Details
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
clang/include/clang/Basic/BuiltinsAMDGPU.def | ||
---|---|---|
221–224 | The intrinsic signature suggests the 1st and 4th/5th arguments are overloadable. How does this handle the various supported types? |
clang/include/clang/Basic/BuiltinsAMDGPU.def | ||
---|---|---|
221–224 | By convention, we do not define overloaded target specific clang builtins. In stead we use postfixes to indicate the types. In this case, we use the following naming convention: h - 4/5-th args are half by default - 1st arg is i32, 4/5-th args are float. Currently we are missing functions for 4/5-th args are double. I will add them. |
clang/include/clang/Basic/BuiltinsAMDGPU.def | ||
---|---|---|
221–224 | It seems we do not support 4/50th args to be double. So the current definitions are all that we support. |
The intrinsic signature suggests the 1st and 4th/5th arguments are overloadable. How does this handle the various supported types?