[SYCL][CUDA] Fix 3D buffer D2D copy slice pitch #2046
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
romanovvlad
requested changes
Jul 8, 2020
| SrcSlicePitch = (3 == DimSrc) ? SrcSize[0] * SrcSize[1] : 0; | ||
| DstSlicePitch = (3 == DimDst) ? DstSize[0] * DstSize[1] : 0; | ||
| } else { | ||
| SrcSlicePitch = (DimSrc > 1) ? SrcSize[0] * SrcSize[1] : SrcSize[0]; |
There was a problem hiding this comment.
Since it is CUDA specific, can't we move it inside CUDA plugin?
There was a problem hiding this comment.
The problem is that it is not possible to determine the full size of the source and destination volumes from what is currently passed to PI API (CUDA requires this, but OpenCL seemingly does not).
I've updated this now so that the calculation is the same for all plugins - this doesn't break the OpenCL backend, and fixes the CUDA backend.
przemektmalon
force-pushed
the
przemek/cuda-fix-3d-d2d-buffer-copy-slice-pitch
branch
2 times, most recently
from
df8083c to
b92cc67
Compare
This commit fixes the slice pitch parameters for the 2D/3D buffer device to device copy operation. This fixes a LIT test (`buffer_full_copy`) that reported passing on CUDA despite throwing a CUDA error. This LIT test is amended to check the validity of 3D buffer copy operations. The CUDA error occured due to the `srcHeight` member of the CUDA_MEMCPY3D struct not reflecting the true extent of the source volume, but the copied region instead. Signed-off-by: Przemek Malon <[email protected]>
przemektmalon
force-pushed
the
przemek/cuda-fix-3d-d2d-buffer-copy-slice-pitch
branch
from
b92cc67 to
f6575a1
Compare
romanovvlad
approved these changes
Jul 15, 2020
przemektmalon
deleted the
przemek/cuda-fix-3d-d2d-buffer-copy-slice-pitch
branch
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This commit fixes the slice pitch parameters
for the 2D/3D buffer device to device
copy operation. This fixes a LIT test
(
buffer_full_copy) that reported passing onCUDA despite throwing a CUDA error. This LIT test
is also fixed so that it can no longer silently fail
upon thrown exceptions.
The CUDA error occured due to the
srcHeightmember of the CUDA_MEMCPY3D struct not reflecting
the true extent of the source volume, but the copied
region instead.
Signed-off-by: Przemek Malon [email protected]