[libc-commits] [libc] [libc][math] Optimize nearest integer functions using builtins when available (PR #98376)

[via libc-commits]

================
@@ -39,11 +47,19 @@ endfunction()
 set(AVAILABLE_COMPILER_FEATURES "")
 
 # Try compile a C file to check if flag is supported.
-set(CMAKE_TRY_COMPILE_TARGET_TYPE STATIC_LIBRARY)
 foreach(feature IN LISTS ALL_COMPILER_FEATURES)
+  set(CMAKE_TRY_COMPILE_TARGET_TYPE STATIC_LIBRARY)
   set(compile_options ${LIBC_COMPILE_OPTIONS_NATIVE})
   if(${feature} STREQUAL "fixed_point")
     list(APPEND compile_options "-ffixed-point")
+  elseif(${feature} MATCHES "^builtin_")
+    set(compile_options ${LIBC_COMPILE_OPTIONS_DEFAULT})
+    # The compiler might handle calls to rounding builtins by generating calls
+    # to the respective libc math functions, in which case we cannot use these
+    # builtins in our implementations of these functions. We check that this is
+    # not the case by trying to link an executable, since linking would fail due
+    # to unresolved references if calls to libc functions were generated.
+    set(CMAKE_TRY_COMPILE_TARGET_TYPE EXECUTABLE)
----------------
overmighty wrote:

This actually doesn't work. Here, `compile_options` doesn't have `-ffreestanding` or anything like it, and the executable just gets linked with the system libc, so linking would only fail if the compiler generated a call to a function that's not implemented by the system libc. Also, I think this CMake script runs too early on to link the feature check executables with our own startup code and use `-ffreestanding`, so I'm not sure what to do.

https://github.com/llvm/llvm-project/pull/98376