__global__
void __launch_bounds__(256) fn_transpose_scale_divide_yield_store___kernel(
  const float* __restrict__ var /* 32,128,14,14 */,
  float* __restrict__ var_3 /* 32,14,14,128 */
) {
  __builtin_assume(blockIdx.x < 896);
  __builtin_assume(threadIdx.x < 32);
  __builtin_assume(threadIdx.y < 8);
  float _var_local_buffer [ 4 ];
  extern __shared__ uint8_t dyn_shared_buffer[];
  uint8_t *transpose_union_shm = (uint8_t*)&dyn_shared_buffer[ 0 ];
  float* var_shared_buffer = (float*)transpose_union_shm;
  float* var_local_buffer = _var_local_buffer;
  __syncthreads();
  for (int k = 0; k < 4; k += 1) {
    if (((((blockIdx.x % 28) / 4) * 32) + threadIdx.x) < 196) {
      float var_local = var[((((((k * 8) + threadIdx.y) + ((blockIdx.x & 3) * 32)) + ((blockIdx.x / 28) * 128)) * 196) + threadIdx.x) + (((blockIdx.x % 28) / 4) * 32)];
      var_shared_buffer[(((k * 8) + threadIdx.y) * 32) + cinn_nvgpu_bitwise_xor_int32(threadIdx.x, ((k * 8) + threadIdx.y))] = var_local;
    }
  }
  __syncthreads();
  for (int k = 0; k < 4; k += 1) {
    if (((((((blockIdx.x % 28) / 4) * 4) + k) * 8) + threadIdx.y) < 196) {
      var_local_buffer[k] = var_shared_buffer[(threadIdx.x * 32) + cinn_nvgpu_bitwise_xor_int32(((k * 8) + threadIdx.y), threadIdx.x)];
    }
  }
  for (int k = 0; k < 4; k += 1) {
    if (((((((blockIdx.x % 28) / 4) * 4) + k) * 8) + threadIdx.y) < 196) {
      var_3[((((((((blockIdx.x % 28) / 4) * 32) + (k * 8)) + threadIdx.y) + ((blockIdx.x / 28) * 196)) * 128) + ((blockIdx.x & 3) * 32)) + threadIdx.x] =
          (var_local_buffer[k] / (var_local_buffer[k] + 1.00000000f));
    }
  }
}