vllm/csrc/cuda_utils.h

#pragma once

#include <stdio.h>

#if defined(__HIPCC__)
  #define HOST_DEVICE_INLINE __host__ __device__
  #define DEVICE_INLINE __device__
  #define HOST_INLINE __host__
#elif defined(__CUDACC__) || defined(_NVHPC_CUDA)
  #define HOST_DEVICE_INLINE __host__ __device__ __forceinline__
  #define DEVICE_INLINE __device__ __forceinline__
  #define HOST_INLINE __host__ __forceinline__
#else
  #define HOST_DEVICE_INLINE inline
  #define DEVICE_INLINE inline
  #define HOST_INLINE inline
#endif

#define CUDA_CHECK(cmd)                                             \
  do {                                                              \
    cudaError_t e = cmd;                                            \
    if (e != cudaSuccess) {                                         \
      printf("Failed: Cuda error %s:%d '%s'\n", __FILE__, __LINE__, \
             cudaGetErrorString(e));                                \
      exit(EXIT_FAILURE);                                           \
    }                                                               \
  } while (0)

int64_t get_device_attribute(int64_t attribute, int64_t device_id);

int64_t get_max_shared_memory_per_block_device_attribute(int64_t device_id);

namespace cuda_utils {

template <typename T>
HOST_DEVICE_INLINE constexpr std::enable_if_t<std::is_integral_v<T>, T>
ceil_div(T a, T b) {
  return (a + b - 1) / b;
}

};  // namespace cuda_utils