Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[Feature]: re-enable gpu-support #325

Open
kitsudaiki opened this issue May 20, 2024 · 4 comments
Open

[Feature]: re-enable gpu-support #325

kitsudaiki opened this issue May 20, 2024 · 4 comments
Assignees
@kitsudaiki
Labels
C++ L: ★★★ Expect high difficulty New feature / Improvement new features and improvements of existing features Topic: Performance improve any kind of performance Topic: Scalability increase scalability of the network
Milestone
0.7.0

Comments

@kitsudaiki
Copy link
Owner

kitsudaiki commented May 20, 2024 •
edited
Loading

Feature

Description

At the moment the gpu-support is disabled, because there were some changes at the core while there was at the same time no gpu available to test the update in the CUDA-code.

Acceptance Criteria

  • enable gpu-support again
  • update CUDA-kernel and the code for the data-transfer from and to the gpu for the current state of the core

Additional Information

Blocked by

How to test
@kitsudaiki kitsudaiki added New feature / Improvement new features and improvements of existing features Topic: Performance improve any kind of performance L: ★★★ Expect high difficulty Topic: Scalability increase scalability of the network labels May 20, 2024
@kitsudaiki kitsudaiki added this to the 0.5.0 milestone May 20, 2024
@kitsudaiki kitsudaiki self-assigned this May 20, 2024
This was referenced May 20, 2024
Closed
Open
Closed
kitsudaiki added a commit that referenced this issue Jul 13, 2024
@kitsudaiki
Change (#325): update cuda-kernel
715f356 
Updated cuda-kernel quite a lot to the current
state of the core data-structure, but not finihed
the process, because there is an idea for a new
update of the core, which would also affect the
cuda-kernel and to avoid double work, this
incomplete state will remain so far for now.
@kitsudaiki kitsudaiki mentioned this issue Jul 13, 2024
Merged
kitsudaiki added a commit that referenced this issue Jul 13, 2024
@kitsudaiki
Merge pull request #363 from kitsudaiki/feat/update-cuda-kernel
a84a64f 
related issue: #325
@kitsudaiki
Copy link
Owner Author

kitsudaiki commented Jul 13, 2024 •
edited
Loading

Update of the cuda-code so far is incomplete and will be set on hold for a while. I had an idea for a new update of the core-structure and the cuda kernel is not abosult necessary at the moment. The current updates were merged anyway, because there was also some cleanup like the split of the kernel-file.
Will be set to on-hold and planned new for version 0.6.0

@kitsudaiki kitsudaiki moved this to On Hold in OpenHanami Jul 13, 2024
@kitsudaiki kitsudaiki modified the milestones: 0.5.0, 0.6.0 Jul 13, 2024
@kitsudaiki kitsudaiki added the on hold paused issues label Jul 13, 2024
@kitsudaiki
Copy link
Owner Author

blocked by issue #370

@kitsudaiki kitsudaiki modified the milestones: 0.6.0, 0.5.0 Jul 15, 2024
@kitsudaiki kitsudaiki mentioned this issue Jul 15, 2024
Closed
10 tasks
@kitsudaiki kitsudaiki modified the milestones: 0.5.0, backlog core Jul 23, 2024
@kitsudaiki
Copy link
Owner Author

cuda-kernel completely disabled form now in context of issue #377

Need to collect new ideas before I make a new attempt to add gpu support to the project.

@kitsudaiki kitsudaiki added the C++ label Aug 30, 2024
@kitsudaiki kitsudaiki modified the milestones: backlog core, 0.6.0 Sep 21, 2024
@kitsudaiki kitsudaiki removed the on hold paused issues label Sep 21, 2024
@kitsudaiki kitsudaiki moved this from On Hold to In Progress in OpenHanami Sep 21, 2024
@kitsudaiki kitsudaiki mentioned this issue Sep 30, 2024
Closed
kitsudaiki added a commit that referenced this issue Oct 5, 2024
@kitsudaiki
feat(#325): new attempt to enable gpu-support again
b686872 
Update the CUDA-kernel and connected functions in order to
re-enable the gpu-support again. Tests failed so far, because
of a very weird bug, which wasn't found so far. Seems like a
memory-corrumption somewhere, but when the new ASan memory-
check is enabled, CUDA doesn't work in general. To not waste
much more time at the moment for the debugging, the current
changes were added anyway to the develop-branch in order to
try it another time again. Until then the CUDA-support is
hard disabled in the source-code.
@kitsudaiki kitsudaiki mentioned this issue Oct 5, 2024
Merged
kitsudaiki added a commit that referenced this issue Oct 5, 2024
@kitsudaiki
feat(#325): new attempt to enable gpu-support again
9181f92 
Update the CUDA-kernel and connected functions in order to
re-enable the gpu-support again. Tests failed so far, because
of a very weird bug, which wasn't found so far. Seems like a
memory-corrumption somewhere, but when the new ASan memory-
check is enabled, CUDA doesn't work in general. To not waste
much more time at the moment for the debugging, the current
changes were added anyway to the develop-branch in order to
try it another time again. Until then the CUDA-support is
hard disabled in the source-code.
kitsudaiki added a commit that referenced this issue Oct 5, 2024
@kitsudaiki
feat(#325): new attempt for gpu-support (incomplete)
ee272fa 
Update the CUDA-kernel and connected functions in order to
re-enable the gpu-support again. Tests failed so far, because
of a very weird bug, which wasn't found so far. Seems like a
memory-corrumption somewhere, but when the new ASan memory-
check is enabled, CUDA doesn't work in general. To not waste
much more time at the moment for the debugging, the current
changes were added anyway to the develop-branch in order to
try it another time again. Until then the CUDA-support is
hard disabled in the source-code.
kitsudaiki added a commit that referenced this issue Oct 5, 2024
@kitsudaiki
feat(#325): new attempt for gpu-support (incomplete)
a7b298c 
Update the CUDA-kernel and connected functions in order to
re-enable the gpu-support again. Tests failed so far, because
of a very weird bug, which wasn't found so far. Seems like a
memory-corrumption somewhere, but when the new ASan memory-
check is enabled, CUDA doesn't work in general. To not waste
much more time at the moment for the debugging, the current
changes were added anyway to the develop-branch in order to
try it another time again. Until then the CUDA-support is
hard disabled in the source-code.
kitsudaiki added a commit that referenced this issue Oct 5, 2024
@kitsudaiki
feat(#325): new attempt for gpu-support (incomplete)
6692597 
Update the CUDA-kernel and connected functions in order to
re-enable the gpu-support again. Tests failed so far, because
of a very weird bug, which wasn't found so far. Seems like a
memory-corrumption somewhere, but when the new ASan memory-
check is enabled, CUDA doesn't work in general. To not waste
much more time at the moment for the debugging, the current
changes were added anyway to the develop-branch in order to
try it another time again. Until then the CUDA-support is
hard disabled in the source-code.
kitsudaiki added a commit that referenced this issue Oct 5, 2024
@kitsudaiki
feat(#325): new attempt for gpu-support (incomplete)
077cce6 
Update the CUDA-kernel and connected functions in order to
re-enable the gpu-support again. Tests failed so far, because
of a very weird bug, which wasn't found so far. Seems like a
memory-corrumption somewhere, but when the new ASan memory-
check is enabled, CUDA doesn't work in general. To not waste
much more time at the moment for the debugging, the current
changes were added anyway to the develop-branch in order to
try it another time again. Until then the CUDA-support is
hard disabled in the source-code.
kitsudaiki added a commit that referenced this issue Oct 5, 2024
@kitsudaiki
feat(#325): new attempt for gpu-support (incomplete)
73aef7a 
Update the CUDA-kernel and connected functions in order to
re-enable the gpu-support again. Tests failed so far, because
of a very weird bug, which wasn't found so far. Seems like a
memory-corrumption somewhere, but when the new ASan memory-
check is enabled, CUDA doesn't work in general. To not waste
much more time at the moment for the debugging, the current
changes were added anyway to the develop-branch in order to
try it another time again. Until then the CUDA-support is
hard disabled in the source-code.
kitsudaiki added a commit that referenced this issue Oct 5, 2024
@kitsudaiki
Merge pull request #462 from kitsudaiki/feat/re-enable-gpu-support
ff0b73f 
related issue: #325
@kitsudaiki
Copy link
Owner Author

updated code, but while testing I ran into a weird error: After the second cycle the worker-task disappeared from the worker-queue of the cuda-worker-thread, right after the task was plased into the queue. Seems like a memory-corruption, but the recently added ASan memory-check doesn't work with the CUDA-code (doesn't detect a CUDA-gpu when ASan is activated). I don't wanted to waste more time at the moment. Will try at again another time. So the CUDA-kernel is still disabled within the Code and the issue here remains as on-hold

@kitsudaiki kitsudaiki moved this from In Progress to On Hold in OpenHanami Oct 5, 2024
@kitsudaiki kitsudaiki modified the milestones: 0.6.0, 0.7.0 Oct 13, 2024
@kitsudaiki kitsudaiki removed the status in OpenHanami Oct 13, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@kitsudaiki kitsudaiki

Labels
C++ L: ★★★ Expect high difficulty New feature / Improvement new features and improvements of existing features Topic: Performance improve any kind of performance Topic: Scalability increase scalability of the network
Projects
OpenHanami
Status: No status
Milestone
0.7.0
Development

No branches or pull requests
1 participant
@kitsudaiki