Freezing OSD/mouse under heavy I/O load : ARM at fault ?

[mahen]

Hi everyone !

My question may sound like a tautology but actually, I'm not that sure that's normal.

When making a copy or extracting an archive or accessing the internet through (UART PPP mode + MiamiDX)
- the whole AmigaOS starts to crawl. I first thought that was normal but also :
- even the mouse pointer freezes. I cannot remember if that happened on my real 030 Amiga.. Isn't it a hardware pointer ? But also what leads me to think there's something wrong is
- even the MiSTer OSD becomes completely unresponsive for several seconds and doesn't catch all my inputs

So, I was wondering. Maybe the culprit is not the simulated Amiga but the way the ARM handles I/O priorities when it's being asked too much at the same time ?

How would a regular Amiga behave ? In my memories, even when the Amiga was significantly slowed down, it always remains at least partly responsive (mouse, dragging a window even if the content wouldn't follow, typing stuff in the terminal, even if the latter wouldn't answer)

Please forgive me if my question sounds stupid / mixes several unrelated topics ?

Cheers
Mahen

[Anything]

Pretty sure I've noticed the same and you're right, the Amiga mouse was a hardware pointer so didn't slow down when the machine was busy.

[ExCyber]

Linux systems have sometimes had similar problems with native desktop environments, even with a hardware cursor and running on much faster processors. A lot of work has gone into improving this over the years, but I assume it's still possible to hit a bad set of circumstances.

In a lot of cases it's related to the details of filesystem and storage performance. What underlying storage device are you using when this happens? Frankly, a lot of flash devices are shockingly bad at handling even a modest rate of random writes. The internet issues might also be explained by this if you're downloading to disk (either explicitly or because something is maintaining a cache).

[rhester72]

@ExCyber - that shouldn't affect MiSTer at all. Storage services are running on the Linux side of the house, and the FPGA shouldn't be subject to that at all.

If a hardware pointer wasn't affected on a real Amiga, it shouldn't be on MiSTer, either.

[ExCyber]

I was under the impression that the Linux side is still responsible for reading/updating the mouse movement in the first place unless using something like SNAC or BlisSter.

[mahen]

Yep, all I/Os go through Linux AFAIK, and usually it's super low latency (1000 Hz), but maybe we hit some corner cases when there is a huge I/O load. Could input be priorized over disk / UART I/O ?

As for the other questions, I'm actually using a fast SD card, very fast compared to the bandwidth actually used by the Amiga.