[libcxx] Add necessary compile flags for targeting the GPU #99333
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
llvmbot
added
the
libc++
|
@llvm/pr-subscribers-libcxx Author: Joseph Huber (jhuber6) ChangesSummary: Full diff: https://github.com/llvm/llvm-project/pull/99333.diff 1 Files Affected:
diff --git a/libcxx/CMakeLists.txt b/libcxx/CMakeLists.txt
index 190a97db9462f..a72e5aca0903f 100644
--- a/libcxx/CMakeLists.txt
+++ b/libcxx/CMakeLists.txt
@@ -491,6 +491,15 @@ include(HandleLibcxxFlags)
# 'config-ix' use them during feature checks. It also adds them to both
# 'LIBCXX_COMPILE_FLAGS' and 'LIBCXX_LINK_FLAGS'
+# Targeting the GPU requires a clang compiler and several extra flags.
+if (${LLVM_RUNTIMES_TARGET} MATCHES "^amdgcn")
+ set(CMAKE_REQUIRED_FLAGS "${CMAKE_REQUIRED_FLAGS} -nogpulib")
+ add_flags("-nogpulib" "-flto" "-fconvergent-functions" "-Xclang" "-mcode-object-version=none")
+elseif (${LLVM_RUNTIMES_TARGET} MATCHES "^nvptx")
+ set(CMAKE_REQUIRED_FLAGS "${CMAKE_REQUIRED_FLAGS} -flto -c -Wno-unused-command-line-argument")
+ add_flags("-nogpulib" "-flto" "-fconvergent-functions" "--cuda-feature=+ptx63")
+endif()
+
if (${CMAKE_SYSTEM_NAME} MATCHES "AIX")
add_flags_if_supported("-mdefault-visibility-export-mapping=explicit")
set(CMAKE_AIX_EXPORT_ALL_SYMBOLS OFF)
|
This was referenced Jul 17, 2024
jhuber6
added a commit
to jhuber6/llvm-project
that referenced
this pull request
Jul 17, 2024
Summary: This patch adds a CMake cache config file for the GPU build. This cache will set the default required options when used from the LLVM runtime interface or directly. These options pretty much disable everything the GPU can't handle. With this and the fllowing patches: llvm#99259, llvm#99243, llvm#99287, and llvm#99333, we will be able to build `libc++` targeting the GPU with an invocation like this. ``` $ cmake ../llvm -C${LLVM_ROOT}/libcxx/cmake/caches/GPU.cmake \ -DRUNTIMES_nvptx64-nvidia-cuda_LLVM_ENABLE_RUNTIMES=compiler-rt;libc;libcxx \ -DRUNTIMES_amdgcn-amd-amdhsa_LLVM_ENABLE_RUNTIMES=compiler-rt;libc;libcxx \ -DLLVM_RUNTIME_TARGETS=amdgcn-amd-amdhsa;nvptx64-nvidia-cuda \ ``` This will then install the libraries and headers into the appropriate locations for use with `clang`.
|
Ping |
ldionne
requested changes
Jul 31, 2024
libcxx/CMakeLists.txt
Outdated
| @@ -491,6 +491,15 @@ include(HandleLibcxxFlags) | |||
| # 'config-ix' use them during feature checks. It also adds them to both | |||
| # 'LIBCXX_COMPILE_FLAGS' and 'LIBCXX_LINK_FLAGS' | |||
|
|
|||
| # Targeting the GPU requires a clang compiler and several extra flags. | |||
There was a problem hiding this comment.
Why does Clang require -nogpulib? Don't we already pass -nodefaultlibs?
Also, why -flto?
We should set this in the CMake caches -- we don't want to add any more target-specific flags in our CMake files, this is too messy.
There was a problem hiding this comment.
-nogpulib is separate, it controls whether or not the GPU toolchain pulls in external vendor libraries.
-flto is 100% required because GPUs have almost no backwards compatibility, so unless you want to build libc++ over 30 times you need to use LLVM-IR + LTO to defer creating object code until it's linked into the user's application. Also ELF linking isn't supported at all for AMDGPU.
I can move this bit to the CMake cache, but CMAKE_REQUIRED_FLAGS might need to stay, otherwise check_cxx_flags will not work.
Summary: The GPU target will always be a sufficiently new `clang`, so we can assume these flags are present. We need to first set `CMAKE_REQUIRED_FLAGS` to these values so that the `check_cxx_compile_flag` utilities work. Then, we need to add several things to the compiler flags that are necessary for correctness and optimal code output.
|
Seems I can do this in the cache like you said, will update that shortly. |
jhuber6
added a commit
to jhuber6/llvm-project
that referenced
this pull request
Jul 31, 2024
Summary: This patch adds a CMake cache config file for the GPU build. This cache will set the default required options when used from the LLVM runtime interface or directly. These options pretty much disable everything the GPU can't handle. With this and the fllowing patches: llvm#99259, llvm#99243, llvm#99287, and llvm#99333, we will be able to build `libc++` targeting the GPU with an invocation like this. ``` $ cmake ../llvm -C${LLVM_ROOT}/libcxx/cmake/caches/GPU.cmake \ -DRUNTIMES_nvptx64-nvidia-cuda_LLVM_ENABLE_RUNTIMES=compiler-rt;libc;libcxx \ -DRUNTIMES_amdgcn-amd-amdhsa_LLVM_ENABLE_RUNTIMES=compiler-rt;libc;libcxx \ -DLLVM_RUNTIME_TARGETS=amdgcn-amd-amdhsa;nvptx64-nvidia-cuda \ ``` This will then install the libraries and headers into the appropriate locations for use with `clang`.
jhuber6
added a commit
to jhuber6/llvm-project
that referenced
this pull request
Aug 2, 2024
Summary: This patch adds a CMake cache config file for the GPU build. This cache will set the default required options when used from the LLVM runtime interface or directly. These options pretty much disable everything the GPU can't handle. With this and the fllowing patches: llvm#99259, llvm#99243, llvm#99287, and llvm#99333, we will be able to build `libc++` targeting the GPU with an invocation like this. ``` $ cmake ../llvm -C${LLVM_ROOT}/libcxx/cmake/caches/GPU.cmake \ -DRUNTIMES_nvptx64-nvidia-cuda_LLVM_ENABLE_RUNTIMES=compiler-rt;libc;libcxx \ -DRUNTIMES_amdgcn-amd-amdhsa_LLVM_ENABLE_RUNTIMES=compiler-rt;libc;libcxx \ -DLLVM_RUNTIME_TARGETS=amdgcn-amd-amdhsa;nvptx64-nvidia-cuda \ ``` This will then install the libraries and headers into the appropriate locations for use with `clang`.
jhuber6
added a commit
to jhuber6/llvm-project
that referenced
this pull request
Aug 14, 2024
Summary: This patch adds a CMake cache config file for the GPU build. This cache will set the default required options when used from the LLVM runtime interface or directly. These options pretty much disable everything the GPU can't handle. With this and the fllowing patches: llvm#99259, llvm#99243, llvm#99287, and llvm#99333, we will be able to build `libc++` targeting the GPU with an invocation like this. ``` $ cmake ../llvm -C${LLVM_ROOT}/libcxx/cmake/caches/GPU.cmake \ -DRUNTIMES_nvptx64-nvidia-cuda_LLVM_ENABLE_RUNTIMES=compiler-rt;libc;libcxx \ -DRUNTIMES_amdgcn-amd-amdhsa_LLVM_ENABLE_RUNTIMES=compiler-rt;libc;libcxx \ -DLLVM_RUNTIME_TARGETS=amdgcn-amd-amdhsa;nvptx64-nvidia-cuda \ ``` This will then install the libraries and headers into the appropriate locations for use with `clang`.
jhuber6
added a commit
to jhuber6/llvm-project
that referenced
this pull request
Aug 15, 2024
Summary: This patch adds a CMake cache config file for the GPU build. This cache will set the default required options when used from the LLVM runtime interface or directly. These options pretty much disable everything the GPU can't handle. With this and the fllowing patches: llvm#99259, llvm#99243, llvm#99287, and llvm#99333, we will be able to build `libc++` targeting the GPU with an invocation like this. ``` $ cmake ../llvm -C${LLVM_ROOT}/libcxx/cmake/caches/GPU.cmake \ -DRUNTIMES_nvptx64-nvidia-cuda_LLVM_ENABLE_RUNTIMES=compiler-rt;libc;libcxx \ -DRUNTIMES_amdgcn-amd-amdhsa_LLVM_ENABLE_RUNTIMES=compiler-rt;libc;libcxx \ -DLLVM_RUNTIME_TARGETS=amdgcn-amd-amdhsa;nvptx64-nvidia-cuda \ ``` This will then install the libraries and headers into the appropriate locations for use with `clang`. Move to separate files
jhuber6
added a commit
to jhuber6/llvm-project
that referenced
this pull request
Aug 21, 2024
Summary: This patch adds a CMake cache config file for the GPU build. This cache will set the default required options when used from the LLVM runtime interface or directly. These options pretty much disable everything the GPU can't handle. With this and the fllowing patches: llvm#99259, llvm#99243, llvm#99287, and llvm#99333, we will be able to build `libc++` targeting the GPU with an invocation like this. ``` $ cmake ../llvm -C${LLVM_ROOT}/libcxx/cmake/caches/GPU.cmake \ -DRUNTIMES_nvptx64-nvidia-cuda_LLVM_ENABLE_RUNTIMES=compiler-rt;libc;libcxx \ -DRUNTIMES_amdgcn-amd-amdhsa_LLVM_ENABLE_RUNTIMES=compiler-rt;libc;libcxx \ -DLLVM_RUNTIME_TARGETS=amdgcn-amd-amdhsa;nvptx64-nvidia-cuda \ ``` This will then install the libraries and headers into the appropriate locations for use with `clang`.
jhuber6
added a commit
that referenced
this pull request
Aug 21, 2024
Summary: This patch adds a CMake cache config file for the GPU build. This cache will set the default required options when used from the LLVM runtime interface or directly. These options pretty much disable everything the GPU can't handle. With this and the following patches: #99259, #99243, #99287, and #99333, we will be able to build `libc++` targeting the GPU with an invocation like this. ``` $ cmake ../llvm -DRUNTIMES_nvptx64-nvidia-cuda_CACHE_FILES=${LLVM_SRC}/../libcxx/cmake/caches/NVPTX.cmake \ -DRUNTIMES_amdgcn-amd-amdhsa_CACHE_FILES=${LLVM_SRC}/../libcxx/cmake/caches/AMDGPU.cmake \ -DRUNTIMES_nvptx64-nvidia-cuda_LLVM_ENABLE_RUNTIMES=compiler-rt;libc;libcxx \ -DRUNTIMES_amdgcn-amd-amdhsa_LLVM_ENABLE_RUNTIMES=compiler-rt;libc;libcxx \ -DLLVM_RUNTIME_TARGETS=amdgcn-amd-amdhsa;nvptx64-nvidia-cuda \ ``` This will then install the libraries and headers into the appropriate locations for use with `clang`.
cjdb
pushed a commit
to cjdb/llvm-project
that referenced
this pull request
Aug 23, 2024
Summary: This patch adds a CMake cache config file for the GPU build. This cache will set the default required options when used from the LLVM runtime interface or directly. These options pretty much disable everything the GPU can't handle. With this and the following patches: llvm#99259, llvm#99243, llvm#99287, and llvm#99333, we will be able to build `libc++` targeting the GPU with an invocation like this. ``` $ cmake ../llvm -DRUNTIMES_nvptx64-nvidia-cuda_CACHE_FILES=${LLVM_SRC}/../libcxx/cmake/caches/NVPTX.cmake \ -DRUNTIMES_amdgcn-amd-amdhsa_CACHE_FILES=${LLVM_SRC}/../libcxx/cmake/caches/AMDGPU.cmake \ -DRUNTIMES_nvptx64-nvidia-cuda_LLVM_ENABLE_RUNTIMES=compiler-rt;libc;libcxx \ -DRUNTIMES_amdgcn-amd-amdhsa_LLVM_ENABLE_RUNTIMES=compiler-rt;libc;libcxx \ -DLLVM_RUNTIME_TARGETS=amdgcn-amd-amdhsa;nvptx64-nvidia-cuda \ ``` This will then install the libraries and headers into the appropriate locations for use with `clang`.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary:
The GPU target will always be a sufficiently new
clang, so we canassume these flags are present. We need to first set
CMAKE_REQUIRED_FLAGSto these values so that thecheck_cxx_compile_flagutilities work. Then, we need to add severalthings to the compiler flags that are necessary for correctness and
optimal code output.