VC8 build

[Vikas N Kumar]

I know this issue with MS Compilers.
MS by default sets the maximum floating point size to 64-bits. I have
faced this problem while matching real number results on my code on
Linux vs code running on Windows. ( I was porting Windows code to
Linux.)
Hence long double gets set to double on Windows.
They by default use the FP control word option. I think one can reset
it manually in their program by calling the appropriate assembly
instrution.
So you might have to override the MS option. I haven't tried it.
Instead what I had to do was set the maximum floating point usage size
to 64-bit on Linux to get the client happy with the results.

Vikas

On 10/3/06, Charles Bailey <yasm at hashpling.org> wrote:
>
> Peter Johnson wrote:
> > On Mon, 2 Oct 2006, Charles Bailey wrote:
> >> I've recently got back into assembler programming.  I used to dabble a
> >> bit in ARM stuff, but now I'm interested in doing some stuff with the
> >> 387 FPU instructions, principally as you can't get access to 80-bit
> >> extended precision from the VC8 C compiler (as far as I can tell).
> >
> > It's been a long while since I looked (probably back in DOS days with
> > Borland compilers), but I seem to recall that "long double" is the magic
> > incantation required to use 80-bit floating point.
> >
>
> And on gcc for linux too; sadly on MS compilers we get this:
>
> Type Name     Bytes            Other Names   Range of Values
> double        8                none          1.7E +/- 308 (15 digits)
> long double   same as double   none          same as double
>
> Anyway, I've sent a patch (not to the list) for the other yasm VC8 build
> file issues which I mentioned.
>
> Thanks,
>
> Charles.
>
> _______________________________________________
> yasm-devel mailing list
> yasm-devel at tortall.net
> http://cvs.tortall.net/mailman/listinfo/yasm-devel
>