Jimerlife added inline comments.
================ Comment at: clang/include/clang/Basic/BuiltinsRISCV.def:36 +// zbf extension +TARGET_BUILTIN(__builtin_riscv_bfp, "LiLiLi", "nc", "experimental-zbf") ---------------- craig.topper wrote: > Capital Z done ================ Comment at: clang/include/clang/Basic/BuiltinsRISCV.def:37 +// zbf extension +TARGET_BUILTIN(__builtin_riscv_bfp, "LiLiLi", "nc", "experimental-zbf") +TARGET_BUILTIN(__builtin_riscv_bfpw, "WiWiWi", "nc", "experimental-zbf,64bit") ---------------- craig.topper wrote: > I think we should have `__builtin_riscv_bfp_32` and `__builtin_riscv_bfp_64`. > It's more convenient and portable for software to be written in terms of > number of bits being operated on rather than changing behavior based on xlen. I have rewrite two intrinsics according to target machine ================ Comment at: llvm/include/llvm/IR/IntrinsicsRISCV.td:98 + def int_riscv_bfp : BitManipGPRGPRIntrinsics; + def int_riscv_bfpw : BitManipGPRGPRIntrinsics; + ---------------- craig.topper wrote: > We only need one intrinsic. `BitManipGPRGPRIntrinsics` is type overloaded. We > can check the type in isel patterns. Only save one intrinsic pattern. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D116994/new/ https://reviews.llvm.org/D116994 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
