Re: [hatari-devel] Hatari and OUTSIDE (virtual memory manager)

[Eero Tamminen <oak@xxxxxxxxxxxxxx>]

Hi,

On 10/10/18 6:46 PM, Uwe Seimet wrote:
> >   From there, what should I do to check the virtual memory ? Start a
> > program that needs a lot of RAM ? If I just open C from the desktop
> > displayed from outside.img, there's no problem.
>
> Yes, you should use some memory-hungry software. Starting an application
> without the Fastload-Flag set should also result in swapping.
>
> > At what point do you get crash / unstability in your case ?
>
> I got bus errors when starting the first application (forgot which one).
> I will test again once MEMWATCH is running with my setup, because
> MEMWATCH is less complex wrt PMMU usage, but in some respects (bus error
> handling) similar to OUTSIDE.
>
> Looks as if I have bad luck these days with Hatari: When I switch on
> cycle exact (but not prefetch or PMMU) the first time HDDRUTIL is
> launched it displays only parts of the name of devices managed by the
> NetFeats SCSI Driver for Linux.

Your NatFeats SCSI Hatari driver code is not flushing caches although
it writes emulated memory from host side (see my other mail).  If all
your MMU testing is done with NatFeats SCSI driver, it could explain
also rest of the issues.


> When launchting HDDRUTIL the second time
> the display is correct. Without cycle exact everything is fine right from
> the start. Looks as if this option causes a side effect that affects the
> SCSI Driver. Any idea what this could be?

Andreas said that although NeXT uses MMU extensively, Previous emulator
doesn't emulate cache, and WinAUE CPU cores didn't emulate cache with
MMU when MMU support was initially added to WinUAE & Hatari.

I think most of Hatari MMU testing was done before caches were emulated
also when MMU is enabled.

Toni, what about the WinUAE testing, was the WinAUE Linux & NetBSD MMU
testing done before or after you had added cache emulation support with
MMU?


	- Eero