[libc] Compile the GPU functions with '-fconvergent-functions' #70229
Conversation
Summary: This patch simply adds the `-fconvergent-functions` flag to the GPU compilation. This is in relation to the behaviour of SIMT architectures under divergence. With the flag, we assume every function is convergent by default and rely on the compiler's divergence analysis to transform it if possible. Fixes: llvm#63853
|
@llvm/pr-subscribers-libc Author: Joseph Huber (jhuber6) ChangesSummary: Fixes: #63853 Full diff: https://github.com/llvm/llvm-project/pull/70229.diff 1 Files Affected:
diff --git a/libc/cmake/modules/LLVMLibCObjectRules.cmake b/libc/cmake/modules/LLVMLibCObjectRules.cmake
index 7caa3b2db75dd2f..cf5097e23a20453 100644
--- a/libc/cmake/modules/LLVMLibCObjectRules.cmake
+++ b/libc/cmake/modules/LLVMLibCObjectRules.cmake
@@ -68,6 +68,7 @@ function(_get_common_compile_options output_var flags)
if (LIBC_TARGET_ARCHITECTURE_IS_GPU)
list(APPEND compile_options "-nogpulib")
list(APPEND compile_options "-fvisibility=hidden")
+ list(APPEND compile_options "-fconvergent-functions")
# Manually disable all standard include paths and include the resource
# directory to prevent system headers from being included.
|
|
Honestly, we should set this flag by default for GPU targets. Other than that, this seems sensible to me. |
Most likely, but this was easier than getting |
…70229) Summary: This patch simply adds the `-fconvergent-functions` flag to the GPU compilation. This is in relation to the behaviour of SIMT architectures under divergence. With the flag, we assume every function is convergent by default and rely on the compiler's divergence analysis to transform it if possible. Fixes: llvm#63853
Summary:
This patch simply adds the
-fconvergent-functionsflag to the GPUcompilation. This is in relation to the behaviour of SIMT
architectures under divergence. With the flag, we assume every function
is convergent by default and rely on the compiler's divergence analysis
to transform it if possible.
Fixes: #63853