Skip to content

fix(veryl): roll HEAD to fd1c3a650789 - #1068

Merged
tignear merged 6 commits into
developfrom
integration/veryl-head
Oct 8, 2026
Merged

tignear merged 6 commits into
developfrom
integration/veryl-head

Conversation

@celox-bot

@celox-bot celox-bot Bot commented Oct 6, 2026

Copy link
Copy Markdown
Contributor

Roll all Veryl crates to one exact upstream master revision.

CI success merges this automatically into the non-release develop lane. CI failure deliberately leaves this PR open so it can be taken over and fixed. Do not retarget this PR to master.

@celox-bot
celox-bot Bot enabled auto-merge October 6, 2026 03:49
Veryl fd1c3a6 keeps an unpacked-array slice as a range on the reference's
VarIndex instead of expanding instance inputs into one expression per
element. Fold that range into Celox's flat part select on the selected
dimension, and reject runtime slice bounds with a diagnostic since SIR
addresses unpacked arrays one element at a time.
A runtime slice (a[i+:k], a[i-:k], a[i step k]) is read as one run of k
elements from a start clamped into the array, then shifted by how far the
clamp moved it; positions shifted in are exactly the elements outside the
array and read X (0 when two-state). Coordinates are read as two-state
values so the run index stays in bounds for X or Z indices.

SIR gains SIROffset::ElementRun, a load of consecutive whole elements
from a runtime element index. The layout keeps arrays read by a run
packed, optimizer passes treat it as a dynamic read of the whole object,
and every backend addresses it as index * element_width.
…head

Resolve register-rename conflicts in favor of the shared use visitors from
develop and teach them SIROffset::ElementRun. Develop's out-of-range guards
assumed a runtime part select is packed; a run of unpacked elements is
instead guarded by its first element and read whole, and a 64-bit anchor no
longer overflows the comb guard's bound check.
@codspeed

codspeed Bot commented Oct 7, 2026 •

Copy link
Copy Markdown

Merging this PR will not alter performance

✅ 4 untouched benchmarks


Comparing integration/veryl-head (b2f982b) with master (6c5b098)1

Open in CodSpeed

Footnotes

  1. No successful run was found on develop (1c33037) during the generation of this report, so master (6c5b098) was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩

@celox-bot
celox-bot Bot added this pull request to the merge queue Oct 7, 2026
@github-merge-queue
github-merge-queue Bot removed this pull request from the merge queue due to no response for status checks Oct 7, 2026
@tignear
tignear added this pull request to the merge queue Oct 7, 2026
@github-merge-queue
github-merge-queue Bot removed this pull request from the merge queue due to no response for status checks Oct 7, 2026
@tignear
tignear added this pull request to the merge queue Oct 7, 2026
@github-merge-queue
github-merge-queue Bot removed this pull request from the merge queue due to no response for status checks Oct 7, 2026
@tignear
tignear added this pull request to the merge queue Oct 7, 2026
@github-merge-queue
github-merge-queue Bot removed this pull request from the merge queue due to failed status checks Oct 7, 2026
@tignear
tignear added this pull request to the merge queue Oct 8, 2026
@github-merge-queue
github-merge-queue Bot removed this pull request from the merge queue due to failed status checks Oct 8, 2026
Use VarIndex.indices and the single InstInput.expr introduced by the pinned Veryl HEAD revision. This fixes compilation after merging develop, which added interface capture lowering against the released API.
@tignear
tignear added this pull request to the merge queue Oct 8, 2026
Merged via the queue into develop with commit 739d191 Oct 8, 2026
33 checks passed
@tignear
tignear deleted the integration/veryl-head branch October 8, 2026 17:21
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant