Re: [hatari-devel] FPU detection, who is wrong?

[ Thread Index | Date Index | More Archives ]

On Tue, Feb 17, 2015 at 10:56 AM, Thomas Huth <th.huth@xxxxxx> wrote:
> I have a 68881 in my MegaSTE, for instance.

Any chance that you can try to find out the revision value of the
68881 that is put on the stack by the FSAVE instruction?

Accidentally, I've gotten hold of two MegaSTEs, one with a 68881 and another with a 68882 so why not to give it a try?

Not only it has proved that the current FreeMiNT kernel (since the 31 of May) is working well with peripheral FPUs now but also a couple of interesting numbers:

- the 68881 indeed reports revision number as 0x3f (and frame size = 0x18)
- the 68882 reports revision number as 0x00! and yes, the stack frame size number is correct (0x38)!

Therefore, the check in FreeMiNT wasn't correct as the revision number field is really unreliable.

Also, I could verify that 68882 is slightly faster, even if not optimized for the interleaved instructions.

MiKRO / Mystic Bytes

Mail converted by MHonArc 2.6.19+