Repository navigation
Decode ALP-RD floats on the GPU - #11707
Draft
vishalkotak wants to merge 1 commit into
Draft
vishalkotak wants to merge 1 commit into
vishalkotak wants to merge 1 commit into
Conversation
Add a standalone CUDA kernel for ALP-RD f32 and f64 arrays and let the CUDA compression preset choose ALP-RD. Signed-off-by: Vishal Kotak <22182951+vishalkotak@users.noreply.github.com>
vishalkotak
force-pushed
the
vishalkotak/alp-rd-cuda-kernel
branch
from
November 20, 2026 20:27
7174955 to
8e3707e
Compare
Contributor
Author
|
Tested on an NVIDIA L4 (CUDA 13.0): the full ALP's standalone kernel seems to ignore the slice offset too; I'll check that separately. P.S. I had to remove some logs from the below snapshot to not bloat the comment text. Test log |
This branch has not been deployed
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Full-precision floats that ALP can't encode fall back to the CPU on the GPU path, because ALP-RD had no CUDA decoder and the CUDA preset never picked it. This adds an ALP-RD kernel for f32 and f64, built on the patch chunk offsets from #11644, and lets the preset choose ALP-RD. Older files without chunk offsets still decode on the CPU.
Patch positions count from the unsliced array's 1024-row chunks, so the kernel decodes chunk-aligned rows, like bit-packing does. Codes outside the dictionary go to the CPU decoder instead of decoding to a wrong value.
Tested on an NVIDIA L4 with the full
vortex-cudasuite; notes in the comment below.I used Claude Code for analysis, debugging, and ideas.