptr; }
   void addCodeGenPrepare() override {
     addPass(createDXILOpLoweringLegacyPass());
-    addPass(createDXILPrepareModulePass());
     addPass(createDXILTranslateMetadataPass());
+    addPass(createDXILPrepareModulePass());
   }
 };
 
 DirectXTargetMachine::DirectXTargetMachine(const Target &T, const Triple &TT,
                                            StringRef CPU, StringRef FS,
                                            const TargetOptions &Options,
                                            std::optional<Reloc::Model> RM,
                                            std::optional<CodeModel::Model> CM,

llvm/test/CodeGen/DirectX/UAVMetadata.ll

 ; RUN: opt -S -dxil-metadata-emit < %s | FileCheck %s
 ; RUN: opt -S --passes="print-dxil-resource" < %s 2>&1 | FileCheck %s --check-prefix=PRINT
-; RUN: llc %s --filetype=asm -o - < %s 2>&1 | FileCheck %s --check-prefixes=CHECK,PRINT
+; RUN: llc --filetype=asm -o - < %s 2>&1 | FileCheck %s --check-prefixes=CHECK,PRINT
 
 target datalayout = "e-m:e-p:32:32-i1:32-i8:8-i16:16-i32:32-i64:64-f16:16-f32:32-f64:64-n8:16:32:64"
 target triple = "dxil-pc-shadermodel6.0-library"
 
 %"class.hlsl::RWBuffer" = type { ptr }
 
 
 ; PRINT:; Resource Bindings:

llvm/test/CodeGen/DirectX/strip-string-fn-attrs-module-flags.ll

+; RUN: opt -S -dxil-prepare < %s | FileCheck %s
+target datalayout = "e-m:e-p:32:32-i1:32-i8:8-i16:16-i32:32-i64:64-f16:16-f32:32-f64:64-n8:16:32:64"
+target triple = "dxil-unknown-shadermodel6.0-compute"
+
+; make sure string function attribute and module flags are clear for llvm3.7.
+
+; CHECK-LABEL:define void @CSMain()
+; CHECK:attributes #0 = { nounwind memory(readwrite, inaccessiblemem: read) }
+; CHECK-NOT:!llvm.module.flags
+
+
+%"class.hlsl::RWBuffer" = type { ptr }
+%dx.types.Handle = type { ptr }
+%dx.types.ResRet.i32 = type { i32, i32, i32, i32, i32 }
+
+@"?srcBuffer@@3V?$RWBuffer@T?$__vector@H$01@__clang@@@hlsl@@A" = local_unnamed_addr global %"class.hlsl::RWBuffer" zeroinitializer, align 4
+@"?dstBuffer@@3V?$RWBuffer@T?$__vector@H$01@__clang@@@hlsl@@A" = local_unnamed_addr global %"class.hlsl::RWBuffer" zeroinitializer, align 4
+
+; Function Attrs: mustprogress nofree nounwind willreturn memory(readwrite, inaccessiblemem: read)
+define void @CSMain() local_unnamed_addr #0 {
+entry:
+  %0 = call %dx.types.Handle @dx.op.createHandle(i32 57, i8 1, i32 1, i32 0, i1 false)
+  %1 = call %dx.types.Handle @dx.op.createHandle(i32 57, i8 1, i32 0, i32 1, i1 false)
+  %2 = call i32 @dx.op.threadId.i32(i32 93, i32 0)
+  %3 = call %dx.types.ResRet.i32 @dx.op.bufferLoad.i32(i32 68, %dx.types.Handle %1, i32 %2, i32 poison)
+  %4 = extractvalue %dx.types.ResRet.i32 %3, 0
+  %5 = extractvalue %dx.types.ResRet.i32 %3, 1
+  %add.i.i0 = add i32 %4, 10
+  %add.i.i1 = add i32 %5, 10
+  call void @dx.op.bufferStore.i32(i32 69, %dx.types.Handle %0, i32 %2, i32 poison, i32 %add.i.i0, i32 %add.i.i1, i32 %add.i.i0, i32 %add.i.i0, i8 15)
+  ret void
+}
+
+declare %dx.types.Handle @dx.op.createHandle(i32 %0, i8 %1, i32 %2, i32 %3, i1 %4)
+
+declare %dx.types.ResRet.i32 @dx.op.bufferLoad.i32(i32 %0, %dx.types.Handle %1, i32 %2, i32 %3)
+
+declare void @dx.op.bufferStore.i32(i32 %0, %dx.types.Handle %1, i32 %2, i32 %3, i32 %4, i32 %5, i32 %6, i32 %7, i8 %8)
+
+declare i32 @dx.op.threadId.i32(i32 %0, i32 %1)
+
+attributes #0 = { mustprogress nofree nounwind willreturn memory(readwrite, inaccessiblemem: read) "frame-pointer"="all" "hlsl.numthreads"="1024,1,1" "hlsl.shader"="compute" "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
+
+!llvm.module.flags = !{!0, !1}
+
+!0 = !{i32 1, !"wchar_size", i32 4}
+!1 = !{i32 7, !"frame-pointer", i32 2}

llvm/test/tools/dxil-dis/attribute-filter.ll

 ; CHECK: Function Attrs: readnone
 ; Function Attrs: norecurse readnone willreturn
 define float @fma2(float %0, float %1, float %2) #1 {
   %4 = fmul float %0, %1
   %5 = fadd float %4, %2
   ret float %5
 }
 
-; CHECK: attributes #0 = { nounwind readnone "disable-tail-calls"="false" }
+; CHECK: attributes #0 = { nounwind readnone }
 attributes #0 = { norecurse nounwind readnone willreturn "disable-tail-calls"="false" }
 
-; CHECK: attributes #1 = { readnone "disable-tail-calls"="false" }
+; CHECK: attributes #1 = { readnone }
 attributes #1 = { norecurse memory(none) willreturn "disable-tail-calls"="false" }

llvm/test/tools/dxil-dis/di-compile-unit.ll

 ; RUN: llc --filetype=obj %s -o - | dxil-dis -o - | FileCheck %s
 target triple = "dxil-unknown-shadermodel6.7-library"
 
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!3, !4}
 
 !0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "Some Compiler", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, splitDebugInlining: false, nameTableKind: None)
 !1 = !DIFile(filename: "di-compile-unit.src", directory: "/some-path")
 !2 = !{}
 !3 = !{i32 7, !"Dwarf Version", i32 2}
 !4 = !{i32 2, !"Debug Info Version", i32 3}
 
+; CHECK: !llvm.module.flags = !{![[DWARF:[0-9]+]], ![[DEBUG:[0-9]+]]}
+
 ; CHECK: !0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "Some Compiler", isOptimized: true, runtimeVersion: 0, emissionKind: 1, enums: !2)
 ; CHECK: !1 = !DIFile(filename: "di-compile-unit.src", directory: "/some-path")
-; CHECK: !2 = !{}
-; CHECK: !3 = !{i32 7, !"Dwarf Version", i32 2}
-; CHECK: !4 = !{i32 2, !"Debug Info Version", i32 3}
+; CHECK: ![[DWARF]] = !{i32 7, !"Dwarf Version", i32 2}
+; CHECK: ![[DEBUG]] = !{i32 2, !"Debug Info Version", i32 3}

llvm/test/tools/dxil-dis/di-subprogram.ll

 ; RUN: llc --filetype=obj %s -o - | dxil-dis -o - | FileCheck %s
 target triple = "dxil-unknown-shadermodel6.7-library"
 
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!3, !4}
 !llvm.used = !{!5}
 !llvm.lines = !{!13, !14, !15, !16}
 
+; CHECK: !llvm.module.flags = !{![[DWARF:[0-9]+]], ![[DEBUG:[0-9]+]]}
+
 ; CHECK: !0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "Some Compiler", isOptimized: true, runtimeVersion: 0, emissionKind: 1, enums: !2)
 !0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "Some Compiler", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, splitDebugInlining: false, nameTableKind: None)
 ; CHECK: !1 = !DIFile(filename: "some-source", directory: "some-path")
 !1 = !DIFile(filename: "some-source", directory: "some-path")
 !2 = !{}
 
-; CHECK: !3 = !{i32 7, !"Dwarf Version", i32 2}
 !3 = !{i32 7, !"Dwarf Version", i32 2}
-; CHECK: !4 = !{i32 2, !"Debug Info Version", i32 3}
 !4 = !{i32 2, !"Debug Info Version", i32 3}
 
-; CHECK: !5 = distinct !DISubprogram(name: "fma", scope: !1, file: !1, line: 1, type: !6, isLocal: false, isDefinition: true, scopeLine: 1, flags: DIFlagPrototyped, isOptimized: true, function: !0, variables: !9)
+; CHECK: ![[FMA:[0-9]+]] = distinct !DISubprogram(name: "fma", scope: !1, file: !1, line: 1, type: ![[FMATY:[0-9]+]], isLocal: false, isDefinition: true, scopeLine: 1, flags: DIFlagPrototyped, isOptimized: true, function: !0, variables: ![[VARS:[0-9]+]])
 !5 = distinct !DISubprogram(name: "fma", scope: !1, file: !1, line: 1, type: !6, scopeLine: 1, flags: DIFlagPrototyped, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !9)
 
-; CHECK: !6 = !DISubroutineType(types: !7)
+; CHECK: ![[FMATY]] = !DISubroutineType(types: ![[FMATYS:[0-9]+]])
 !6 = !DISubroutineType(types: !7)
 
-; CHECK: !7 = !{!8, !8, !8, !8}
+; CHECK: ![[FMATYS]] = !{![[F32TY:[0-9]+]], ![[F32TY]], ![[F32TY]], ![[F32TY]]}
 !7 = !{!8, !8, !8, !8}
 
-; CHECK: !8 = !DIBasicType(name: "float", size: 32, encoding: DW_ATE_float)
+; CHECK: ![[F32TY]] = !DIBasicType(name: "float", size: 32, encoding: DW_ATE_float)
 !8 = !DIBasicType(name: "float", size: 32, encoding: DW_ATE_float)
 
-; CHECK: !9 = !{!10, !11, !12}
+; CHECK: ![[VARS]] = !{![[X:[0-9]+]], ![[Y:[0-9]+]], ![[Z:[0-9]+]]}
 !9 = !{!10, !11, !12}
 
-; CHECK: !10 = !DILocalVariable(tag: DW_TAG_variable, name: "x", arg: 1, scope: !5, file: !1, line: 1, type: !8)
+; CHECK: ![[X]] = !DILocalVariable(tag: DW_TAG_variable, name: "x", arg: 1, scope: ![[FMA]], file: !1, line: 1, type: ![[F32TY]])
 !10 = !DILocalVariable(name: "x", arg: 1, scope: !5, file: !1, line: 1, type: !8)
 
-; CHECK: !11 = !DILocalVariable(tag: DW_TAG_variable, name: "y", arg: 2, scope: !5, file: !1, line: 1, type: !8)
+; CHECK: ![[Y]] = !DILocalVariable(tag: DW_TAG_variable, name: "y", arg: 2, scope: ![[FMA]], file: !1, line: 1, type: ![[F32TY]])
 !11 = !DILocalVariable(name: "y", arg: 2, scope: !5, file: !1, line: 1, type: !8)
 
-; CHECK: !12 = !DILocalVariable(tag: DW_TAG_variable, name: "z", arg: 3, scope: !5, file: !1, line: 1, type: !8)
+; CHECK: ![[Z]] = !DILocalVariable(tag: DW_TAG_variable, name: "z", arg: 3, scope: ![[FMA]], file: !1, line: 1, type: ![[F32TY]])
 !12 = !DILocalVariable(name: "z", arg: 3, scope: !5, file: !1, line: 1, type: !8)
 
 
-; CHECK: !13 = !DILocation(line: 0, scope: !5)
-; CHECK: !14 = !DILocation(line: 2, column: 12, scope: !5)
-; CHECK: !15 = !DILocation(line: 2, column: 16, scope: !5)
-; CHECK: !16 = !DILocation(line: 2, column: 3, scope: !5)
+; CHECK: !{{[0-9]+}} = !DILocation(line: 0, scope: ![[FMA]])
+; CHECK: !{{[0-9]+}} = !DILocation(line: 2, column: 12, scope: ![[FMA]])
+; CHECK: !{{[0-9]+}} = !DILocation(line: 2, column: 16, scope: ![[FMA]])
+; CHECK: !{{[0-9]+}} = !DILocation(line: 2, column: 3, scope: ![[FMA]])
 
 !13 = !DILocation(line: 0, scope: !5)
 !14 = !DILocation(line: 2, column: 12, scope: !5)
 !15 = !DILocation(line: 2, column: 16, scope: !5)
 !16 = !DILocation(line: 2, column: 3, scope: !5)
+
+; CHECK: ![[DWARF]] = !{i32 7, !"Dwarf Version", i32 2}
+; CHECK: ![[DEBUG]] = !{i32 2, !"Debug Info Version", i32 3}