diff options
Diffstat (limited to 'CONTEXT.md')
| -rw-r--r-- | CONTEXT.md | 14 |
1 files changed, 9 insertions, 5 deletions
| @@ -162,11 +162,15 @@ Single Rust crate (`x86_64-unknown-uefi` target). Module breakdown: | |||
| 162 | 162 | ||
| 163 | ## Known issues / perf notes | 163 | ## Known issues / perf notes |
| 164 | 164 | ||
| 165 | - **Scroll is slow under QEMU without KVM**: each scroll shifts the whole | 165 | - **Scroll cost** — scrolling still shifts the whole framebuffer (a |
| 166 | framebuffer up (~3 MB copy at 1280×800) with TCG-emulated stores. | 166 | ~3-8 MB pixel copy per row), but a RAM backbuffer (`gop::SHADOW`) |
| 167 | Deferred optimisation: ring-buffer rows / deferred pan so scrolling | 167 | makes it fast on real hardware: every raster write goes to the shadow |
| 168 | never touches the whole framebuffer. The vt's cell grid already avoids | 168 | too, and scrolling is a cached shadow memcpy + one write-only push to |
| 169 | re-rasterising on scroll (only the exposed bottom row is blitted). | 169 | the display. The display is never *read* — that was the killer, since |
| 170 | real-hardware GOP framebuffers are uncacheable/write-combining and | ||
| 171 | uncached reads cost a bus round-trip per cache line. Still on the | ||
| 172 | someday list: a pan-based ring so scrolling never touches the whole | ||
| 173 | framebuffer at all. | ||
| 170 | - **Editor redraw** re-blits changed cells per keystroke (fine at 1280×800); | 174 | - **Editor redraw** re-blits changed cells per keystroke (fine at 1280×800); |
| 171 | the "optimize drawing someday" item. | 175 | the "optimize drawing someday" item. |
| 172 | - **ramfs only**: edited files live in memory, lost on reboot until disk | 176 | - **ramfs only**: edited files live in memory, lost on reboot until disk |
