[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

68060 FP problem (fwd)



I have been forwarding 68060 gcc bug reports from this fellow to the egcs
group.  In conversations, he mentioned that this... I think RTEMS is
broken since we only account for 332 bytes for FP context  where he says
we need 400 bytes on both the 040 and 060.  I would appreciate some
feedback from some of you who have actually run RTEMS on an 040. :)

---------- Forwarded message ----------
Date: Thu, 26 Mar 1998 10:19:24 -0800
From: David Barto <barto at rabbit.visionpro.com>
To: Joel Sherrill <joel at oarcorp.com>
Subject: Re: 68060 FP problem (fwd) 

In message <Pine.BSF.3.96.980326112109.9859d-100000 at vespucci.advicom.net>you wr
ite:

>> Has RTEMS been updated to manage the '40 FPU on a task switch? When
>> I updated to the '40, I just left this bit out, because I don't use
>> the FPU very much and I \know/ that only one task ever uses it.
>> 
>
>What's different about the 040 from previous generations from a context
>switch point of view?  The 68881 was the first FPU supported in RTEMS --
>way back when. :)
>
>I know the 060 requires different context switch code for the FPU bit was
>not aware of this on the 040.
>
>
The 040 and 060 fpu are more similar in layout than the 881/040 are

The 040 changed the way the registers are saved, what is saved
and the way to save.

The 68040 Users Manual, (Motorola 1989, Page 9-30) Shows the results of
FSAVE/FRESTORE and the fact that it now saves up to 0x64 long words of
information on the stack.

The 68060 (Motorola 1994, page 6-36) Always generates the 3 long word
execption vector.

This is much different from the '881 code.

	-barto
-- 
David Barto	barto at network.ucsd.edu	barto at visionpro.com
>From a Marketing type:
    Don't give me any technical reason why something can't be done.
    If you really believed in the product you'd make it work.