https://github.com/changpeng updated
https://github.com/llvm/llvm-project/pull/86707
>From 485dff66813104ad73d8eada7cd7d43edf9d093d Mon Sep 17 00:00:00 2001
From: Changpeng Fang <changpeng.f...@amd.com>
Date: Tue, 26 Mar 2024 11:06:48 -0700
Subject: [PATCH 1/2] AMDGPU: Simplify EmitAMDGPUBuiltinExpr for load
transposes, NFC
We should not manually get the types of the loading data.
Instead, we can get the types from the intrinsics directly.
---
clang/lib/CodeGen/CGBuiltin.cpp | 12 ++----------
1 file changed, 2 insertions(+), 10 deletions(-)
diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index 46a815155e7b87..d05ea9f626ba86 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -18544,31 +18544,23 @@ Value
*CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID,
case AMDGPU::BI__builtin_amdgcn_global_load_tr_b128_v8i16: {
Intrinsic::ID IID;
- llvm::Type *ArgTy;
switch (BuiltinID) {
case AMDGPU::BI__builtin_amdgcn_global_load_tr_b64_i32:
- ArgTy = llvm::Type::getInt32Ty(getLLVMContext());
IID = Intrinsic::amdgcn_global_load_tr_b64;
break;
case AMDGPU::BI__builtin_amdgcn_global_load_tr_b64_v2i32:
- ArgTy = llvm::FixedVectorType::get(
- llvm::Type::getInt32Ty(getLLVMContext()), 2);
IID = Intrinsic::amdgcn_global_load_tr_b64;
break;
case AMDGPU::BI__builtin_amdgcn_global_load_tr_b128_v4i16:
- ArgTy = llvm::FixedVectorType::get(
- llvm::Type::getInt16Ty(getLLVMContext()), 4);
IID = Intrinsic::amdgcn_global_load_tr_b128;
break;
case AMDGPU::BI__builtin_amdgcn_global_load_tr_b128_v8i16:
- ArgTy = llvm::FixedVectorType::get(
- llvm::Type::getInt16Ty(getLLVMContext()), 8);
IID = Intrinsic::amdgcn_global_load_tr_b128;
break;
}
-
+ llvm::Type *LoadTy = ConvertType(E->getType());
llvm::Value *Addr = EmitScalarExpr(E->getArg(0));
- llvm::Function *F = CGM.getIntrinsic(IID, {ArgTy});
+ llvm::Function *F = CGM.getIntrinsic(IID, {});
return Builder.CreateCall(F, {Addr});
}
case AMDGPU::BI__builtin_amdgcn_get_fpenv: {
>From 2375ed4269c4feedf5767383838bb6c5d4cfd80c Mon Sep 17 00:00:00 2001
From: Changpeng Fang <changpeng.f...@amd.com>
Date: Tue, 26 Mar 2024 11:17:40 -0700
Subject: [PATCH 2/2] AMDGPU: Simplify EmitAMDGPUBuiltinExpr for load
transposes, NFC
---
clang/lib/CodeGen/CGBuiltin.cpp | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index d05ea9f626ba86..d25d79d085a8eb 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -18560,7 +18560,7 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned
BuiltinID,
}
llvm::Type *LoadTy = ConvertType(E->getType());
llvm::Value *Addr = EmitScalarExpr(E->getArg(0));
- llvm::Function *F = CGM.getIntrinsic(IID, {});
+ llvm::Function *F = CGM.getIntrinsic(IID, {LoadTy});
return Builder.CreateCall(F, {Addr});
}
case AMDGPU::BI__builtin_amdgcn_get_fpenv: {
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
