aboutsummaryrefslogtreecommitdiff
path: root/clang/lib/CodeGen
diff options
context:
space:
mode:
Diffstat (limited to 'clang/lib/CodeGen')
-rw-r--r--clang/lib/CodeGen/CGAtomic.cpp12
-rw-r--r--clang/lib/CodeGen/CGOpenMPRuntime.cpp5
-rw-r--r--clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp11
-rw-r--r--clang/lib/CodeGen/TargetBuiltins/ARM.cpp24
-rw-r--r--clang/lib/CodeGen/Targets/RISCV.cpp24
-rw-r--r--clang/lib/CodeGen/Targets/SystemZ.cpp4
6 files changed, 48 insertions, 32 deletions
diff --git a/clang/lib/CodeGen/CGAtomic.cpp b/clang/lib/CodeGen/CGAtomic.cpp
index 4a3446a..d95dab3 100644
--- a/clang/lib/CodeGen/CGAtomic.cpp
+++ b/clang/lib/CodeGen/CGAtomic.cpp
@@ -507,6 +507,18 @@ static llvm::Value *EmitPostAtomicMinMax(CGBuilderTy &Builder,
bool IsSigned,
llvm::Value *OldVal,
llvm::Value *RHS) {
+ const bool IsFP = OldVal->getType()->isFloatingPointTy();
+
+ if (IsFP) {
+ llvm::Intrinsic::ID IID = (Op == AtomicExpr::AO__atomic_max_fetch ||
+ Op == AtomicExpr::AO__scoped_atomic_max_fetch)
+ ? llvm::Intrinsic::maxnum
+ : llvm::Intrinsic::minnum;
+
+ return Builder.CreateBinaryIntrinsic(IID, OldVal, RHS, llvm::FMFSource(),
+ "newval");
+ }
+
llvm::CmpInst::Predicate Pred;
switch (Op) {
default:
diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.cpp b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
index fa94692..1ff2be7 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntime.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
@@ -1762,8 +1762,11 @@ void CGOpenMPRuntime::emitDeclareTargetFunction(const FunctionDecl *FD,
// access its value.
llvm::GlobalValue *Addr = GV;
if (CGM.getLangOpts().OpenMPIsTargetDevice) {
+ llvm::PointerType *FnPtrTy = llvm::PointerType::get(
+ CGM.getLLVMContext(),
+ CGM.getModule().getDataLayout().getProgramAddressSpace());
Addr = new llvm::GlobalVariable(
- CGM.getModule(), CGM.VoidPtrTy,
+ CGM.getModule(), FnPtrTy,
/*isConstant=*/true, llvm::GlobalValue::ExternalLinkage, GV, Name,
nullptr, llvm::GlobalValue::NotThreadLocal,
CGM.getModule().getDataLayout().getDefaultGlobalsAddressSpace());
diff --git a/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp b/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
index 3613b6a..fddeba9 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
@@ -7,7 +7,7 @@
//===----------------------------------------------------------------------===//
//
// This provides a generalized class for OpenMP runtime code generation
-// specialized by GPU targets NVPTX and AMDGCN.
+// specialized by GPU targets NVPTX, AMDGCN and SPIR-V.
//
//===----------------------------------------------------------------------===//
@@ -1242,12 +1242,13 @@ void CGOpenMPRuntimeGPU::emitParallelCall(
CGBuilderTy &Bld = CGF.Builder;
llvm::Value *NumThreadsVal = NumThreads;
llvm::Function *WFn = WrapperFunctionsMap[OutlinedFn];
- llvm::Value *ID = llvm::ConstantPointerNull::get(CGM.Int8PtrTy);
- if (WFn)
- ID = Bld.CreateBitOrPointerCast(WFn, CGM.Int8PtrTy);
- llvm::Type *FnPtrTy = llvm::PointerType::get(
+ llvm::PointerType *FnPtrTy = llvm::PointerType::get(
CGF.getLLVMContext(), CGM.getDataLayout().getProgramAddressSpace());
+ llvm::Value *ID = llvm::ConstantPointerNull::get(FnPtrTy);
+ if (WFn)
+ ID = Bld.CreateBitOrPointerCast(WFn, FnPtrTy);
+
llvm::Value *FnPtr = Bld.CreateBitOrPointerCast(OutlinedFn, FnPtrTy);
// Create a private scope that will globalize the arguments
diff --git a/clang/lib/CodeGen/TargetBuiltins/ARM.cpp b/clang/lib/CodeGen/TargetBuiltins/ARM.cpp
index 82b71e3..2429a43 100644
--- a/clang/lib/CodeGen/TargetBuiltins/ARM.cpp
+++ b/clang/lib/CodeGen/TargetBuiltins/ARM.cpp
@@ -7795,7 +7795,7 @@ Value *CodeGenFunction::EmitAArch64BuiltinExpr(unsigned BuiltinID,
}
case NEON::BI__builtin_neon_vcvt1_low_bf16_mf8_fpm:
ExtractLow = true;
- LLVM_FALLTHROUGH;
+ [[fallthrough]];
case NEON::BI__builtin_neon_vcvt1_bf16_mf8_fpm:
case NEON::BI__builtin_neon_vcvt1_high_bf16_mf8_fpm:
return EmitFP8NeonCvtCall(Intrinsic::aarch64_neon_fp8_cvtl1,
@@ -7803,7 +7803,7 @@ Value *CodeGenFunction::EmitAArch64BuiltinExpr(unsigned BuiltinID,
Ops[0]->getType(), ExtractLow, Ops, E, "vbfcvt1");
case NEON::BI__builtin_neon_vcvt2_low_bf16_mf8_fpm:
ExtractLow = true;
- LLVM_FALLTHROUGH;
+ [[fallthrough]];
case NEON::BI__builtin_neon_vcvt2_bf16_mf8_fpm:
case NEON::BI__builtin_neon_vcvt2_high_bf16_mf8_fpm:
return EmitFP8NeonCvtCall(Intrinsic::aarch64_neon_fp8_cvtl2,
@@ -7811,7 +7811,7 @@ Value *CodeGenFunction::EmitAArch64BuiltinExpr(unsigned BuiltinID,
Ops[0]->getType(), ExtractLow, Ops, E, "vbfcvt2");
case NEON::BI__builtin_neon_vcvt1_low_f16_mf8_fpm:
ExtractLow = true;
- LLVM_FALLTHROUGH;
+ [[fallthrough]];
case NEON::BI__builtin_neon_vcvt1_f16_mf8_fpm:
case NEON::BI__builtin_neon_vcvt1_high_f16_mf8_fpm:
return EmitFP8NeonCvtCall(Intrinsic::aarch64_neon_fp8_cvtl1,
@@ -7819,7 +7819,7 @@ Value *CodeGenFunction::EmitAArch64BuiltinExpr(unsigned BuiltinID,
Ops[0]->getType(), ExtractLow, Ops, E, "vbfcvt1");
case NEON::BI__builtin_neon_vcvt2_low_f16_mf8_fpm:
ExtractLow = true;
- LLVM_FALLTHROUGH;
+ [[fallthrough]];
case NEON::BI__builtin_neon_vcvt2_f16_mf8_fpm:
case NEON::BI__builtin_neon_vcvt2_high_f16_mf8_fpm:
return EmitFP8NeonCvtCall(Intrinsic::aarch64_neon_fp8_cvtl2,
@@ -7854,7 +7854,7 @@ Value *CodeGenFunction::EmitAArch64BuiltinExpr(unsigned BuiltinID,
case NEON::BI__builtin_neon_vdot_lane_f16_mf8_fpm:
case NEON::BI__builtin_neon_vdotq_lane_f16_mf8_fpm:
ExtendLaneArg = true;
- LLVM_FALLTHROUGH;
+ [[fallthrough]];
case NEON::BI__builtin_neon_vdot_laneq_f16_mf8_fpm:
case NEON::BI__builtin_neon_vdotq_laneq_f16_mf8_fpm:
return EmitFP8NeonFDOTCall(Intrinsic::aarch64_neon_fp8_fdot2_lane,
@@ -7866,7 +7866,7 @@ Value *CodeGenFunction::EmitAArch64BuiltinExpr(unsigned BuiltinID,
case NEON::BI__builtin_neon_vdot_lane_f32_mf8_fpm:
case NEON::BI__builtin_neon_vdotq_lane_f32_mf8_fpm:
ExtendLaneArg = true;
- LLVM_FALLTHROUGH;
+ [[fallthrough]];
case NEON::BI__builtin_neon_vdot_laneq_f32_mf8_fpm:
case NEON::BI__builtin_neon_vdotq_laneq_f32_mf8_fpm:
return EmitFP8NeonFDOTCall(Intrinsic::aarch64_neon_fp8_fdot4_lane,
@@ -7898,37 +7898,37 @@ Value *CodeGenFunction::EmitAArch64BuiltinExpr(unsigned BuiltinID,
"vmlall");
case NEON::BI__builtin_neon_vmlalbq_lane_f16_mf8_fpm:
ExtendLaneArg = true;
- LLVM_FALLTHROUGH;
+ [[fallthrough]];
case NEON::BI__builtin_neon_vmlalbq_laneq_f16_mf8_fpm:
return EmitFP8NeonFMLACall(Intrinsic::aarch64_neon_fp8_fmlalb_lane,
ExtendLaneArg, HalfTy, Ops, E, "vmlal_lane");
case NEON::BI__builtin_neon_vmlaltq_lane_f16_mf8_fpm:
ExtendLaneArg = true;
- LLVM_FALLTHROUGH;
+ [[fallthrough]];
case NEON::BI__builtin_neon_vmlaltq_laneq_f16_mf8_fpm:
return EmitFP8NeonFMLACall(Intrinsic::aarch64_neon_fp8_fmlalt_lane,
ExtendLaneArg, HalfTy, Ops, E, "vmlal_lane");
case NEON::BI__builtin_neon_vmlallbbq_lane_f32_mf8_fpm:
ExtendLaneArg = true;
- LLVM_FALLTHROUGH;
+ [[fallthrough]];
case NEON::BI__builtin_neon_vmlallbbq_laneq_f32_mf8_fpm:
return EmitFP8NeonFMLACall(Intrinsic::aarch64_neon_fp8_fmlallbb_lane,
ExtendLaneArg, FloatTy, Ops, E, "vmlall_lane");
case NEON::BI__builtin_neon_vmlallbtq_lane_f32_mf8_fpm:
ExtendLaneArg = true;
- LLVM_FALLTHROUGH;
+ [[fallthrough]];
case NEON::BI__builtin_neon_vmlallbtq_laneq_f32_mf8_fpm:
return EmitFP8NeonFMLACall(Intrinsic::aarch64_neon_fp8_fmlallbt_lane,
ExtendLaneArg, FloatTy, Ops, E, "vmlall_lane");
case NEON::BI__builtin_neon_vmlalltbq_lane_f32_mf8_fpm:
ExtendLaneArg = true;
- LLVM_FALLTHROUGH;
+ [[fallthrough]];
case NEON::BI__builtin_neon_vmlalltbq_laneq_f32_mf8_fpm:
return EmitFP8NeonFMLACall(Intrinsic::aarch64_neon_fp8_fmlalltb_lane,
ExtendLaneArg, FloatTy, Ops, E, "vmlall_lane");
case NEON::BI__builtin_neon_vmlallttq_lane_f32_mf8_fpm:
ExtendLaneArg = true;
- LLVM_FALLTHROUGH;
+ [[fallthrough]];
case NEON::BI__builtin_neon_vmlallttq_laneq_f32_mf8_fpm:
return EmitFP8NeonFMLACall(Intrinsic::aarch64_neon_fp8_fmlalltt_lane,
ExtendLaneArg, FloatTy, Ops, E, "vmlall_lane");
diff --git a/clang/lib/CodeGen/Targets/RISCV.cpp b/clang/lib/CodeGen/Targets/RISCV.cpp
index 0ef39b6..0d0941e 100644
--- a/clang/lib/CodeGen/Targets/RISCV.cpp
+++ b/clang/lib/CodeGen/Targets/RISCV.cpp
@@ -680,22 +680,22 @@ ABIArgInfo RISCVABIInfo::classifyArgumentType(QualType Ty, bool IsFixed,
if (const auto *ED = Ty->getAsEnumDecl())
Ty = ED->getIntegerType();
- // All integral types are promoted to XLen width
- if (Size < XLen && Ty->isIntegralOrEnumerationType()) {
- return extendType(Ty, CGT.ConvertType(Ty));
- }
-
if (const auto *EIT = Ty->getAs<BitIntType>()) {
- if (EIT->getNumBits() < XLen)
+
+ if (XLen == 64 && EIT->getNumBits() == 32)
return extendType(Ty, CGT.ConvertType(Ty));
- if (EIT->getNumBits() > 128 ||
- (!getContext().getTargetInfo().hasInt128Type() &&
- EIT->getNumBits() > 64))
- return getNaturalAlignIndirect(
- Ty, /*AddrSpace=*/getDataLayout().getAllocaAddrSpace(),
- /*ByVal=*/false);
+
+ if (EIT->getNumBits() <= 2 * XLen)
+ return ABIArgInfo::getExtend(Ty, CGT.ConvertType(Ty));
+ return getNaturalAlignIndirect(
+ Ty, /*AddrSpace=*/getDataLayout().getAllocaAddrSpace(),
+ /*ByVal=*/false);
}
+ // All integral types are promoted to XLen width
+ if (Size < XLen && Ty->isIntegralOrEnumerationType())
+ return extendType(Ty, CGT.ConvertType(Ty));
+
return ABIArgInfo::getDirect();
}
diff --git a/clang/lib/CodeGen/Targets/SystemZ.cpp b/clang/lib/CodeGen/Targets/SystemZ.cpp
index 9b6b72b1..e50f06c 100644
--- a/clang/lib/CodeGen/Targets/SystemZ.cpp
+++ b/clang/lib/CodeGen/Targets/SystemZ.cpp
@@ -193,11 +193,11 @@ llvm::Type *SystemZABIInfo::getFPArgumentType(QualType Ty,
case BuiltinType::Float16:
if (Size == 16)
return llvm::Type::getHalfTy(getVMContext());
- LLVM_FALLTHROUGH;
+ [[fallthrough]];
case BuiltinType::Float:
if (Size == 32)
return llvm::Type::getFloatTy(getVMContext());
- LLVM_FALLTHROUGH;
+ [[fallthrough]];
case BuiltinType::Double:
return llvm::Type::getDoubleTy(getVMContext());
default:
generated by cgit v1.1 at 2025-10-14 11:14:07 +0000