[Beowulf] Question about amd64 architecture and floating pointoperations

[Vincent Diepeveen]

Well it might be even worse, assuming you're toying in a higher level 
language than assembly;
in that case the compiler might optimize your code to SIMD code in which 
case you no longer have
80 bits long doubles, but 64 bits doubles, and 52 bits mantissa; which 
exactly is the idea of SIMD.

80 bits really is not a very useful size at all (though i realize you can do 
a few tricks with it which can give you up to 62 bits significance in 64 
bits integers, but that's for a few exceptionnel nerds who try to get the 
maximum out of something, which doesn't describe 99.99% of the programming 
population).

Vincent
----- Original Message ----- 
From: "Mark Hahn" <hahn at physics.mcmaster.ca>
To: "Vincent Diepeveen" <diep at xs4all.nl>
Cc: <beowulf at beowulf.org>
Sent: Friday, November 24, 2006 7:25 PM
Subject: Re: [Beowulf] Question about amd64 architecture and floating 
pointoperations


>> Hopefully your 80 bits logics code is not critical to anything.
>> I wouldn't count at keeping the entire 62 bits (?) mantissa.
>>
>> Context switch and dang it's gone.
>
> why do you think that?
>