[clang][HLSL][SPRI-V] Add convergence intrinsics (#80680)

Keenuts · web-flow · commit 0f61051f541a · 2024-03-28T17:18:05.000+01:00
HLSL has wave operations and other kind of function which required the
control flow to either be converged, or respect certain constraints as
where and how to re-converge.

At the HLSL level, the convergence are mostly obvious: the control flow
is expected to re-converge at the end of a scope.
Once translated to IR, HLSL scopes disapear. This means we need a way to
communicate convergence restrictions down to the backend.

For this, the SPIR-V backend uses convergence intrinsics. So this commit
adds some code to generate convergence intrinsics when required.

---------

Signed-off-by: Nathan Gauër &lt;brioche@google.com&gt;
diff --git a/clang/include/clang/Basic/Builtins.td b/clang/include/clang/Basic/Builtins.td
@@ -4599,6 +4599,12 @@ def HLSLWaveActiveCountBits : LangBuiltin<"HLSL_LANG"> {
   let Prototype = "unsigned int(bool)";
 }
 
+def HLSLWaveGetLaneIndex : LangBuiltin<"HLSL_LANG"> {
+  let Spellings = ["__builtin_hlsl_wave_get_lane_index"];
+  let Attributes = [NoThrow, Const];
+  let Prototype = "unsigned int()";
+}
+
 def HLSLClamp : LangBuiltin<"HLSL_LANG"> {
   let Spellings = ["__builtin_hlsl_elementwise_clamp"];
   let Attributes = [NoThrow, Const];
diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -1131,8 +1131,92 @@ struct BitTest {
 
   static BitTest decodeBitTestBuiltin(unsigned BuiltinID);
 };
+
+// Returns the first convergence entry/loop/anchor instruction found in |BB|.
+// std::nullptr otherwise.
+llvm::IntrinsicInst *getConvergenceToken(llvm::BasicBlock *BB) {
+  for (auto &I : *BB) {
+    auto *II = dyn_cast<llvm::IntrinsicInst>(&I);
+    if (II && isConvergenceControlIntrinsic(II->getIntrinsicID()))
+      return II;
+  }
+  return nullptr;
+}
+
 } // namespace
 
+llvm::CallBase *
+CodeGenFunction::addConvergenceControlToken(llvm::CallBase *Input,
+                                            llvm::Value *ParentToken) {
+  llvm::Value *bundleArgs[] = {ParentToken};
+  llvm::OperandBundleDef OB("convergencectrl", bundleArgs);
+  auto Output = llvm::CallBase::addOperandBundle(
+      Input, llvm::LLVMContext::OB_convergencectrl, OB, Input);
+  Input->replaceAllUsesWith(Output);
+  Input->eraseFromParent();
+  return Output;
+}
+
+llvm::IntrinsicInst *
+CodeGenFunction::emitConvergenceLoopToken(llvm::BasicBlock *BB,
+                                          llvm::Value *ParentToken) {
+  CGBuilderTy::InsertPoint IP = Builder.saveIP();
+  Builder.SetInsertPoint(&BB->front());
+  auto CB = Builder.CreateIntrinsic(
+      llvm::Intrinsic::experimental_convergence_loop, {}, {});
+  Builder.restoreIP(IP);
+
+  auto I = addConvergenceControlToken(CB, ParentToken);
+  return cast<llvm::IntrinsicInst>(I);
+}
+
+llvm::IntrinsicInst *
+CodeGenFunction::getOrEmitConvergenceEntryToken(llvm::Function *F) {
+  auto *BB = &F->getEntryBlock();
+  auto *token = getConvergenceToken(BB);
+  if (token)
+    return token;
+
+  // Adding a convergence token requires the function to be marked as
+  // convergent.
+  F->setConvergent();
+
+  CGBuilderTy::InsertPoint IP = Builder.saveIP();
+  Builder.SetInsertPoint(&BB->front());
+  auto I = Builder.CreateIntrinsic(
+      llvm::Intrinsic::experimental_convergence_entry, {}, {});
+  assert(isa<llvm::IntrinsicInst>(I));
+  Builder.restoreIP(IP);
+
+  return cast<llvm::IntrinsicInst>(I);
+}
+
+llvm::IntrinsicInst *
+CodeGenFunction::getOrEmitConvergenceLoopToken(const LoopInfo *LI) {
+  assert(LI != nullptr);
+
+  auto *token = getConvergenceToken(LI->getHeader());
+  if (token)
+    return token;
+
+  llvm::IntrinsicInst *PII =
+      LI->getParent()
+          ? emitConvergenceLoopToken(
+                LI->getHeader(), getOrEmitConvergenceLoopToken(LI->getParent()))
+          : getOrEmitConvergenceEntryToken(LI->getHeader()->getParent());
+
+  return emitConvergenceLoopToken(LI->getHeader(), PII);
+}
+
+llvm::CallBase *
+CodeGenFunction::addControlledConvergenceToken(llvm::CallBase *Input) {
+  llvm::Value *ParentToken =
+      LoopStack.hasInfo()
+          ? getOrEmitConvergenceLoopToken(&LoopStack.getInfo())
+          : getOrEmitConvergenceEntryToken(Input->getFunction());
+  return addConvergenceControlToken(Input, ParentToken);
+}
+
 BitTest BitTest::decodeBitTestBuiltin(unsigned BuiltinID) {
   switch (BuiltinID) {
     // Main portable variants.
@@ -5809,6 +5893,15 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
         {NDRange, Kernel, Block}));
   }
 
+  case Builtin::BI__builtin_hlsl_wave_get_lane_index: {
+    auto *CI = EmitRuntimeCall(CGM.CreateRuntimeFunction(
+        llvm::FunctionType::get(IntTy, {}, false), "__hlsl_wave_get_lane_index",
+        {}, false, true));
+    if (getTarget().getTriple().isSPIRVLogical())
+      CI = dyn_cast<CallInst>(addControlledConvergenceToken(CI));
+    return RValue::get(CI);
+  }
+
   case Builtin::BI__builtin_store_half:
   case Builtin::BI__builtin_store_halff: {
     Value *Val = EmitScalarExpr(E->getArg(0));
diff --git a/clang/lib/CodeGen/CGCall.cpp b/clang/lib/CodeGen/CGCall.cpp
@@ -5715,6 +5715,9 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
   if (!CI->getType()->isVoidTy())
     CI->setName("call");
 
+  if (getTarget().getTriple().isSPIRVLogical() && CI->isConvergent())
+    CI = addControlledConvergenceToken(CI);
+
   // Update largest vector width from the return type.
   LargestVectorWidth =
       std::max(LargestVectorWidth, getMaxVectorWidth(CI->getType()));
diff --git a/clang/lib/CodeGen/CGLoopInfo.h b/clang/lib/CodeGen/CGLoopInfo.h
@@ -110,6 +110,10 @@ class LoopInfo {
   /// been processed.
   void finish();
 
+  /// Returns the first outer loop containing this loop if any, nullptr
+  /// otherwise.
+  const LoopInfo *getParent() const { return Parent; }
+
 private:
   /// Loop ID metadata.
   llvm::TempMDTuple TempLoopID;
@@ -291,12 +295,13 @@ class LoopInfoStack {
   /// Set no progress for the next loop pushed.
   void setMustProgress(bool P) { StagedAttrs.MustProgress = P; }
 
-private:
   /// Returns true if there is LoopInfo on the stack.
   bool hasInfo() const { return !Active.empty(); }
   /// Return the LoopInfo for the current loop. HasInfo should be called
   /// first to ensure LoopInfo is present.
   const LoopInfo &getInfo() const { return *Active.back(); }
+
+private:
   /// The set of attributes that will be applied to the next pushed loop.
   LoopAttributes StagedAttrs;
   /// Stack of active loops.
diff --git a/clang/lib/CodeGen/CodeGenFunction.h b/clang/lib/CodeGen/CodeGenFunction.h
@@ -4985,6 +4985,25 @@ class CodeGenFunction : public CodeGenTypeCache {
   llvm::Value *emitBoolVecConversion(llvm::Value *SrcVec,
                                      unsigned NumElementsDst,
                                      const llvm::Twine &Name = "");
+  // Adds a convergence_ctrl token to |Input| and emits the required parent
+  // convergence instructions.
+  llvm::CallBase *addControlledConvergenceToken(llvm::CallBase *Input);
+
+private:
+  // Emits a convergence_loop instruction for the given |BB|, with |ParentToken|
+  // as it's parent convergence instr.
+  llvm::IntrinsicInst *emitConvergenceLoopToken(llvm::BasicBlock *BB,
+                                                llvm::Value *ParentToken);
+  // Adds a convergence_ctrl token with |ParentToken| as parent convergence
+  // instr to the call |Input|.
+  llvm::CallBase *addConvergenceControlToken(llvm::CallBase *Input,
+                                             llvm::Value *ParentToken);
+  // Find the convergence_entry instruction |F|, or emits ones if none exists.
+  // Returns the convergence instruction.
+  llvm::IntrinsicInst *getOrEmitConvergenceEntryToken(llvm::Function *F);
+  // Find the convergence_loop instruction for the loop defined by |LI|, or
+  // emits one if none exists. Returns the convergence instruction.
+  llvm::IntrinsicInst *getOrEmitConvergenceLoopToken(const LoopInfo *LI);
 
 private:
   llvm::MDNode *getRangeForLoadFromType(QualType Ty);
diff --git a/clang/lib/Headers/hlsl/hlsl_intrinsics.h b/clang/lib/Headers/hlsl/hlsl_intrinsics.h
@@ -1389,7 +1389,12 @@ float4 trunc(float4);
 /// true, across all active lanes in the current wave.
 _HLSL_AVAILABILITY(shadermodel, 6.0)
 _HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_count_bits)
-uint WaveActiveCountBits(bool Val);
+__attribute__((convergent)) uint WaveActiveCountBits(bool Val);
+
+/// \brief Returns the index of the current lane within the current wave.
+_HLSL_AVAILABILITY(shadermodel, 6.0)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_get_lane_index)
+__attribute__((convergent)) uint WaveGetLaneIndex();
 
 } // namespace hlsl
 #endif //_HLSL_HLSL_INTRINSICS_H_
diff --git a/clang/test/CodeGenHLSL/builtins/wave_get_lane_index_do_while.hlsl b/clang/test/CodeGenHLSL/builtins/wave_get_lane_index_do_while.hlsl
@@ -0,0 +1,40 @@
+// RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -x hlsl -triple \
+// RUN:   spirv-pc-vulkan-library %s -emit-llvm -disable-llvm-passes -o - | FileCheck %s
+
+// CHECK: define spir_func void @main() [[A0:#[0-9]+]] {
+void main() {
+// CHECK: entry:
+// CHECK:   %[[CT_ENTRY:[0-9]+]] = call token @llvm.experimental.convergence.entry()
+// CHECK:   br label %[[LABEL_WHILE_COND:.+]]
+  int cond = 0;
+
+// CHECK: [[LABEL_WHILE_COND]]:
+// CHECK:   %[[CT_LOOP:[0-9]+]] = call token @llvm.experimental.convergence.loop() [ "convergencectrl"(token %[[CT_ENTRY]]) ]
+// CHECK:   br label %[[LABEL_WHILE_BODY:.+]]
+  while (true) {
+
+// CHECK: [[LABEL_WHILE_BODY]]:
+// CHECK:   br i1 {{%.+}}, label %[[LABEL_IF_THEN:.+]], label %[[LABEL_IF_END:.+]]
+
+// CHECK: [[LABEL_IF_THEN]]:
+// CHECK:   call i32 @__hlsl_wave_get_lane_index() [ "convergencectrl"(token %[[CT_LOOP]]) ]
+// CHECK:   br label %[[LABEL_WHILE_END:.+]]
+    if (cond == 2) {
+      uint index = WaveGetLaneIndex();
+      break;
+    }
+
+// CHECK: [[LABEL_IF_END]]:
+// CHECK:   br label %[[LABEL_WHILE_COND]]
+    cond++;
+  }
+
+// CHECK: [[LABEL_WHILE_END]]:
+// CHECK:   ret void
+}
+
+// CHECK-DAG: declare i32 @__hlsl_wave_get_lane_index() [[A1:#[0-9]+]]
+
+// CHECK-DAG: attributes [[A0]] = {{{.*}}convergent{{.*}}}
+// CHECK-DAG: attributes [[A1]] = {{{.*}}convergent{{.*}}}
+
diff --git a/clang/test/CodeGenHLSL/builtins/wave_get_lane_index_simple.hlsl b/clang/test/CodeGenHLSL/builtins/wave_get_lane_index_simple.hlsl
@@ -0,0 +1,14 @@
+// RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -x hlsl -triple \
+// RUN:   spirv-pc-vulkan-library %s -emit-llvm -disable-llvm-passes -o - | FileCheck %s
+
+// CHECK: define spir_func noundef i32 @_Z6test_1v() [[A0:#[0-9]+]] {
+// CHECK: %[[CI:[0-9]+]] = call token @llvm.experimental.convergence.entry()
+// CHECK: call i32 @__hlsl_wave_get_lane_index() [ "convergencectrl"(token %[[CI]]) ]
+uint test_1() {
+  return WaveGetLaneIndex();
+}
+
+// CHECK: declare i32 @__hlsl_wave_get_lane_index() [[A1:#[0-9]+]]
+
+// CHECK-DAG: attributes [[A0]] = { {{.*}}convergent{{.*}} }
+// CHECK-DAG: attributes [[A1]] = { {{.*}}convergent{{.*}} }
diff --git a/clang/test/CodeGenHLSL/builtins/wave_get_lane_index_subcall.hlsl b/clang/test/CodeGenHLSL/builtins/wave_get_lane_index_subcall.hlsl
@@ -0,0 +1,21 @@
+// RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -x hlsl -triple \
+// RUN:   spirv-pc-vulkan-library %s -emit-llvm -disable-llvm-passes -o - | FileCheck %s
+
+// CHECK: define spir_func noundef i32 @_Z6test_1v() [[A0:#[0-9]+]] {
+// CHECK: %[[C1:[0-9]+]] = call token @llvm.experimental.convergence.entry()
+// CHECK: call i32 @__hlsl_wave_get_lane_index() [ "convergencectrl"(token %[[C1]]) ]
+uint test_1() {
+  return WaveGetLaneIndex();
+}
+
+// CHECK-DAG: declare i32 @__hlsl_wave_get_lane_index() [[A1:#[0-9]+]]
+
+// CHECK: define spir_func noundef i32 @_Z6test_2v() [[A0]] {
+// CHECK: %[[C2:[0-9]+]] = call token @llvm.experimental.convergence.entry()
+// CHECK: call spir_func noundef i32 @_Z6test_1v() [ "convergencectrl"(token %[[C2]]) ]
+uint test_2() {
+  return test_1();
+}
+
+// CHECK-DAG: attributes [[A0]] = {{{.*}}convergent{{.*}}}
+// CHECK-DAG: attributes [[A1]] = {{{.*}}convergent{{.*}}}
diff --git a/llvm/include/llvm/IR/IntrinsicInst.h b/llvm/include/llvm/IR/IntrinsicInst.h
@@ -1782,6 +1782,19 @@ class ConvergenceControlInst : public IntrinsicInst {
   static bool classof(const Value *V) {
     return isa<IntrinsicInst>(V) && classof(cast<IntrinsicInst>(V));
   }
+
+  // Returns the convergence intrinsic referenced by |I|'s convergencectrl
+  // attribute if any.
+  static IntrinsicInst *getParentConvergenceToken(Instruction *I) {
+    auto *CI = dyn_cast<llvm::CallInst>(I);
+    if (!CI)
+      return nullptr;
+
+    auto Bundle = CI->getOperandBundle(llvm::LLVMContext::OB_convergencectrl);
+    assert(Bundle->Inputs.size() == 1 &&
+           Bundle->Inputs[0]->getType()->isTokenTy());
+    return dyn_cast<llvm::IntrinsicInst>(Bundle->Inputs[0].get());
+  }
 };
 
 } // end namespace llvm
