Repository navigation
Conversation
… hc-read second half stops overwriting the next layer's base stamp kProfPer 32 -> 33: fused_gr_read_multi writes three stamps per half and the second half's base was 30, so its up-stamp landed at slot 32 = (l+1)*32 + 0, the next layer's stage-0 base. Derived columns (gap/head/hc0) also wrapped to ~1.8e19 ns whenever a slot was unstamped, printing ~1e14 ms per window. A delta now counts only when both stamps are nonzero and monotonic; the hc0 split replaces the single hc-read0 column only when the split chain is complete. Metrics-only: no computed model value changes. Reason recorded in PORTING.md 19.
Owner
|
Thanks! This is in the 0.1.32 batch. We take it as is; every change goes through our identity gate (byte-identical output on four models) and a speed A/B before it ships. |
Owner
|
Released in 0.1.32 as is. Merged under your name (GitHub doesn't show it as merged because it went in by hand). Thank you! |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Fixes the
STRATA_VERIFY_PROFILEstage report so its derived columns ((gap),head,hc0 norm/down/up) print sane ms/window values instead of\~1e14 ms. Metrics-only — no computed model value changes; this only affects the profiler's reported columns.Two integer-arithmetic defects:
Slot collision (
kProfPer32 -> 33). The hc-read second half'supstamp was written to slot 32, which is the next layer's slot 0. That both corrupted the next layer's base stamp and left the up-stamp unreadable.Unsigned wrap on absent/out-of-order stamps. A slot no kernel stamped is
0; an unsignedto - fromwherefromwrapped to\~1e19 ns, which is the\~1e14 mswe were seeing.The fix introduces a
gap(to, from)helper that contributes0unless both stamps are nonzero and monotonic (to >= from), bumpskProfPerto 33, and shows the hc-read0 norm/down/up split only when the whole split chain is present (so it isn't counted twice). The column-header array is extended to match.With the columns sane, the verify-window residual could be attributed: it is the
wait for ringscolumn (28 ms/round), and a no-doorbell floor (STRATA_VERIFY_DEVICE_PLAN=1) leaves it unchanged, so it is host-ring submission latency, not the expert-pool doorbell round-trip. Deliverable 3 of #228, re-scoped to stand alone per your request.