[MISC] Move register-tile Cholesky into quadrants.#2860
Merged
duburcqa merged 4 commits intoJun 5, 2026
Conversation
Delete `genesis/utils/_tile16.py` and `_tile32.py`, whose contents (the inlined column-setter cascades + qd.static-unrolled cholesky_ body) now live in `quadrants/lang/simt/_tile16.py` and `_tile32.py` upstream. The constraint solver switches from `Tile16x16Cholesky` / `Tile32x32Cholesky` to `qd.simt.Tile16x16` / `qd.simt.Tile32x32`, and the blocked-Cholesky bodies go back to the readable slice syntax (`tile[:] = arr[r:r2, c:c2]`, `tile -= qd.outer(v, v)`) instead of the explicit `_load3d` / `_store3d` / `_resolve_vec3d` / `_ger_sub` private calls that were required while two distinct tile classes co-existed in one kernel. Pairs with quadrants PR hp/import-tiles-from-genesis.
duburcqa
reviewed
Jun 1, 2026
Comment on lines
2113
to
2133
| @@ -2130,11 +2125,11 @@ def func_cholesky_and_solve_fused_tiled( | |||
| """Tile-size dispatcher; see _cholesky_and_solve_fused_tiled_impl for the algorithm and dispatch rule.""" | |||
| if qd.static(static_rigid_sim_config.cholesky_tile_size == 32): | |||
| _cholesky_and_solve_fused_tiled_impl( | |||
| constraint_state, rigid_global_info, static_rigid_sim_config, Tile32x32Cholesky, write_L_to_nt_H | |||
| constraint_state, rigid_global_info, static_rigid_sim_config, qd.simt.Tile32x32, write_L_to_nt_H | |||
| ) | |||
| else: | |||
| _cholesky_and_solve_fused_tiled_impl( | |||
| constraint_state, rigid_global_info, static_rigid_sim_config, Tile16x16Cholesky, write_L_to_nt_H | |||
| constraint_state, rigid_global_info, static_rigid_sim_config, qd.simt.Tile16x16, write_L_to_nt_H | |||
| ) | |||
Collaborator
There was a problem hiding this comment.
Would it be possible to have qd.simt.Tile that accept static "size" argument, so that ce can pass size as input argument instead of entire class? This would be more idiomatic.
Collaborator
There was a problem hiding this comment.
And even more natural and cleaner, because you would just have to pass "static_rigid_sim_config.cholesky_tile_size" directly inside _cholesky_and_solve_fused_tiled_impl instead of adding one more indirection level as it is currently the case.
Collaborator
Author
There was a problem hiding this comment.
would be better to comment these on the quadrants-side PR. I've already merged that though.
Collaborator
Author
There was a problem hiding this comment.
(No way to modify such parameters from genesis side)
duburcqa
previously approved these changes
Jun 1, 2026
|
🔴 Benchmark Regression Detected ➡️ Report |
duburcqa
changed the title
Collaborator
Author
|
🙌 |
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
2 participants
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.
Delete
genesis/utils/_tile16.pyand_tile32.py, whose contents (the inlined column-setter cascades + qd.static-unrolled cholesky_ body) now live inquadrants/lang/simt/_tile16.pyand_tile32.pyupstream. The constraint solver switches fromTile16x16Cholesky/Tile32x32Choleskytoqd.simt.Tile16x16/qd.simt.Tile32x32, and the blocked-Cholesky bodies go back to the readable slice syntax (tile[:] = arr[r:r2, c:c2],tile -= qd.outer(v, v)) instead of the explicit_load3d/_store3d/_resolve_vec3d/_ger_subprivate calls that were required while two distinct tile classes co-existed in one kernel.Pairs with quadrants PR hp/import-tiles-from-genesis.
Description
Related Issue
Resolves Genesis-Embodied-AI/Genesis#
Motivation and Context
How Has This Been / Can This Be Tested?
Screenshots (if appropriate):
Checklist:
Submitting Code Changessection of CONTRIBUTING document.