jhuber6 created this revision.
jhuber6 added reviewers: JonChesterfield, jdoerfert, tra, yaxunl.
Herald added a subscriber: Anastasia.
Herald added a project: All.
jhuber6 requested review of this revision.
Herald added subscribers: cfe-commits, jplehr, sstefan1, MaskRay.
Herald added a project: clang.
These headers are currently broken when included from the offloading
languages like OpenMP, OpenCL, CUDA, and HIP. Turn this logic off so we
can compile these languages when the GPU libc is installed. I am
currently trying to remedy this and have made an RFC for it in libc,
see
https://discourse.llvm.org/t/rfc-implementing-gpu-headers-in-the-llvm-c-library/71523.
Repository:
rG LLVM Github Monorepo
https://reviews.llvm.org/D153578
Files:
clang/lib/Driver/ToolChains/Clang.cpp
clang/test/Driver/gpu-libc-headers.c
Index: clang/test/Driver/gpu-libc-headers.c
===================================================================
--- clang/test/Driver/gpu-libc-headers.c
+++ clang/test/Driver/gpu-libc-headers.c
@@ -1,18 +1,6 @@
// REQUIRES: nvptx-registered-target
// REQUIRES: amdgpu-registered-target
-// RUN: %clang -### --target=x86_64-unknown-linux-gnu -fopenmp=libomp
--sysroot=./ \
-// RUN: -fopenmp-targets=amdgcn-amd-amdhsa
-Xopenmp-target=amdgcn-amd-amdhsa --offload-arch=gfx908 \
-// RUN: -nogpulib %s 2>&1 | FileCheck %s --check-prefix=CHECK-HEADERS
-// RUN: %clang -### --target=x86_64-unknown-linux-gnu -fopenmp=libomp
--sysroot=./ \
-// RUN: -fopenmp-targets=nvptx64-nvidia-cuda
-Xopenmp-target=nvptx64-nvidia-cuda --offload-arch=sm_70 \
-// RUN: -nogpulib %s 2>&1 | FileCheck %s --check-prefix=CHECK-HEADERS
-// RUN: %clang -### --target=nvptx64-nvidia-cuda -march=sm_70 -nogpulib
--sysroot=./ %s 2>&1 | \
-// RUN: FileCheck %s --check-prefix=CHECK-HEADERS
-// RUN: %clang -### --target=amdgcn-amd-amdhsa -mcpu=gfx1030 -nogpulib
--sysroot=./ %s 2>&1 | \
-// RUN: FileCheck %s --check-prefix=CHECK-HEADERS
-// CHECK-HEADERS: "-cc1"{{.*}}"-c-isystem"
"{{.*}}include{{.*}}gpu-none-llvm"{{.*}}"-isysroot" "./"
-
// RUN: %clang -### --target=amdgcn-amd-amdhsa -mcpu=gfx1030 -nogpulib \
// RUN: -nogpuinc %s 2>&1 | FileCheck %s
--check-prefix=CHECK-HEADERS-DISABLED
// RUN: %clang -### --target=amdgcn-amd-amdhsa -mcpu=gfx1030 -nogpulib \
Index: clang/lib/Driver/ToolChains/Clang.cpp
===================================================================
--- clang/lib/Driver/ToolChains/Clang.cpp
+++ clang/lib/Driver/ToolChains/Clang.cpp
@@ -1181,16 +1181,18 @@
// If we are compiling for a GPU target we want to override the system
headers
// with ones created by the 'libc' project if present.
+ // FIXME: We need to find a way to make these headers compatible with the
+ // host environment so they can be included from offloading languages. For
now
+ // these are only active when targeting the GPU with cross-compilation.
if (!Args.hasArg(options::OPT_nostdinc) &&
!Args.hasArg(options::OPT_nogpuinc) &&
!Args.hasArg(options::OPT_nobuiltininc) &&
+ C.getActiveOffloadKinds() == Action::OFK_None &&
(getToolChain().getTriple().isNVPTX() ||
getToolChain().getTriple().isAMDGCN())) {
// Add include/gpu-none-libc/* to our system include path. This lets us
use
// GPU-specific system headers first.
- // TODO: We need to find a way to make these headers compatible with the
- // host environment.
SmallString<128> P(llvm::sys::path::parent_path(D.InstalledDir));
llvm::sys::path::append(P, "include");
llvm::sys::path::append(P, "gpu-none-llvm");
Index: clang/test/Driver/gpu-libc-headers.c
===================================================================
--- clang/test/Driver/gpu-libc-headers.c
+++ clang/test/Driver/gpu-libc-headers.c
@@ -1,18 +1,6 @@
// REQUIRES: nvptx-registered-target
// REQUIRES: amdgpu-registered-target
-// RUN: %clang -### --target=x86_64-unknown-linux-gnu -fopenmp=libomp --sysroot=./ \
-// RUN: -fopenmp-targets=amdgcn-amd-amdhsa -Xopenmp-target=amdgcn-amd-amdhsa --offload-arch=gfx908 \
-// RUN: -nogpulib %s 2>&1 | FileCheck %s --check-prefix=CHECK-HEADERS
-// RUN: %clang -### --target=x86_64-unknown-linux-gnu -fopenmp=libomp --sysroot=./ \
-// RUN: -fopenmp-targets=nvptx64-nvidia-cuda -Xopenmp-target=nvptx64-nvidia-cuda --offload-arch=sm_70 \
-// RUN: -nogpulib %s 2>&1 | FileCheck %s --check-prefix=CHECK-HEADERS
-// RUN: %clang -### --target=nvptx64-nvidia-cuda -march=sm_70 -nogpulib --sysroot=./ %s 2>&1 | \
-// RUN: FileCheck %s --check-prefix=CHECK-HEADERS
-// RUN: %clang -### --target=amdgcn-amd-amdhsa -mcpu=gfx1030 -nogpulib --sysroot=./ %s 2>&1 | \
-// RUN: FileCheck %s --check-prefix=CHECK-HEADERS
-// CHECK-HEADERS: "-cc1"{{.*}}"-c-isystem" "{{.*}}include{{.*}}gpu-none-llvm"{{.*}}"-isysroot" "./"
-
// RUN: %clang -### --target=amdgcn-amd-amdhsa -mcpu=gfx1030 -nogpulib \
// RUN: -nogpuinc %s 2>&1 | FileCheck %s --check-prefix=CHECK-HEADERS-DISABLED
// RUN: %clang -### --target=amdgcn-amd-amdhsa -mcpu=gfx1030 -nogpulib \
Index: clang/lib/Driver/ToolChains/Clang.cpp
===================================================================
--- clang/lib/Driver/ToolChains/Clang.cpp
+++ clang/lib/Driver/ToolChains/Clang.cpp
@@ -1181,16 +1181,18 @@
// If we are compiling for a GPU target we want to override the system headers
// with ones created by the 'libc' project if present.
+ // FIXME: We need to find a way to make these headers compatible with the
+ // host environment so they can be included from offloading languages. For now
+ // these are only active when targeting the GPU with cross-compilation.
if (!Args.hasArg(options::OPT_nostdinc) &&
!Args.hasArg(options::OPT_nogpuinc) &&
!Args.hasArg(options::OPT_nobuiltininc) &&
+ C.getActiveOffloadKinds() == Action::OFK_None &&
(getToolChain().getTriple().isNVPTX() ||
getToolChain().getTriple().isAMDGCN())) {
// Add include/gpu-none-libc/* to our system include path. This lets us use
// GPU-specific system headers first.
- // TODO: We need to find a way to make these headers compatible with the
- // host environment.
SmallString<128> P(llvm::sys::path::parent_path(D.InstalledDir));
llvm::sys::path::append(P, "include");
llvm::sys::path::append(P, "gpu-none-llvm");
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
