diff --git a/include/cutlass/uint128.h b/include/cutlass/uint128.h index 0a41e95c..096d90f1 100644 --- a/include/cutlass/uint128.h +++ b/include/cutlass/uint128.h @@ -194,7 +194,7 @@ struct alignas(16) uint128_t uint64_t remainder{0}; #if defined(CUTLASS_UINT128_NATIVE) remainder = uint64_t(native % divisor); -#elif defined(CUTLASS_INT128_ARITHMETIC_DIV) +#elif defined(CUTLASS_INT128_ARITHMETIC_DIV) && ! defined (__CUDA_ARCH__) // implemented using MSVC's arithmetic intrinsics (void)_udiv128(hilo_.hi, hilo_.lo, divisor, &remainder); #else