Benchmark Breeze on Kryon's desktop host against raylib

Both Linux executables, built from the same tree, ran the scripted session
alternately on a private Xvfb. On the desktop host idle CPU falls from 62%
to 3% of a core, CPU per frame from 26 ms to 2.5 ms (raylib draws through
llvmpipe there) and memory by about a quarter. A missing audio device is
tried again at most every ten seconds on the desktop host, where each failed
open holds a frame while SDL tries its audio servers.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
This commit is contained in:
waoziandClaude Opus 5.5 committed 2026-11-16 19:55:08 -03:00
1 parent 307e3c05ac
commit bd6d5a2c41
2 files changed
+82 -14

No files matched your search

@@ -120,8 +120,66 @@ host under the same load.
CPU per frame above the frame interval means more than one core works per
frame: llvmpipe rasterizes on its own threads.
Not measured yet: clean and incremental build times and generated C for the
raylib state; take them from the last raylib commit before the port lands.
## After: raylib host and Kryon's desktop host, same source
Measured 2026-11-16 19:18–19:44 on a private Xvfb (1280x1000x24, llvmpipe),
load average 13–24 on 32 CPUs. Both executables were built from the same
working tree on 307e3c05 with Ziran 0.6.0 (d17c7aa1), Kryon 1b290f19, the same
embedded bundles and `KRYON_TEST_DRIVER`. Each run is one copy of the
scripted session; the two builds alternated, five runs each, so both saw the
same load. The desktop host build still links a throwaway stub file for
K6–K8 (outside the repository).
| Measurement | raylib host, median (range) | desktop host, median (range) | Change |
|---|---|---|---|
| Time to first frame (ms) | 415.8 (377.4–459) | 474.3 (371.1–749.3) | +14% |
| Frame rate while interacting, 120-frame reports (fps) | 60.47 (60.44–60.48) | 60.43 (58.31–60.44) | 0% |
| App frame time, 120-frame report median (ms) | 0.92 (0.83–0.96) | 2.19 (1.83–2.50) | +138% |
| Longest app frame while interacting (ms) | 39.05 (37.43–40.79) | 101.5 (44.77–107.8) | +160% |
| Frame interval p50 (ms) | 16.68 (16.67–16.68) | 16.81 (16.79–16.86) | +1% |
| Frame interval p95 (ms) | 33.34 (33.34–33.34) | 67.06 (67.03–67.10) | +101% |
| CPU per frame p50 (ms) | 26.17 (25.21–27.09) | 2.54 (2.25–2.90) | −90% |
| CPU per frame p95 (ms) | 32.95 (31.55–34.58) | 4.30 (3.66–5.38) | −87% |
| Frames during interaction | 1336 (1330–1342) | 1191 (1188–1191) | −11% |
| Idle CPU over 30 s (%) | 61.72 (60.82–63.19) | 2.63 (1.90–3.86) | −96% |
| Steady RSS, idle median (MiB) | 494.8 (489.8–495.8) | 361.3 (327.6–381.0) | −27% |
| Peak RSS (MiB) | 494.7 (489.7–495.8) | 362.6 (329.1–380.5) | −27% |
| Executable | raylib host | desktop host |
|---|---|---|
| Size (bytes, stripped) | 100,296,744 | 99,303,400 |
| Generated C | 863 files, 432,234 lines | 873 files, 437,203 lines |
| raylib markers | 3 of 4 | 0 of 4 |
| Linked libraries | 10: SDL2, GLESv2, gio, glib, ssl, crypto, z, m, atomic, c | 15: SDL2, pangocairo, pango, cairo, fontconfig, freetype, gio, gobject, glib, ssl, crypto, z, m, atomic, c |
Build time of the same two private builds (Ziran C generation, `cc` on 8
jobs, link; embedded bundles and vendor libraries already built), alternating
two rounds, load average 18–34:
| Build | raylib host | desktop host |
|---|---|---|
| Clean (s) | 47.5, 45.6 | 51.0, 41.4 |
| No change (s) | 10.5, 8.8 | 14.0, 9.2 |
Within the noise of a shared machine, the two hosts build in the same time.
Reading the numbers:
- Xvfb has no GPU, so raylib draws through llvmpipe: about 26 ms of CPU per
frame spread over llvmpipe's threads, and it keeps drawing at 60 fps while
nothing changes. On a real GPU raylib's per-frame and idle CPU are far
lower; the desktop host draws with Cairo on the CPU either way. The idle
and memory results carry over best to real hardware: the desktop host
reuses an unchanged frame and drops to 15 fps.
- App frame time on the desktop host includes Cairo's painting, which raylib
hands to the GPU (llvmpipe) outside the app's frame.
- Frame interval p95 of 67 ms and fewer frames are the desktop frame clock
pacing unchanged frames at 15 fps, not dropped frames; the 120-frame fps
matches raylib.
- The longest desktop frames: one 100 ms wait per run while SDL tries the
missing audio device (the benchmark has none; Breeze retries at most every
10 s on the desktop host) and one 115 ms frame where a screen first lays
out. Before the 10 s back-off every cue retried and stalled 100–200 ms.
## Port map: raylib call sites and their replacements
@@ -211,10 +269,8 @@ branch for the other platforms and a desktop branch:
## Per-platform checklist
Linux desktop:
- [x] Every raylib call site above ported; the desktop graph generates and
compiles. A private link (build/desktop-host/breeze, 96.3 MiB) has no
raylib markers and no GLES; it links SDL2, Pango, Cairo, fontconfig,
FreeType, GLib/GIO and X11.
- [x] Every raylib call site above ported and committed (307e3c05); the
desktop graph generates, compiles and links with no raylib and no GLES.
- [ ] Kryon K5–K8 (with the parent coordinator): Cairo picture-cache
invalidation; `DisposeFrameTexture` and route-fade frames;
`ImagePreviewDecode`/`ImagePreviewRelease`; `notification_desktop`'s
@@ -223,8 +279,7 @@ Linux desktop:
- [ ] Switch the shared Makefile's native build to `DESKTOP_HOST`
(desktop libraries from pkg-config, the audio package's MP3 and Vorbis
archives) once the link needs no stubs.
- [ ] Run the scenario on a private Xvfb; then the after benchmark,
alternating with the saved raylib build so both see the same load.
- [x] The scripted session runs on a private Xvfb; after benchmark above.
- [ ] Delete the Linux path's raylib build rules (`RAYLIB_A` for native,
`KRYON_BACKEND_RAYLIB`, `AVIF_RAYLIB`, raylib `SUPPORT_*` flags).
- [ ] Rename `raylib_text_input.zi` and move the key constants out of
@@ -252,9 +307,7 @@ sent to the parent coordinator on 2026-11-16:
## Steps
1. Kryon K5–K8, routed by the parent coordinator, pushed, then pinned here.
2. Remove the live-Kryon `ziran.local.toml` someone added on 2026-11-16 15:57
(its owner, via the coordinator) so builds use the pinned Kryon again.
3. Finish the Linux checklist above, then the after benchmark and the
comparison here.
4. Windows, web and Android, platform by platform, each with its raylib code
2. Switch the shared Linux build to the desktop host and delete its raylib
rules (checklist above); rerun the after benchmark on that build.
3. Windows, web and Android, platform by platform, each with its raylib code
deleted as it moves.
+16 -1
View File
@@ -15,6 +15,7 @@ MixedAudioCallback :: #type (buffer: *void, frames: u32) -> ();
#import "audio_decode_mp3"
#import "audio_decode_vorbis"
#import "byte_text_linux"
#import, file "host_clock.zi";
AUDIO_SOUND_SLOTS :: 64
AUDIO_MUSIC_SLOTS :: 16
@@ -228,8 +229,22 @@ MixedAudioCallback :: #type (buffer: *void, frames: u32) -> ();
}
}
// A device that failed to open is tried again at most every
// AUDIO_RETRY_SECONDS: each failed open can hold a frame for a tenth of a
// second while SDL tries its audio servers.
AUDIO_RETRY_SECONDS :: 10.0
audio_retry_at: float64;
audio_init :: () -> bool {
return Sink.AudioSinkOpen()
now := HostSeconds()
if now < audio_retry_at {
return false
}
if Sink.AudioSinkOpen() {
return true
}
audio_retry_at = now + AUDIO_RETRY_SECONDS
return false
}
audio_close :: () {