Repository navigation
Conversation
Create a host-only Cargo workspace that specializes selected CUB forms through NVRTC and reports exact scratch layouts with their LTO output. Bind specialization identities to source, headers and toolkit bytes. Include a runnable provider example, Rust guide and CPU checks. Signed-off-by: Trent Nelson <[email protected]>
Install the pinned toolchain without updating rustup itself. Signed-off-by: Trent Nelson <[email protected]>
Serve native specialization batches through the versioned Oxide boundary and exercise Load and Store through a standalone Cargo consumer with a generic helper dependency. Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Reuse the same helper sites with 64- and 128-thread kernels and retain one device static across their emitted bodies. Check each launch plan and preserve fixed external-symbol diagnostics. Signed-off-by: Trent Nelson <[email protected]>
Select fixture output locations from raw CUDA coordinates and exercise payload initialization, access and copying at both item counts. Annotate compile-fail examples and clarify hierarchy ordering in the guide. Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]> # Conflicts: # rust/Cargo.lock
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Support both CUB warp-transpose algorithms for complete blocks. Preserve payload validity through Store permutation and query exact scratch layouts for regular and timesliced exchanges. Signed-off-by: Trent Nelson <[email protected]>
Use the group-first API for independent 32-lane tiles and preserve carrier ownership grain across block and warp operations. Select the matching Oxide compiler for scoped participation and access checks. Signed-off-by: Trent Nelson <[email protected]>
Combine the six block Load/Store algorithms with explicit storage and physical-warp Direct dispatch. Preserve group and storage forwarding through the common boundary and qualify timesliced storage reuse. Signed-off-by: Trent Nelson <[email protected]>
Use the common group-first API with power-of-two lane groups, preserving fixed carrier ownership and scope-correct controls through helpers. Signed-off-by: Trent Nelson <[email protected]>
Explain the user workflow, compilation and generic specialization before provider internals. Keep the source listings tied to the Cargo example and move provider details to their own page. Signed-off-by: Trent Nelson <[email protected]>
Allow algorithm.into() for full-tile calls and From constructors for option builders while preserving existing default inference. Use the conversion in the executable generic helper. Signed-off-by: Trent Nelson <[email protected]>
Observe the finite loop and later wider results, and distinguish participation from memory completion. Cover nested breaks and lane-local scalar work before a wider operation. Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Signed-off-by: Trent Nelson <[email protected]>
Contributor
|
Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually. Contributors can view more details about this message here. |
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
1 participant
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.
Physical and logical warps can now select Striped, Vectorize or Transpose through the same group-first Load/Store API. Supported logical widths are powers of two from 1 through 32. Transpose gets independent compiler-planned scratch per group and permutes values and validity on mutable Store; Striped and Vectorize preserve the payload.
The guide documents the supported algorithms and native shared-memory composition. Tests cover group origins, offsets, partial and scattered validity, mixed scopes, larger blocks and byte/64-bit accesses. The documented query example now uses the matching compiler revision too.
This draft follows #12114 and requires NVIDIA/cuda-rust#1586. Review this slice.
Validation: all 156 compiler controls, host tests, strict Clippy, rustdoc and guide builds pass. Focused Warp GPU checks pass, and separate-helper examples pass with thin/fat host LTO. Rust host CI passes on this head. The combined GPU suite passes, including 18 sanitizer runs and 20 expected-error checks. Direct/Transpose CUB comparisons are complete; overhead varies by workload. Final foundation review remains open.