Error(Segmentation fault) while using cudaHostAlloc ,Does parameter size require size?

Accelerated Computing CUDA CUDA Programming and Performance
1 
inline bool cudaAllocMapped( void** cpuPtr, void** gpuPtr, size_t size )
{
	if( !cpuPtr || !gpuPtr || size == 0 )
		return false;

	//CUDA(cudaSetDeviceFlags(cudaDeviceMapHost));

	if( CUDA_FAILED(cudaHostAlloc(cpuPtr, size, cudaHostAllocMapped)) )
		return false;

	if( CUDA_FAILED(cudaHostGetDevicePointer(gpuPtr, *cpuPtr, 0)) )
		return false;

	memset(*cpuPtr, 0, size);
	printf("[cuda]  cudaAllocMapped %zu bytes, CPU %p GPU %p\n", size, *cpuPtr, *gpuPtr);
	return true;
}

in the project i want to get NV12 and malloc memory， 5125121.5 can success，2562561.5can success. but when i use 3003001.5 which for SSD network, Segmentation fault arise on cudaHostAlloc

2

Your code works fine for me. My guess would be that the way you are calling it (the variables you are passing) is the problem. If you are passing a bare ** pointer, that won’t work, because it has no allocation for the thing it is pointing to.

$ cat t481.cu
#define CUDA_FAILED(x) x
#include <stdio.h>

inline bool cudaAllocMapped( void** cpuPtr, void** gpuPtr, size_t size )
{
        if( !cpuPtr || !gpuPtr || size == 0 )
                return false;

        //CUDA(cudaSetDeviceFlags(cudaDeviceMapHost));

        if( CUDA_FAILED(cudaHostAlloc(cpuPtr, size, cudaHostAllocMapped)) )
                return false;

        if( CUDA_FAILED(cudaHostGetDevicePointer(gpuPtr, *cpuPtr, 0)) )
                return false;

        memset(*cpuPtr, 0, size);
        printf("[cuda]  cudaAllocMapped %zu bytes, CPU %p GPU %p\n", size, *cpuPtr, *gpuPtr);
        return true;
}
int main(){

  float *data, *d_data;
  const size_t sz = 1000;
  cudaAllocMapped((void**)&data, (void**)&d_data, sz);
  return 0;
}
$ nvcc -o t481 t481.cu
$ cuda-memcheck ./t481
========= CUDA-MEMCHECK
[cuda]  cudaAllocMapped 1000 bytes, CPU 0x200c00000 GPU 0x200c00000
========= ERROR SUMMARY: 0 errors
$