Hi, Is memcpy from pinned host memory to GPU memory inside a kernel performed by a DMA engine? So that the accordingly SM is free to switch to another warp and perform some arithmetic operations.

[image] huaouo: Is memcpy from pinned host memory to GPU memory inside a kernel performed by a DMA engine? No, it is not. (Just study PTX or SASS to convince yourself.) Those will create LD and ST instructions, and are performed by the SM/LSU/Memory controller, which basically issues PCIE c…

Is memcpy from pinned host memory to GPU memory inside a kernel performed by a DMA engine?

Accelerated Computing CUDA CUDA Programming and Performance

system Closed November 25, 2021, 4:25pm 4

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.

Topic		Replies	Views
DMA CUDA Programming and Performance	4	8249	October 8, 2009
Issuing copies to host from within device code CUDA Programming and Performance	4	456	October 15, 2020
transfer from pageable host memory to page-locked host memory? CUDA Programming and Performance	3	1148	June 1, 2012
Overlapping CPU<->GPU trasnfer and kernel computation only for pinned memory CUDA Programming and Performance	3	970	March 29, 2011
selfmade cudeMallocHost()? CUDA Programming and Performance	9	8761	February 14, 2008
DMA Transfer between Third party device, host, and GPU CUDA Programming and Performance hw , cuda	0	715	June 8, 2020
Is cudaMemcpyAsync inside a kernel controlled by the GPU? CUDA Programming and Performance	9	3676	July 28, 2019
Is it possible to use pinned memory? Outside of CUDA CUDA Programming and Performance	14	6582	January 22, 2025
why using pinned memory is faster? CUDA Programming and Performance	3	2948	November 30, 2007
cudaHostAlloc and memcpy CUDA Programming and Performance	1	1172	June 29, 2015

Is memcpy from pinned host memory to GPU memory inside a kernel performed by a DMA engine?

Related topics