radeontop
Similar to htop, but for radeon gpu
Download/extract the source https://github.com/clbr/radeontop
Load Fatdog devx. Run make; make install
Then run it : radeontop
Similar to htop, but for radeon gpu
Download/extract the source https://github.com/clbr/radeontop
Load Fatdog devx. Run make; make install
Then run it : radeontop
On my regular boot, vesa framebuffer (and overlayfs fatdog in a Xvnc (run as spot) that the framebuffer vnc's into), the gpu ... is idle. CPU works harder, but at a acceptable 16 bit colour depth and reduced frame rate is still mostly pretty quiet. The battery lasts ages without the gpu power drain
whilst the general look-n-feel is OK/good.
But then again the otherwise idle fast gpu vram ... could be used as a ram-disk. Not easily set up, is hardware specific and requires driver modifications, however if enabled then Fatdog could be loaded-into/run-from that
Something for the back-burner, as my current focus in on lossless remote low bandwidth/reasonable quality vnc
fatdoguser wrote: ↑Sat Feb 10, 2024 2:33 ammy current focus in on lossless remote low bandwidth/reasonable quality vnc
So far I've assumed a 640x480 input ... whatever resolution/screen scaled to that (early days simplification for more of a proof of concept).
Reduce 32 bit color down to 16 bit is still OK quality, halves the amount of data.
Reduce the frame rate and for remote working even 8 fps is still usable. Down from 60 fps = 13% of the data (87% 'compression). So combined with the color depth reduction, 6.66% of the data (93% 'compression).
With a additional phase alternate lines, phase alternate pixels, that's like scaling down to 25% (75% reduction), so collectively 98.3% 'compression'. For that last stage, each frame alternates/cycles, so after 4 frames every pixels color has been conveyed.
Less than 10Mb/sec. Extremely fast/simple processing, more suited to 'live'. For non-live, watching a youtube for instance and buffering has low entropy, high compression potential (sequential frames tend to be similar), on average another 66% reduction.
Lossless in the sense that each second will see every pixel updated multiple times with the actual pixel color, but in a phased manner. Lower color depth, lower frame rate, slower color transitions for each pixel, but all still fast enough that the eye sees it as being 'acceptable/OK'.
If/when sound is encapsulated with that then as generally that involves a element of sound buffering so also might the video be buffered/compressed and sent/delivered as combined synchronized audio/video.
Fatdog as a development system/tool is great
@fatdoguser thank you. Good find. radeontop will be included in the base for the next release.