Repository navigation
Conversation
…ts for rows The one Engram figure was an undivided host step, and a decode that followed a prompt added its own Engram work to it. The step is now timed only while a prompt's buffers are open, and split into the wait for the reader's rows and the normalization.
giveen
force-pushed
the
engram-prompt-timing
branch
from
November 19, 2026 22:27
3f35369 to
b7df086
Compare
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Split out of #466 at review 5982149603, item 3, so that PR is only the ring retry.
What changed
engram_steptimed the whole host step into one figure, and it ran on every call: a decode that followed a prompt added its own Engram work to the prompt line. Now:big_.rows != 0), so the line reports a prompt;PleReader::collectand the normalization loop, printed as(%.0f ms waiting for rows, %.0f ms normalizing);The log line's prefix, level and the figures the server parses are unchanged; the new text is inside the existing Engram clause.
Verification
cmake --build build -j --target model_hy4flash: clean.collect, two around the normalization, per Engram layer per prompt chunk). No bench run: the before/after numbers live in hy4flash: improve single-GPU NVFP4 decode on hybrid CPUs #466 and are not affected by this.