MPIR FFT Failure

[Cactus]

There is an FFT assertion failure reported here:

http://www.mersenneforum.org/showthread.php?t=15327

on the Windows x64 version of MPIR. It seems to be a very rare issue
but I wonder if there is anything that needs to be done about it.

It seems that there might also be some issues that arise when a lot of
memory is being used.

Brian

[Bill Hart]

I don't know. It looks suspiciously like an actual bug to me.

> --
> You received this message because you are subscribed to the Google Groups "mpir-devel" group.
> To post to this group, send email to mpir-...@googlegroups.com.
> To unsubscribe from this group, send email to mpir-devel+...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en.
>
>

[jason]

On Mar 4, 5:01 pm, Bill Hart <goodwillh...@googlemail.com> wrote:
> I don't know. It looks suspiciously like an actual bug to me.
>

> On 4 March 2011 10:09, Cactus <rieman...@gmail.com> wrote:
>
> > There is an FFT assertion failure reported here:
>
> >http://www.mersenneforum.org/showthread.php?t=15327
>
> > on the Windows x64 version of MPIR. It seems to be a very rare issue
> > but I wonder if there is anything that needs to be done about it.
>
> > It seems that there might also be some issues that arise when a lot of
> > memory is being used.
>
> > Brian
>
> > --
> > You received this message because you are subscribed to the Google Groups "mpir-devel" group.
> > To post to this group, send email to mpir-...@googlegroups.com.
> > To unsubscribe from this group, send email to mpir-devel+...@googlegroups.com.
> > For more options, visit this group athttp://groups.google.com/group/mpir-devel?hl=en.

Hi

I assumed this thread was connected with the Valgrind FFT error ,
clearly its not , I can build ECM and mpir-2.* under the MinGW64
"latest" if you want some binarys to test , just give me the exact
incantation used (and cpu type , as I assume you dont use a fat build,
and I will have to force ECM and mpir to "detect" that cpu)

Jason

[Cactus]

Hi Jason

This was a failure reported by someone on MersenneForum, not one I
haave experienced myself so I have no idea of the ccircumstances
needed to trigger it.

I can consistently invoke a memory allocation fault when using
nedmalloc as a custom allocator but I have been unable to localise
this as I am not very familiar with the nedmalloc debugging features.

Brian

[jason]

Hi Jason

Brian

I wonder if your nedmalloc failure is the same as the valgrind error , do
you try deleting the SSE copy.asm files?

Jason

[Cactus]

> For more options, visit this group athttp://groups.google.com/group/mpir-devel?hl=en.

>
> I wonder if your nedmalloc failure is the same as the valgrind error , do
> you try deleting the SSE copy.asm files?
>
> Jason

It fails with the generic C version so its not an assembler issue.

Brian

[jason]

Brian

nedmalloc runs on linux by the look of it , I'll give it a go

Jason

[Cactus]

> For more options, visit this group athttp://groups.google.com/group/mpir-devel?hl=en.

>
> nedmalloc runs on linux by the look of it , I'll give it a go
>
> Jason

I am using it in a modified version of bench_two. The original
version has a lot of memory leaks so I will have to put a revised one
together if you are willing to try it.

Brian