https://github.com/wenju-he created
https://github.com/llvm/llvm-project/pull/153137
According to OpenCL spec, native_* functions have implementation-defined
accuracy and typically have better performance. We can enable floating- point
contraction optimizations for them.
>From 719a6914321afc04b44fda1a7ba38ecffa95d080 Mon Sep 17 00:00:00 2001
From: Wenju He <wenju...@intel.com>
Date: Tue, 12 Aug 2025 06:27:11 +0200
Subject: [PATCH] [libclc] Enable -ffp-contract=fast compile option for math
native_* functions
According to OpenCL spec, native_* functions have implementation-defined
accuracy and typically have better performance. We can enable floating-
point contraction optimizations for them.
---
libclc/CMakeLists.txt | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/libclc/CMakeLists.txt b/libclc/CMakeLists.txt
index 756e097dcf123..55072785e6f79 100644
--- a/libclc/CMakeLists.txt
+++ b/libclc/CMakeLists.txt
@@ -304,7 +304,7 @@ set_source_files_properties(
${CMAKE_CURRENT_SOURCE_DIR}/opencl/lib/generic/math/native_sin.cl
${CMAKE_CURRENT_SOURCE_DIR}/opencl/lib/generic/math/native_sqrt.cl
${CMAKE_CURRENT_SOURCE_DIR}/opencl/lib/generic/math/native_tan.cl
- PROPERTIES COMPILE_OPTIONS -fapprox-func
+ PROPERTIES COMPILE_OPTIONS "-fapprox-func;-ffp-contract=fast"
)
enable_testing()
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
