Repository navigation
perf(gpui): stream the editor preview into one texture (default on Linux) - #2854
Draft
richiemcilroy wants to merge 1012 commits into
Draft
richiemcilroy wants to merge 1012 commits into
richiemcilroy wants to merge 1012 commits into
Conversation
Contributor
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
This comment has been minimized.
This comment has been minimized.
richiemcilroy
changed the title
…moval transaction
perf(enc-gif): hand frames to gifski zero-copy via strided ImgVec
fix(mobile): preserve independent settings updates
fix(web): preserve hours in transcript timestamps
fix(web): recover failed voice-note playback
#3148) * feat(desktop-gpui): record panics, fatal signals and Windows exceptions in a crash log * feat(utils): include crash records and captured stderr in uploaded logs * style(desktop-gpui): rustfmt the crash capture changes * fix(desktop-gpui): log share export, upload and export task failures * fix(desktop-gpui): leave a running instance's session marker and stderr alone * fix(desktop-gpui): start the crash session after the instance lock, bound and fully capture stderr
ee60119 started writing start-times.json beside every segment for crash recovery, but the preparing projection's inventory never accounted for it. Every Studio recording then failed the projection with "Unrepresented source files or respawn layouts require ordinary finalization", and the editor showed the preparing shell without its live preview. Account for the file per segment; a cleanly stopped segment's timing still comes from its metadata.
…ding instead of dropping them (#3163)
…ng on Windows (#3162) * fix(desktop-gpui): turn camera blur off after a crash with blur running on Windows * style(desktop-gpui): drop comments that restate the code
…; keep Cap's windows out of the X11 picker (#3288) * Cover Linux X11 capture targets, audio, export, area and Instant uploads in CI; keep Cap's own windows out of the X11 picker * Leave the KDE harness to its own change
Remove the doc sentence that restates bgrx_plane_to_rgb24. Keep the swscale fallback note, the safety contracts, and the shuffle layout.
…itor (#3299) * Use the X display when WAYLAND_DISPLAY names no running compositor * Name the logged field after the variable * Check a stale WAYLAND_DISPLAY launch in the Linux CI * Drop comments that repeat the code
…a-nv12-9048 perf(rendering): AVX2 software RGBA→NV12
…dule refactor: move caption word utilities into cap-project
…grz-rgb24 perf(linux-screenshot): SSSE3 BGRZ/BGRA to RGB24
… app (#3348) * On GNOME, give Cap's windows the app id it registered with the portals * Run Cap once before the terminal-launch screenshot * Keep the window app id helper Linux-only
…3350) * Linux cancelled or crashed Instant starts delete their empty video, as on macOS * Check a cancelled Linux start for media after capture stops, and keep its video in the sweep
…dlines are not starved (#3366) * Run macOS playback and seek verify tests alone so frame deadlines are not starved * Split the macOS full verify tier into three shards * Split the macOS full verify tier into four shards
perf(web): commit instant recording sources with listings and one fewer workflow hop
…-callbacks fix(gpui): defer window callbacks AppKit sends while the app is mid-update
|
Review the following changes in direct dependencies. Learn more about Socket for GitHub.
|
This branch was successfully 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.
Stacked on #2852 (phase 2 of the gpui streaming-image design). Moves the editor preview's RGBA path to a
StreamingImage. It is on by default on Linux, and opt-in on Windows and for macOS's RGBA fallback.Change
EditorPreviewFrame::Stream(StreamingImage). The editor pump keeps one Rgba8 stream per editor instance.frame_arrived's update closure hands it to the stream on the main thread. This happens after theholds_instancecheck, so a stale pump can't change what the preview shows.Arc.frame_image)drop_imageof the previous frame: the window's dead-stream sweep releases the texture nowconvert_nanos/convert_samples, so the bench counters stay comparable.Kill switch
CAP_GPUI_STREAMING_IMAGE:0: back to per-frameRenderImages1: opts Windows and macOS's RGBA fallback inWindows is off by default because no CI runner could run the in-app A/B there (see below).
Tests
In-app editor A/B
CI scratch branch
ci/streaming-editor-ab, run 39670701971..capwith mic, generated withcap_verify::fixture.CAP_GPUI_AUTO_EDITOR=<fixture> CAP_GPUI_AUTO_TIMELINE_BENCH=1 CAP_GPUI_MUTE_AUDIO=1.CAP_GPUI_STREAMING_IMAGE=0/1interleaved (order alternating per round), 3 rounds, medians.timed_draw(paint) per framePumpStatsconvert avgPumpStats.presentedgpu_specsconfirms the passes ran on lavapipe Vulkan and llvmpipe GL 4.5.Linux lavapipe (Vulkan)
Linux llvmpipe (GL)
cleanhas no preview frames; its differences are noise.An earlier version wrote wgpu stream frames during paint. On GL that made draw p50 12–24% worse, while main-thread CPU was already lower. #2852 now defers that write to
before_frame, where the atlas uploads happen, and the tables above are from after that change.Windows
The in-app A/B could not run on any available CI runner:
The Windows evidence is therefore the D3D11 WARP atlas micro-benchmark in #2852: 1.2–2.1x faster, flat at 1080p, 0 large allocations per frame. That's why Windows is opt-in until the A/B runs on a real Windows machine.
Atlas micro-benchmark
See #2852. On Linux, wgpu on lavapipe Vulkan / llvmpipe GL makes the atlas step 1.4–3.8x faster at p50. Textures created per 600 frames drop from 600 to 1, and large allocations per frame drop from 2 to 0.