Alpaka: Vecmem Interface #960
Open
Conversation
This file contains hidden or 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
Meant as an example/guidance for Ryan Cross.
Needs testing on both HIP and SYCL still.
Can be removed when vecmem is updated.
Alpaka, unlike CUDA, doesn't have an equivalent of "Get Current Device", such that we can't really use that. This is hopefully a reasonable middle-ground, as if a user does care about what device they are using, they should pass a device over. Otherwise, they get a reasonable default (and per device environment arguments can be used to change this, as they would for native CUDA, HIP & SYCL code.
CrossR
commented
May 2, 2025
|
Tested with my async branch too, and after a few minor changes to the Queue using code, works there too. |
krasznaa
requested changes
May 7, 2025
There was a problem hiding this comment.
I like how things look, but let's do one other kind of cleanup as part of this PR!
Now that traccc::alpaka takes care of everything internally, practically all executables should become "simple C++ code". I.e.
https://github.com/acts-project/traccc/blob/main/examples/run/alpaka/CMakeLists.txt#L12-L13
should no longer be needed I believe. (But you'll need to test this.)
CrossR
commented
May 8, 2025
| @@ -38,7 +35,7 @@ add_library( traccc_examples_alpaka STATIC | |||
| "full_chain_algorithm.hpp" | |||
| "full_chain_algorithm.cpp" ) | |||
| target_link_libraries( traccc_examples_alpaka | |||
| PUBLIC alpaka::alpaka vecmem::core detray::core detray::detectors | |||
| PUBLIC vecmem::core detray::core detray::detectors | |||
There was a problem hiding this comment.
alpaka calls now removed from the examples CMakeLists.
CrossR
force-pushed
the
AlpakaVecmemOrganization-main-20250501
branch
from
4e6c1e9 to
b29c2c2
Compare
CrossR
force-pushed
the
AlpakaVecmemOrganization-main-20250501
branch
from
3905ebe to
2c4e954
Compare
|
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.
Based on the code in #959, just with a few small changes:
ifdefguard aroundasync_copyinit in theCUDA || HIPbit, just whilst we are still on an older vecmem release. Can be updated when we update that, though no rush as the async copy is not used in Alpaka yet.alpaka::Queueto default to device 0 in cases where there is no device given. Any opinions on this welcomed. I didn't / couldn't go down the CUDA route (which isif device == INVALID_DEVICE get_current_device()), as Alpaka has no concept of the current device. Could instead just remove the default and force a device ID to be given? I think either is preferable to just silently failing.I've deleted the previous code now as well, which removes Alpaka from the
traccc::alpakapublic headers once again.