cuda - Appropriate usage of cudaGetSymbolAddress and cudaMemcpyToSymbol with global memory? -
cuda - Appropriate usage of cudaGetSymbolAddress and cudaMemcpyToSymbol with global memory? -
i new cuda , familiar normal usage of cudamalloc , cudamemcpy , cudamemcpytosymbol copying constant memory.
however, have been given code makes frequent utilize of cudagetsymboladdress , cudamemcpytosymbol re-create global memory , i'm not sure why have chosen instead of cudamalloc/cudamemcpy.
would able explain when advantageous , appropriate utilize cudagetsymboladdress , cudamemcpytosymbol?
thank you!
when global memory allocated dynamically using cudamalloc, right copying api utilize cudamemcpy.
when global memory allocated statically:
__device__ int my_data[dsize]; then right api utilize cudamemcpytosymbol or cudamemcpyfromsymbol
cuda
Comments
Post a Comment