| Re: [hatari-devel] add support for TT ram -> problem with HD emulation and DTA |
[ Thread Index |
Date Index
| More lists.tuxfamily.org/hatari-devel Archives
]
- To: hatari-devel@xxxxxxxxxxxxxxxxxxx
- Subject: Re: [hatari-devel] add support for TT ram -> problem with HD emulation and DTA
- From: Uwe Seimet <Uwe.Seimet@xxxxxxxxx>
- Date: Wed, 17 Dec 2014 17:06:52 +0100
- Dkim-signature: v=1; a=rsa-sha1; c=relaxed/relaxed; t=1418832412; l=2149; s=domk; d=seimet.de; h=In-Reply-To:Content-Transfer-Encoding:Content-Disposition: Content-Type:MIME-Version:References:Subject:To:From:Date; bh=PhMOL1rgG80+RzFWYfY3CX64jNs=; b=m7rZTTNOHk99lFJPSEmihzn+KHk8rXeUIQFYfSRxINk3ApL7oZhjyVIuWyt8hIQynKo t73ltoeq4YDy4QON+fNp4IKgEBQt81vTAaWzh+zA/uwQShZdwiqoVaskuc/hIv3C0LXK4 NEKl1COPLpYFvGl9BYs75TlLU4xEg0iUX0g=
Hi,
It's probably not documented what exactly happens, but it's documented that
it doesn't work. (And it does not work, indeed.)
Take care
Uwe
> Le 17/12/2014 16:50, Roger Burrows a écrit :
> >On 17 Dec 2014 at 14:19, Nicolas Pomarède wrote:
> >
> >>Le 17/12/2014 14:05, Uwe Seimet a écrit :
> >>>Hi,
> >>>
> >>>By the way, remember that one cannot transfer data from/to TT RAM using
> >>>ACSI/DMA. So for full compatibility it should be ensured that such a
> >>>transfer does not work.
> >>>
> >>
> >>Hi,
> >>
> >>yes, but it's the TOS or the specific program's responsability to
> >>allocate the correct kind of memory for DMA operations, there's nothing
> >>we can do at the HW emulation level to fix this (at best, we can print a
> >>log and ignore the read or only send '0' bytes if it's a write)
> >>
> >Exactly! If you're using Hatari to develop programs for real machines, you
> >don't want Hatari to allow something that real machines don't.
>
> yes, but problem is that those bad behaviours are not necessarily
> documented.
>
> For example, does the ASCI write 0 bytes when dma points to bad memory, or
> does it reads bytes after masking the lower 24 bits ? Sometimes you will
> point to a region where RAM is mirrored, but sometimes you will point to
> some bus error regions (in which case the DMA won't do a bus error, but what
> value will be sent to the HDC ?)
>
> This is not obvious to handle in all cases, as it requires lots of tests on
> all types of machines we don't always have access to.
>
> Those memory issues are really difficult to handle, it requires lot of time
> to test/document them. We can add some cases as we encounter them with real
> examples, but it would take too much time to try to handle everything.
>
> Nicolas
>
>
--
Dr. Uwe Seimet
Senior Software Developer
Quisma GmbH
Quality Internet Services & Marketing
Benzstrasse 15
76185 Karlsruhe
Amtsgericht München HRB 153042
Geschäftsführung: Johann Hermann, Sascha Wittler, Markus Plattner
Tel: +49 721 13249323
E-Mail: uwe.seimet@xxxxxxxxxx
http://www.quisma.com