Prefetch in cuda (through C code)

Question

I am working on data prefetch in CUDA (Fermi GPU) through C code. Cuda reference manual talks about the prefetching at ptx level code not at C level code.

Can anyon

Answer 1

According to PTX manual here is how prefetch works in PTX:

You can embed the PTX instructions into the CUDA kernel. Here is a tiny sample from NVIDIA's documentation:

__device__ int cube (int x)
{
  int y;
  asm("{\n\t"                       // use braces for local scope
      " .reg .u32 t1;\n\t"           // temp reg t1,
      " mul.lo.u32 t1, %1, %1;\n\t" // t1 = x * x
      " mul.lo.u32 %0, t1, %1;\n\t" // y = t1 * x
      "}"
      : "=r"(y) : "r" (x));
  return y;
}

You may come to conclude with the following prefetch function in C:

__device__ void prefetch_l1 (unsigned int addr)
{

  asm(" prefetch.global.L1 [ %1 ];": "=r"(addr) : "r"(addr));
}

NOTICE: You need the GPU of Compute Capability 2.0 or higher for prefetch. Pass proper compile flags accordingly -arch=sm_20

Answer 2

According to this thread, below is the code for different cache prefetching techniques:

#define DEVICE_STATIC_INTRINSIC_QUALIFIERS  static __device__ __forceinline__

#if (defined(_MSC_VER) && defined(_WIN64)) || defined(__LP64__)
#define PXL_GLOBAL_PTR   "l"
#else
#define PXL_GLOBAL_PTR   "r"
#endif

DEVICE_STATIC_INTRINSIC_QUALIFIERS void __prefetch_global_l1(const void* const ptr)
{
  asm("prefetch.global.L1 [%0];" : : PXL_GLOBAL_PTR(ptr));
}

DEVICE_STATIC_INTRINSIC_QUALIFIERS void __prefetch_global_uniform(const void* const ptr)
{
  asm("prefetchu.L1 [%0];" : : PXL_GLOBAL_PTR(ptr));
}

DEVICE_STATIC_INTRINSIC_QUALIFIERS void __prefetch_global_l2(const void* const ptr)
{
  asm("prefetch.global.L2 [%0];" : : PXL_GLOBAL_PTR(ptr));
}