Changes

Jump to: navigation, search

K2

88 bytes removed, 20:17, 16 April 2018
Assignment3
==Assignment3==
While working on the project, we discovered that cudaMemcpy(HtoD) and (DtoH) takes long time.​
 
so, we decided to use pinned memory instead of pageable memory​ cudaHostAlloc();​
"cudaHostAlloc() will allocate pinned memory which is always stored in RAM and can be accessed by GPU's DMA(direct memory access) directly"
"cudaHostAlloc() will allocate pinned memory which is always stored in RAM and can be accessed by GPU's DMA(direct memory access) directly"
[[File:Resultone.png|500px|thumb|left|pinned memory vs pageable memory]]
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
using pinned memory is 2 times faster than pageable memory
44
edits

Navigation menu