AMDGPU: Add MF independent version of getImplicitParameterOffset

llvm · Jun 7, 2023 · 3d0350b · 3d0350b
1 parent bc61bc8
commit 3d0350b
Show file tree

Hide file tree

Showing 6 changed files with 17 additions and 11 deletions.
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUCallLowering.cpp b/llvm/lib/Target/AMDGPU/AMDGPUCallLowering.cpp
@@ -521,7 +521,7 @@ bool AMDGPUCallLowering::lowerFormalArgumentsKernel(
 
   unsigned i = 0;
   const Align KernArgBaseAlign(16);
-  const unsigned BaseOffset = Subtarget->getExplicitKernelArgOffset(F);
+  const unsigned BaseOffset = Subtarget->getExplicitKernelArgOffset();
   uint64_t ExplicitArgOffset = 0;
 
   // TODO: Align down to dword alignment and extract bits for extending loads.

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
@@ -1066,7 +1066,7 @@ void AMDGPUTargetLowering::analyzeFormalArgumentsCompute(
   const Function &Fn = MF.getFunction();
   LLVMContext &Ctx = Fn.getParent()->getContext();
   const AMDGPUSubtarget &ST = AMDGPUSubtarget::get(MF);
-  const unsigned ExplicitOffset = ST.getExplicitKernelArgOffset(Fn);
+  const unsigned ExplicitOffset = ST.getExplicitKernelArgOffset();
   CallingConv::ID CC = Fn.getCallingConv();
 
   Align MaxAlign = Align(1);
@@ -4563,13 +4563,11 @@ SDValue AMDGPUTargetLowering::loadInputValue(SelectionDAG &DAG,
 }
 
 uint32_t AMDGPUTargetLowering::getImplicitParameterOffset(
-    const MachineFunction &MF, const ImplicitParameter Param) const {
-  const AMDGPUMachineFunction *MFI = MF.getInfo<AMDGPUMachineFunction>();
-  unsigned ExplicitArgOffset =
-      Subtarget->getExplicitKernelArgOffset(MF.getFunction());
+    uint64_t ExplicitKernArgSize, const ImplicitParameter Param) const {
+  unsigned ExplicitArgOffset = Subtarget->getExplicitKernelArgOffset();
   const Align Alignment = Subtarget->getAlignmentForImplicitArgPtr();
-  uint64_t ArgOffset = alignTo(MFI->getExplicitKernArgSize(), Alignment) +
-                       ExplicitArgOffset;
+  uint64_t ArgOffset =
+      alignTo(ExplicitKernArgSize, Alignment) + ExplicitArgOffset;
   switch (Param) {
   case FIRST_IMPLICIT:
     return ArgOffset;
@@ -4583,6 +4581,12 @@ uint32_t AMDGPUTargetLowering::getImplicitParameterOffset(
   llvm_unreachable("unexpected implicit parameter type");
 }
 
+uint32_t AMDGPUTargetLowering::getImplicitParameterOffset(
+    const MachineFunction &MF, const ImplicitParameter Param) const {
+  const AMDGPUMachineFunction *MFI = MF.getInfo<AMDGPUMachineFunction>();
+  return getImplicitParameterOffset(MFI->getExplicitKernArgSize(), Param);
+}
+
 #define NODE_NAME_CASE(node) case AMDGPUISD::node: return #node;
 
 const char* AMDGPUTargetLowering::getTargetNodeName(unsigned Opcode) const {

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h
@@ -344,6 +344,8 @@ class AMDGPUTargetLowering : public TargetLowering {
   /// type of implicit parameter.
   uint32_t getImplicitParameterOffset(const MachineFunction &MF,
                                       const ImplicitParameter Param) const;
+  uint32_t getImplicitParameterOffset(const uint64_t ExplicitKernArgSize,
+                                      const ImplicitParameter Param) const;
 
   MVT getFenceOperandTy(const DataLayout &DL) const override {
     return MVT::i32;

diff --git a/llvm/lib/Target/AMDGPU/AMDGPULowerKernelArguments.cpp b/llvm/lib/Target/AMDGPU/AMDGPULowerKernelArguments.cpp
@@ -70,7 +70,7 @@ bool AMDGPULowerKernelArguments::runOnFunction(Function &F) {
   IRBuilder<> Builder(&*getInsertPt(EntryBlock));
 
   const Align KernArgBaseAlign(16); // FIXME: Increase if necessary
-  const uint64_t BaseOffset = ST.getExplicitKernelArgOffset(F);
+  const uint64_t BaseOffset = ST.getExplicitKernelArgOffset();
 
   Align MaxAlign;
   // FIXME: Alignment is broken with explicit arg offset.;

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp b/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp
@@ -576,7 +576,7 @@ unsigned AMDGPUSubtarget::getKernArgSegmentSize(const Function &F,
                                                 Align &MaxAlign) const {
   uint64_t ExplicitArgBytes = getExplicitKernArgSize(F, MaxAlign);
 
-  unsigned ExplicitOffset = getExplicitKernelArgOffset(F);
+  unsigned ExplicitOffset = getExplicitKernelArgOffset();
 
   uint64_t TotalSize = ExplicitOffset + ExplicitArgBytes;
   unsigned ImplicitBytes = getImplicitArgNumBytes(F);

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.h b/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.h
@@ -226,7 +226,7 @@ class AMDGPUSubtarget {
 
   /// Returns the offset in bytes from the start of the input buffer
   ///        of the first explicit kernel argument.
-  unsigned getExplicitKernelArgOffset(const Function &F) const {
+  unsigned getExplicitKernelArgOffset() const {
     switch (TargetTriple.getOS()) {
     case Triple::AMDHSA:
     case Triple::AMDPAL: