fix: include cuda_bf16.h unconditionally, add --expt-relaxed-constexpr

dsv4/kernels/attention/fmha_sm100.cuh

@@ -30,16 +30,8 @@
 #pragma once
 
 #include <cuda_runtime.h>
-#include <cstdint>
-
-// NOTE: cuda_bf16.h has a C++17 compatibility bug on CUDA 13.2.
-// We define a minimal BF16 type using the __bf16 built-in (CUDA 13+).
-// __bf16 is a built-in storage type; we wrap it for type safety.
-#if defined(__CUDA_ARCH__)
-// Device code: __nv_bfloat16 IS available via the built-in
-#undef __BF16_COMPAT
 #include <cuda_bf16.h>
-#endif
+#include <cstdint>
 
 // CUTLASS C++ includes (CUDA device code only)
 #if defined(__CUDA_ARCH__)

tests/unit/test_fmha_sm100.py

@@ -42,6 +42,7 @@ nvcc_cmd = [
     "--x", "cu",
     "-o", "/tmp/fmha_sm100_test.o",
     "--ptxas-options=-v",
+    "--expt-relaxed-constexpr",
 ]
 
 print(f"nvcc command: {' '.join(nvcc_cmd)}")