apparent bug with cmucl compiles to sse2



Raymond Toy wrote:

Thank you for your interest, i see there is no way to recompile a particular
fasl as i feared.

> 
> Anyway, I ran your test after compiling eigens-by-jacobi.lisp for sse2.
> I get pretty much the same plot there as I do on a Sun sparc machine.
> 
> Can you tell me what the smallest and largest eigenvalues you get?   For
> sse2, I get:
>  - 2.045740417937529 and 2.876115587743105
> 
> These are exactly the same values on sparc.

I get exactly these same values  on my laptop which doesn't have .sse2
binaries, but .xf86 ones.
I have also computed with bfloats (fppprec:80) which takes 5mn
instead of 3s, but gives the same result up to 15 significative figures.

So i think these are correct. But yesterday, on my colleague macbook,
the extreme values where like -4 and +6, while values in between looked
correct.

There is another potential problem, the plot produced has a "glitch" around
80 (horizontal axis). I certainly get that here, even computing with
bfloats. My colleague has just sent me the graphs computed with maxima and 
mathematica, and the last one doesn't have any glitch. I have put a copy of
the graphs here:
http://www.lpthe.jussieu.fr/~talon/Spin_50.pdf
http://www.lpthe.jussieu.fr/~talon/spin50_Mathematica.pdf

so maybe there is another problem somewhere. Anyways, it is wonderful how
the eigenvalues are regularly spaced for this matrix,this is very unusual.

> 
> What version of cmucl are you using?  I did the test using 2009-01.
> 
> Ray

It was cmucl-2009-02-x86-darwin.tar.bz2   

-- 
Michel Talon