From: redelm@nospicedham.ev1.net.invalid
R.Wieser wrote in part:
> Robert,
>> You will need FSAVE/FRSTOR (and varients)
> Wolfgang gave that suggestion too. Alas, the F(N)SAVE resets
> the FPU stack, and for some reason I can't get the FXSAVE to work
> (my assembler shows its age by not knowing the opcode, and trying
> to use a "db 0Fh,0AEH, ....." sequence crashes the program).
It might have some safeguards against executing data :)
>> Your first FLD will clobber the stack top,
>
> I don't get that - why only the first one, and why would
> it clobber (the value at) the stack top ?
The stack is eight FP registers, any load pushes the one on
the top into the bit bucket. Actually, I believe the registers
are a circular file, and the load overwrites and decrements TOS.
>> As another poster has said, I don't think the x87 automagically
>> sets value flags
>
> I don't quite get this either. Value flags ? I'm reading the
> "Status Word" and in it look at the ST bits (at 11-13).
Aren't those three bits (0-7) the Top-of-Stack pointer?
People sometimes compare the FPSW with the x86 flags register.
It is not.
> Remark : I later found out/realized that the "Stack Top"
> is just the starting offset for the ST(x) arguments. IOW :
> whats in it isn't really relevant.
Exactly.
>> Dump and examine in main memory.
>
> :-) The problem was that I had no idea that I could or how to do that .
Well, debugging always requires more space. x86 assumes
sufficient stack space (or switches to priviliged memory).
34 years ago I wrote an extention to MS-DOS DEBUG.COM to
examine the x87. Converting binaryFP to decimal FP was hard.
> Ofcourse it didn't help that I got confused by (and by it focussed on)
> the "Stack Top" value. :-\
Well, quite forgivable. The x87 is focussed on the stack.
-- Robert
--- SoupGate-Win32 v1.05
* Origin: you cannot sedate... all the things you hate (1:229/2)
|
