[libc-commits] [libc] [libc] Fix standard cross build targeting the GPU (PR #82724)

[Joseph Huber via libc-commits]

https://github.com/jhuber6 created https://github.com/llvm/llvm-project/pull/82724

Summary:
The GPU target has recently been changed to support standard `libc`
build rules. This means we should be able to build for it both in
`LLVM_ENABLE_PROJECTS` mode, or targeting the runtimes directory
directly as in the LLVM `libc` documentation. Previously this failed
because the version check on the compiler was too strict and the
`--target=` options were not being set on the link jobs unless in CMake
cross compiliation mode. This patch fixes those so the following config
should work now to build the GPU target directly if using NVPTX.

```
cmake ../runtimes -DCMAKE_BUILD_TYPE=Release \
  -DCMAKE_CXX_COMPILER=clang++ -DCMAKE_C_COMPILER=clang \
  -DLLVM_ENABLE_RUNTIMES=libc -DLLVM_RUNTIMES_TARGET=nvptx64-nvidia-cuda \
  -DLLVM_DEFAULT_TARGET_TRIPLE=nvptx64-nvidia-cuda \
  -DLIBC_HDRGEN_EXE=/path/to/hdrgen/libc-hdrgen \
  -DLLVM_LIBC_FULL_BUILD=ON -GNinja
```


>From 758bc3c578e3027eb15664641df64a9010a5d7d9 Mon Sep 17 00:00:00 2001
From: Joseph Huber <huberjn at outlook.com>
Date: Thu, 22 Feb 2024 20:07:15 -0600
Subject: [PATCH] [libc] Fix standard cross build targeting the GPU

Summary:
The GPU target has recently been changed to support standard `libc`
build rules. This means we should be able to build for it both in
`LLVM_ENABLE_PROJECTS` mode, or targeting the runtimes directory
directly as in the LLVM `libc` documentation. Previously this failed
because the version check on the compiler was too strict and the
`--target=` options were not being set on the link jobs unless in CMake
cross compiliation mode. This patch fixes those so the following config
should work now to build the GPU target directly if using NVPTX.

```
cmake ../runtimes -DCMAKE_BUILD_TYPE=Release \
  -DCMAKE_CXX_COMPILER=clang++ -DCMAKE_C_COMPILER=clang \
  -DLLVM_ENABLE_RUNTIMES=libc -DLLVM_RUNTIMES_TARGET=nvptx64-nvidia-cuda \
  -DLLVM_DEFAULT_TARGET_TRIPLE=nvptx64-nvidia-cuda \
  -DLIBC_HDRGEN_EXE=/path/to/hdrgen/libc-hdrgen \
  -DLLVM_LIBC_FULL_BUILD=ON -GNinja
```
---
 libc/cmake/modules/LLVMLibCTestRules.cmake      | 4 ++++
 libc/cmake/modules/prepare_libc_gpu_build.cmake | 4 ++--
 2 files changed, 6 insertions(+), 2 deletions(-)

diff --git a/libc/cmake/modules/LLVMLibCTestRules.cmake b/libc/cmake/modules/LLVMLibCTestRules.cmake
index 1166c26e4d8a56..9d1e426a5b7b31 100644
--- a/libc/cmake/modules/LLVMLibCTestRules.cmake
+++ b/libc/cmake/modules/LLVMLibCTestRules.cmake
@@ -463,6 +463,7 @@ function(add_integration_test test_name)
 
   if(LIBC_TARGET_ARCHITECTURE_IS_AMDGPU)
     target_link_options(${fq_build_target_name} PRIVATE 
+      ${LIBC_COMPILE_OPTIONS_DEFAULT}
       -mcpu=${LIBC_GPU_TARGET_ARCHITECTURE} -flto
       "-Wl,-mllvm,-amdgpu-lower-global-ctor-dtor=0" -nostdlib -static
       "-Wl,-mllvm,-amdhsa-code-object-version=${LIBC_GPU_CODE_OBJECT_VERSION}")
@@ -470,6 +471,7 @@ function(add_integration_test test_name)
     # We need to use the internal object versions for NVPTX.
     set(internal_suffix ".__internal__")
     target_link_options(${fq_build_target_name} PRIVATE 
+      ${LIBC_COMPILE_OPTIONS_DEFAULT}
       "-Wl,--suppress-stack-size-warning"
       -march=${LIBC_GPU_TARGET_ARCHITECTURE} -nostdlib -static
       "--cuda-path=${LIBC_CUDA_ROOT}")
@@ -644,6 +646,7 @@ function(add_libc_hermetic_test test_name)
 
   if(LIBC_TARGET_ARCHITECTURE_IS_AMDGPU)
     target_link_options(${fq_build_target_name} PRIVATE 
+      ${LIBC_COMPILE_OPTIONS_DEFAULT}
       -mcpu=${LIBC_GPU_TARGET_ARCHITECTURE} -flto
       "-Wl,-mllvm,-amdgpu-lower-global-ctor-dtor=0" -nostdlib -static
       "-Wl,-mllvm,-amdhsa-code-object-version=${LIBC_GPU_CODE_OBJECT_VERSION}")
@@ -651,6 +654,7 @@ function(add_libc_hermetic_test test_name)
     # We need to use the internal object versions for NVPTX.
     set(internal_suffix ".__internal__")
     target_link_options(${fq_build_target_name} PRIVATE 
+      ${LIBC_COMPILE_OPTIONS_DEFAULT}
       "-Wl,--suppress-stack-size-warning"
       -march=${LIBC_GPU_TARGET_ARCHITECTURE} -nostdlib -static
       "--cuda-path=${LIBC_CUDA_ROOT}")
diff --git a/libc/cmake/modules/prepare_libc_gpu_build.cmake b/libc/cmake/modules/prepare_libc_gpu_build.cmake
index 75beef86760c8d..8e5d1bb93d77cf 100644
--- a/libc/cmake/modules/prepare_libc_gpu_build.cmake
+++ b/libc/cmake/modules/prepare_libc_gpu_build.cmake
@@ -5,8 +5,8 @@ endif()
 
 # Ensure the compiler is a valid clang when building the GPU target.
 set(req_ver "${LLVM_VERSION_MAJOR}.${LLVM_VERSION_MINOR}.${LLVM_VERSION_PATCH}")
-if(NOT (CMAKE_CXX_COMPILER_ID MATCHES "[Cc]lang" AND
-        ${CMAKE_CXX_COMPILER_VERSION} VERSION_EQUAL "${req_ver}"))
+if(LLVM_VERSION_MAJOR AND NOT (CMAKE_CXX_COMPILER_ID MATCHES "[Cc]lang" AND
+   ${CMAKE_CXX_COMPILER_VERSION} VERSION_EQUAL "${req_ver}"))
   message(FATAL_ERROR "Cannot build libc for GPU. CMake compiler "
                       "'${CMAKE_CXX_COMPILER_ID} ${CMAKE_CXX_COMPILER_VERSION}' "
                       " is not 'Clang ${req_ver}'.")