andros/emacs-gpu
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

README: chart the benchmark tables and align the macOS one

Browse source 
Add a mermaid xychart under each performance table (GPU/cairo ratio
with a parity line for the two Linux tables, CPU cost of sustained
scroll for the macOS one) so the takeaway is visible at a glance; the
tables stay as the source of record.  Right-align the numeric columns
of the macOS table to match the Linux ones.
This commit is contained in:
Andros Fenollosa 2026-06-12 22:09:12 +02:00
parent e22b30263a
commit 9c77b8e1fe
1 changed files with 27 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

28
README.md
View file

@ -237,6 +237,15 @@ rather than the 60 Hz cap (median of 3 runs, redisplays per second):
| Typing (1 char + redisplay) | 1857 fps | 1311 fps | 0.71x |
| Image scroll | 1359 fps | 1239 fps | 0.91x |
```mermaid
xychart-beta
title "1616x912 laptop frame: GPU throughput relative to cairo (1.0 = parity)"
x-axis ["Line scroll", "Page scroll", "Full redraw", "Typing", "Image scroll"]
y-axis "GPU / cairo ratio" 0 --> 1.4
bar [0.92, 1.00, 1.19, 0.71, 0.91]
line [1.0, 1.0, 1.0, 1.0, 1.0]
```
The big win is structural: glyphs **and** solid fills (backgrounds,
underlines, boxes) share one submission-ordered vertex batch, clipped
on the CPU, so a whole redraw flushes as a handful of draw calls --
@ -279,6 +288,15 @@ side excludes on-screen present) flip the result:
| Typing | 238 fps | 1766 fps | **7.4x** |
| Image scroll | 115 fps | 1328 fps | **11.5x** |
```mermaid
xychart-beta
title "4K frame: GPU speedup over cairo (1.0 = parity)"
x-axis ["Line scroll", "Page scroll", "Full redraw", "Typing", "Image scroll"]
y-axis "speedup (x)" 0 --> 12
bar [2.05, 1.22, 1.84, 7.4, 11.5]
line [1.0, 1.0, 1.0, 1.0, 1.0]
```
cairo slows down roughly linearly with the pixel count; the GPU barely
moves. Image scrolling is the extreme case (cairo re-blits the image from
CPU memory every frame, the GPU re-composites a cached texture). This,
@ -307,13 +325,21 @@ font-locked `xdisp.c`, same binary with and without the GPU backend;
`/usr/bin/time -l` over scripted workloads):
| Workload | Stock (Cocoa) | GPU, vsync on (default) | GPU, vsync off |
|---|---|---|---|
|---|---:|---:|---:|
| Sustained scroll, redisplays/s | 481 | 324 | 475 |
| CPU for 15 s of that scroll | 16.0 s | **10.5 s** | 15.5 s |
| Typing throughput (chars/s, machine-paced) | 108 | 52 | 106 |
| Idle (8 s) CPU | 1.21 s | 1.19 s | same |
| Peak RSS | ~140 MB | ~144 MB | same |
```mermaid
xychart-beta
title "M1 Pro: CPU seconds burned during 15s of sustained scroll (lower is better)"
x-axis ["Stock (Cocoa)", "GPU, vsync on", "GPU, vsync off"]
y-axis "CPU seconds" 0 --> 18
bar [16.0, 10.5, 15.5]
```
Honest reading:
- **Machine-paced throughput and CPU cost match the stock backend**