Skip to content

Tags

Tags

archive/stash/20261113-0

Toggle archive/stash/20261113-0's commit message
WIP on fix/launch-budget-on-cpu: 6a640fd regress: content_shown_cpu_m…

…s before content_shown_ms, clear of #540's named()

archive/fix/settle-until-input-taken

Toggle archive/fix/settle-until-input-taken's commit message
Parked: a step waiting for its input taken needs a Ping answered with…

…out idle

Not merged. A full gate run on seldon (with fix/glide-sampled-clock merged
in) failed progress-turning every time, alone at load 2 too: a Bloom host
answers a Ping read in a turn that queued nothing only once JavaScript is
idle (Window::answerPingsItself), and a turning spinner keeps it busy, so
the step waited the full five seconds and its work finished first. That
happens whenever the step's events reach the host before the Ping (a key
through the input method). Pinging only after steps that give input fixed
the first version's login-setup-qr-* and notifications-reader failures,
not this one.

Doing it right needs a Ping the host answers once JavaScript ran what came
with it, without waiting for idle: a protocol message serving the director
alone. transform (the reason for this branch) did not reproduce in 30 runs
at load 64-87, so the branch is parked; revisit only if transform recurs in
the gate.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

archive/wip/saturation-s6-pages

Toggle archive/wip/saturation-s6-pages's commit message
Saturation benchmark: S6's pages compress about 2:1, as an applicatio…

…n's do

The memory hog wrote one byte in each page and left the rest zero, so its pages compressed to almost
nothing: zswap (ADR 0066, decision 7) would have made S6 free. Each page is now half a xorshift's bytes
and half zeros, about 2:1 with lz4, as an application's anonymous memory compresses.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

archive/wip/saturation-benchmark-history

Toggle archive/wip/saturation-benchmark-history's commit message
WIP: tests

archive/wip/s6-claims-B

Toggle archive/wip/s6-claims-B's commit message
Merge remote-tracking branch 'origin/fix/gpu-buffer-released-first' i…

…nto HEAD

archive/wip/s6-claims-A

Toggle archive/wip/s6-claims-A's commit message
Merge remote-tracking branch 'origin/fix/gpu-buffer-released-first' i…

…nto HEAD

archive/wip/s6-ab-with-scopes-B

Toggle archive/wip/s6-ab-with-scopes-B's commit message
Measurement only: step 6 (memory) with one scope per application (#516)…

…, for step 6's A/B

archive/wip/s6-ab-with-scopes-A

Toggle archive/wip/s6-ab-with-scopes-A's commit message
Measurement only: the saturation benchmark (#538) with one scope per …

…application (#516), for step 6's A/B

archive/wip/memory-pressure-before-rebase

Toggle archive/wip/memory-pressure-before-rebase's commit message
Memory pressure: a level comes down only under half its threshold for…

… 10 s, and is told again at most once a window

archive/wip/host-ui-thread-concurrent-2

Toggle archive/wip/host-ui-thread-concurrent-2's commit message
js-thread: formatted as .clang-format says

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>