diff --git a/tinygrad/runtime/support/compiler_amd.py b/tinygrad/runtime/support/compiler_amd.py index 3331e20631..91052aa7fa 100644 --- a/tinygrad/runtime/support/compiler_amd.py +++ b/tinygrad/runtime/support/compiler_amd.py @@ -103,7 +103,7 @@ class HIPCCCompiler(Compiler): subprocess.run(["hipcc", "-c", "-emit-llvm", "--cuda-device-only", "-O3", "-mcumode", f"--offload-arch={self.arch}", "-I/opt/rocm/include/hip", "-o", bcf.name, srcf.name] + self.extra_options, check=True) subprocess.run(["hipcc", "-target", "amdgcn-amd-amdhsa", f"-mcpu={self.arch}", - "-O3", "-mllvm", "-amdgpu-internalize-symbols", "-c", "-o", libf.name, bcf.name], check=True) + "-O3", "-mllvm", "-amdgpu-internalize-symbols", "-c", "-o", libf.name, bcf.name] + self.extra_options, check=True) return pathlib.Path(libf.name).read_bytes() def disassemble(self, lib:bytes): amdgpu_disassemble(lib)