On 04/03/2017 20:57, kesse...@schema.de wrote:

> first of all, thanks for all the great MPIR work! I've been using it for 
> about 4 years to compute visually compelling deep Mandelbrot zoom videos.
> 
> Yesterday I've downloaded 3.0.0 and compiled it using VS 2015 U3 on an Intel 
> Core i7 6900K (8 cores, Broadwell) on Windows 10 64.
> 
> Unfortunately, 3.0.0 seems to be slower than 2.7.2 by about 8% when small 
> floats are used. By small floats I mean a precision of up to 256 bits (4 
> limbs on x64).
> 
> Compilation worked flawlessly for all of the 10 architectures I've selected. 
> Just to make sure Visual Studio updates are not the source of the problem, I 
> also recompiled the 7 architectures I've been testing with 2.7.2.

[snip details]

Hi Marcus,

As the maintainer of the Windows builds for MPIR, I would like to thank
you for your work on this issue.  Even though it was not the focus of
your email, I much appreciate your confirmation that the Windows build
process is working as intended.

Turning to the performance issue, i don't have much to add to what you
and Bill have said.  My first thought was that tuning has been messed up
but it cannot be this since the tuning files for the two versions are
identical.

As Bill has said, tuning on Windows is not very stable so we simply use
the tuning values for the *nix/GCC versions.  But this too cannot
explain the difference as it also applies to both versions.

One thing that I would like to see, if possible, is the result of
running the MPIR benchmark code on your system for the two MPIR
versions.  The benchmark covers various internal routines and might show
us whether particular routines are at fault.  It would also be a useful
second look at the issue.

   with my regards,

     Brian

-- 
You received this message because you are subscribed to the Google Groups 
"mpir-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to mpir-devel+unsubscr...@googlegroups.com.
To post to this group, send email to mpir-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/mpir-devel.
For more options, visit https://groups.google.com/d/optout.

Reply via email to