Links to Precompiled GMP-ECM versions

[Dubslow]

Originally Posted by pinhodecarlos

I'm using your version but do you have now an early one?

What do you mean early one? 7.0-dev perhaps?

[pinhodecarlos]

Originally Posted by Dubslow

What do you mean early one? 7.0-dev perhaps?

Yes.

[Dubslow]

Okay then.

Attached is a GMP-ECM 7.0-dev, SVN r2345 binary for GNU-Linux (x86-64), compiled with gcc 4.7.2 and statically linked against GMP 5.1.0. Documentation included.

[pinhodecarlos]

Originally Posted by Dubslow

Okay then.

Attached is a GMP-ECM 7.0-dev, SVN r2345 binary for GNU-Linux (x86-64), compiled with gcc 4.7.2 and statically linked against GMP 5.1.0. Documentation included.

Thank you.

Carlos

[RichD]

Originally Posted by Dubslow

Attached is a GMP-ECM 7.0-dev, SVN r2345 binary for GNU-Linux (x86-64), compiled with gcc 4.7.2 and statically linked against GMP 5.1.0. Documentation included.

I don't think Bill is around these parts but could anyone else build a Linux x86_64 build from the latest SVN which has several fixes in the past year?

[pinhodecarlos]

Originally Posted by RichD

I don't think Bill is around these parts but could anyone else build a Linux x86_64 build from the latest SVN which has several fixes in the past year?

I am also interested. Thank you in advance.

[ChristianB]

Hi,

I'm currently investigating new applications for yoyo@home. We want to update the server code of yoyo@home in the future and this would give us the possibility to run optimized applications for specific hardware. I already set up a test project for this with the yoyo@home ecm application.

With my next step I would like to update ecm to a current version and also support optimized versions (if available) since a cuda version seems to be not an option.

Could someone provide me with an optimized version? Especially a recent windows one would be nice as I don't have a windows build system. I already have a recent stock gmp-ecm for Linux 64 bit.

[xilman]

Originally Posted by ChristianB

With my next step I would like to update ecm to a current version and also support optimized versions (if available) since a cuda version seems to be not an option.

That may be the case; it may not be. On hardware that supports it the CUDA version tends to be much more efficient than a pure cpu solution. I see an improvement of at least twice on a relatively lowly 460 compared with a 6-core AMD box clocked at 3.2GHz.

If I remember correctly what Cedric told me, the CC>=2.0 limitation is to do with atomic operations that could be implemented (at lower efficiency of course) in other ways on older hardware. As run-time selection of the correct code for different compute capabilities is straightforward, it shouldn't be too hard to port GMP-ECM to cc<2.0 cards.

It is the case that only stage 1 can presently run under CUDA. I've already been thinking about how to incorporate this behaviour into the ECMNET framework.

We (you, me, Cedric and anyone else seriously interested) should probably take this to email.


Paul

[ChristianB]

Originally Posted by xilman

If I remember correctly what Cedric told me, the CC>=2.0 limitation is to do with atomic operations that could be implemented (at lower efficiency of course) in other ways on older hardware. As run-time selection of the correct code for different compute capabilities is straightforward, it shouldn't be too hard to port GMP-ECM to cc<2.0 cards.

Porting to CC<2.0 cards seems to be too much effort for what I want to do. I successfully compiled binaries for ecm-cuda for the three CC classes. I will switch my GPU for one with CC=2.1 and see if it works (for CC<=2.1). I still have to work on the possible Pipeline here. As only stage 1 is done on the GPU I then have to distribute the stage 2 work as a normal CPU app which must be a different app than the one doing stage 1 and stage 2 on CPU (because of BOINC limitations).

It would be easier if ecm could call the BOINC API directly, I will investigate this a little more I think.

[kracker]

I noticed a lot of binaries floating around here for Windows are linked with MPFR. Is there a reason? (vs GMP)

[wombatman]

I can't speak for anyone else, but it's much simpler for me to compile with MPIR in Visual Studio on Windows. I also can't compile the GPU-enabled version in MinGW, where GMP makes more sense.