https://github.com/jhuber6 updated
https://github.com/llvm/llvm-project/pull/79873
>From 35e12c3d83f3be93618805ffaf05e3424689f32f Mon Sep 17 00:00:00 2001
From: Joseph Huber <hube...@outlook.com>
Date: Mon, 29 Jan 2024 11:08:04 -0600
Subject: [PATCH 1/2] [NVPTX] Allow compiling LLVM-IR without `-march` set
Summary:
The NVPTX tools require an architecture to be used, however if we are
creating generic LLVM-IR we should be able to leave it unspecified. This
will result in the `target-cpu` attributes not being set on the
functions so it can be changed when linked into code. This allows the
standalone `--target=nvptx64-nvidia-cuda` toolchain to create LLVM-IR
simmilar to how CUDA's deviceRTL looks from C/C++
---
.../clang/Basic/DiagnosticDriverKinds.td | 2 ++
clang/lib/Basic/Targets/NVPTX.cpp | 7 +++++-
clang/lib/Basic/Targets/NVPTX.h | 3 ++-
clang/lib/Driver/ToolChains/Cuda.cpp | 19 ++++++++++-----
clang/test/Driver/cuda-cross-compiling.c | 24 +++++++++++--------
.../Preprocessor/predefined-arch-macros.c | 12 ++++++++++
6 files changed, 49 insertions(+), 18 deletions(-)
diff --git a/clang/include/clang/Basic/DiagnosticDriverKinds.td
b/clang/include/clang/Basic/DiagnosticDriverKinds.td
index 094fe19509412..476528375fb88 100644
--- a/clang/include/clang/Basic/DiagnosticDriverKinds.td
+++ b/clang/include/clang/Basic/DiagnosticDriverKinds.td
@@ -57,6 +57,8 @@ def warn_drv_avr_stdlib_not_linked: Warning<
InGroup<AVRRtlibLinkingQuirks>;
def err_drv_cuda_bad_gpu_arch : Error<"unsupported CUDA gpu architecture: %0">;
def err_drv_offload_bad_gpu_arch : Error<"unsupported %0 gpu architecture:
%1">;
+def err_drv_offload_missing_gpu_arch : Error<
+ "Must pass in an explicit %0 gpu architecture to '%1'">;
def err_drv_no_cuda_installation : Error<
"cannot find CUDA installation; provide its path via '--cuda-path', or pass "
"'-nocudainc' to build without CUDA includes">;
diff --git a/clang/lib/Basic/Targets/NVPTX.cpp
b/clang/lib/Basic/Targets/NVPTX.cpp
index 0b9d97f69d146..7687e3faad770 100644
--- a/clang/lib/Basic/Targets/NVPTX.cpp
+++ b/clang/lib/Basic/Targets/NVPTX.cpp
@@ -59,7 +59,7 @@ NVPTXTargetInfo::NVPTXTargetInfo(const llvm::Triple &Triple,
// Define available target features
// These must be defined in sorted order!
NoAsmVariants = true;
- GPU = CudaArch::SM_20;
+ GPU = CudaArch::UNUSED;
if (TargetPointerWidth == 32)
resetDataLayout("e-p:32:32-i64:64-i128:128-v16:16-v32:32-n16:32:64");
@@ -169,6 +169,11 @@ void NVPTXTargetInfo::getTargetDefines(const LangOptions
&Opts,
MacroBuilder &Builder) const {
Builder.defineMacro("__PTX__");
Builder.defineMacro("__NVPTX__");
+
+ // Skip setting architecture dependent macros if undefined.
+ if (GPU == CudaArch::UNUSED && !HostTarget)
+ return;
+
if (Opts.CUDAIsDevice || Opts.OpenMPIsTargetDevice || !HostTarget) {
// Set __CUDA_ARCH__ for the GPU specified.
std::string CUDAArchCode = [this] {
diff --git a/clang/lib/Basic/Targets/NVPTX.h b/clang/lib/Basic/Targets/NVPTX.h
index 20d76b702a942..f476d49047c01 100644
--- a/clang/lib/Basic/Targets/NVPTX.h
+++ b/clang/lib/Basic/Targets/NVPTX.h
@@ -79,7 +79,8 @@ class LLVM_LIBRARY_VISIBILITY NVPTXTargetInfo : public
TargetInfo {
initFeatureMap(llvm::StringMap<bool> &Features, DiagnosticsEngine &Diags,
StringRef CPU,
const std::vector<std::string> &FeaturesVec) const override {
- Features[CudaArchToString(GPU)] = true;
+ if (GPU != CudaArch::UNUSED)
+ Features[CudaArchToString(GPU)] = true;
Features["ptx" + std::to_string(PTXVersion)] = true;
return TargetInfo::initFeatureMap(Features, Diags, CPU, FeaturesVec);
}
diff --git a/clang/lib/Driver/ToolChains/Cuda.cpp
b/clang/lib/Driver/ToolChains/Cuda.cpp
index 8a9d0caaccf30..ca54d2d55426b 100644
--- a/clang/lib/Driver/ToolChains/Cuda.cpp
+++ b/clang/lib/Driver/ToolChains/Cuda.cpp
@@ -389,7 +389,11 @@ void NVPTX::Assembler::ConstructJob(Compilation &C, const
JobAction &JA,
GPUArchName = JA.getOffloadingArch();
} else {
GPUArchName = Args.getLastArgValue(options::OPT_march_EQ);
- assert(!GPUArchName.empty() && "Must have an architecture passed in.");
+ if (GPUArchName.empty()) {
+ C.getDriver().Diag(diag::err_drv_offload_missing_gpu_arch)
+ << getToolChain().getArchName() << getShortName();
+ return;
+ }
}
// Obtain architecture from the action.
@@ -593,7 +597,11 @@ void NVPTX::Linker::ConstructJob(Compilation &C, const
JobAction &JA,
CmdArgs.push_back("-v");
StringRef GPUArch = Args.getLastArgValue(options::OPT_march_EQ);
- assert(!GPUArch.empty() && "At least one GPU Arch required for nvlink.");
+ if (GPUArch.empty()) {
+ C.getDriver().Diag(diag::err_drv_offload_missing_gpu_arch)
+ << getToolChain().getArchName() << getShortName();
+ return;
+ }
CmdArgs.push_back("-arch");
CmdArgs.push_back(Args.MakeArgString(GPUArch));
@@ -726,9 +734,8 @@ NVPTXToolChain::NVPTXToolChain(const Driver &D, const
llvm::Triple &Triple,
llvm::opt::DerivedArgList *
NVPTXToolChain::TranslateArgs(const llvm::opt::DerivedArgList &Args,
StringRef BoundArch,
- Action::OffloadKind DeviceOffloadKind) const {
- DerivedArgList *DAL =
- ToolChain::TranslateArgs(Args, BoundArch, DeviceOffloadKind);
+ Action::OffloadKind OffloadKind) const {
+ DerivedArgList *DAL = ToolChain::TranslateArgs(Args, BoundArch, OffloadKind);
if (!DAL)
DAL = new DerivedArgList(Args.getBaseArgs());
@@ -738,7 +745,7 @@ NVPTXToolChain::TranslateArgs(const
llvm::opt::DerivedArgList &Args,
if (!llvm::is_contained(*DAL, A))
DAL->append(A);
- if (!DAL->hasArg(options::OPT_march_EQ)) {
+ if (!DAL->hasArg(options::OPT_march_EQ) && OffloadKind != Action::OFK_None) {
DAL->AddJoinedArg(nullptr, Opts.getOption(options::OPT_march_EQ),
CudaArchToString(CudaArch::CudaDefault));
} else if (DAL->getLastArgValue(options::OPT_march_EQ) == "native") {
diff --git a/clang/test/Driver/cuda-cross-compiling.c
b/clang/test/Driver/cuda-cross-compiling.c
index 12d0af3b45f32..6c9e2cac736b7 100644
--- a/clang/test/Driver/cuda-cross-compiling.c
+++ b/clang/test/Driver/cuda-cross-compiling.c
@@ -59,16 +59,6 @@
// LINK: nvlink{{.*}}"-o" "a.out" "-arch" "sm_61" {{.*}} "{{.*}}.cubin"
-//
-// Test the generated arguments default to a value with no architecture.
-//
-// RUN: %clang --target=nvptx64-nvidia-cuda -###
--cuda-path=%S/Inputs/CUDA/usr/local/cuda %s 2>&1 \
-// RUN: | FileCheck -check-prefix=DEFAULT %s
-
-// DEFAULT: -cc1" "-triple" "nvptx64-nvidia-cuda" "-S" {{.*}}
"-target-cpu" "sm_52" "-target-feature" "+ptx{{[0-9]+}}" {{.*}} "-o"
"[[PTX:.+]].s"
-// DEFAULT-NEXT: ptxas{{.*}}"-m64" "-O0" "--gpu-name" "sm_52" "--output-file"
"[[CUBIN:.+]].cubin" "[[PTX]].s" "-c"
-// DEFAULT-NEXT: nvlink{{.*}}"-o" "a.out" "-arch" "sm_52" {{.*}}
"[[CUBIN]].cubin"
-
//
// Test to ensure that we enable handling global constructors in a freestanding
// Nvidia compilation.
@@ -77,3 +67,17 @@
// RUN: | FileCheck -check-prefix=LOWERING %s
// LOWERING: -cc1" "-triple" "nvptx64-nvidia-cuda" {{.*}} "-mllvm"
"--nvptx-lower-global-ctor-dtor"
+
+//
+// Tests for handling a missing architecture.
+//
+// RUN: not %clang -target nvptx64-nvidia-cuda %s -### 2>&1 \
+// RUN: | FileCheck -check-prefix=MISSING %s
+
+// MISSING: error: Must pass in an explicit nvptx64 gpu architecture to 'ptxas'
+// MISSING: error: Must pass in an explicit nvptx64 gpu architecture to
'nvlink'
+
+// RUN: %clang -target nvptx64-nvidia-cuda -flto -c %s -### 2>&1 \
+// RUN: | FileCheck -check-prefix=GENERIC %s
+
+// GENERIC-NOT: -cc1" "-triple" "nvptx64-nvidia-cuda" {{.*}} "-target-cpu"
diff --git a/clang/test/Preprocessor/predefined-arch-macros.c
b/clang/test/Preprocessor/predefined-arch-macros.c
index 27c7b4a271fee..be9323adb4f64 100644
--- a/clang/test/Preprocessor/predefined-arch-macros.c
+++ b/clang/test/Preprocessor/predefined-arch-macros.c
@@ -4292,6 +4292,18 @@
// RUN: | FileCheck -match-full-lines %s -check-prefix=CHECK_SYSTEMZ_ZVECTOR
// CHECK_SYSTEMZ_ZVECTOR: #define __VEC__ 10304
+// Begin nvptx tests ----------------
+
+// RUN: %clang -march=sm_75 -E -dM %s -o - 2>&1 \
+// RUN: -target nvptx64-unknown-unknown \
+// RUN: | FileCheck -match-full-lines %s
-check-prefixes=CHECK_NVPTX,CHECK_ARCH_SM_75
+// RUN: %clang -E -dM %s -o - 2>&1 \
+// RUN: -target nvptx64-unknown-unknown \
+// RUN: | FileCheck -match-full-lines %s
-check-prefixes=CHECK_NVPTX,CHECK_ARCH_UNSET
+// CHECK_ARCH_SM_75: #define __CUDA_ARCH__ 750
+// CHECK_ARCH_UNSET-NOT: #define __CUDA_ARCH__
+// CHECK_NVPTX: #define __NVPTX__ 1
+
// Begin amdgcn tests ----------------
// RUN: %clang -march=amdgcn -E -dM %s -o - 2>&1 \
>From a76203edb9d46ac0e0e75a21dcfe4993b3fc6cfd Mon Sep 17 00:00:00 2001
From: Joseph Huber <hube...@outlook.com>
Date: Tue, 30 Jan 2024 17:16:59 -0600
Subject: [PATCH 2/2] Add TODO
---
clang/lib/Driver/ToolChains/Cuda.cpp | 1 +
1 file changed, 1 insertion(+)
diff --git a/clang/lib/Driver/ToolChains/Cuda.cpp
b/clang/lib/Driver/ToolChains/Cuda.cpp
index ca54d2d55426b..ed5924c3b73b5 100644
--- a/clang/lib/Driver/ToolChains/Cuda.cpp
+++ b/clang/lib/Driver/ToolChains/Cuda.cpp
@@ -745,6 +745,7 @@ NVPTXToolChain::TranslateArgs(const
llvm::opt::DerivedArgList &Args,
if (!llvm::is_contained(*DAL, A))
DAL->append(A);
+ // TODO: We should accept 'generic' as a valid architecture.
if (!DAL->hasArg(options::OPT_march_EQ) && OffloadKind != Action::OFK_None) {
DAL->AddJoinedArg(nullptr, Opts.getOption(options::OPT_march_EQ),
CudaArchToString(CudaArch::CudaDefault));
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
