LLR Version 4.0.1 released.

[Jean Penné]

Hi All,

I uploaded today the version 4.0.1 of the LLR program.
You can find it now on my personal site :

http://jpenne.free.fr/

The 32bit Windows and Linux compressed binaries are available as usual.
The Linux 64bit binaries are released here, and also the Mac OS 64bit binaries.
The Mac OS 32bit is not released here because I have not the 32bit hwloc library which is needed, and could not build it on my Mac mini...
I uploaded also the complete source in a compressed file ; it may be used to build the 64bit Windows binaries.

What is new in this version :

In previous Version 4.0.0, one call to free() function was missing in Gerbicz error checking code ;
this caused an important memory leak... This issue is now fixed here!

This version is linked with the last version 30.6 of George Woltman's gwnum library.
No really new feature, from 3.8.24, but some improvements related to reliability and speed.
I avoid now the use of giants functions invg() and gcdg() which are slow and seem not to be very reliable.
To do that, I am using gtompz() and mpztog() conversion functions.
Also, I replaced everywhere the gwnum squaring and multiplication functions gwsquare() and gwmul() by their new forms :
gwsquare2(), gwmul3(), gwmul3_carefully(), etc...
As usual, I need help to build the 64bit Windows binaries.
I uploaded also the GNU gmp6.1.0 compressed source I used on 32bit VC6.0
I hope it can be used to build this library on Windows 64bit and link it with LLR...
Please, inform me if you encountered any problem while using this new version.
Best Regards,
Jean

[kruoli]

Finally, I was able to compile a 64 bit Windows command line version that supports AVX-512.
One of the workarounds I had to do was to comment av512case0: ;read8 4096, 16384 ; Read 4KB (so jumping back immediately) in timeit.asm, because there was a problem with UASM saying it could not build an EVEX from this for some reason. I wanted to mention this specifically since I am not fully sure for what this is used; a few test runs on AVX512 numbers went smoothly.

[Prime95]

Originally Posted by kruoli

in timeit.asm, because there was a problem with UASM saying it could not build an EVEX from this for some reason. I wanted to mention this specifically since I am not fully sure for what this is used;.

timeit.asm is not used at all

[kruoli]

For some reason, the build is around 15 % slower on the same exponent when comparing to version 3.8.23 as the 32 bit build but running an LLR test instead of PRP.

[rebirther]

Originally Posted by kruoli

For some reason, the build is around 15 % slower on the same exponent when comparing to version 3.8.23 as the 32 bit build but running an LLR test instead of PRP.

All versions above 3.8.23 are slower, tested it last time with 64bit 3.8.24

[henryzz]

Originally Posted by kruoli

Finally, I was able to compile a 64 bit Windows command line version that supports AVX-512.
One of the workarounds I had to do was to comment av512case0: ;read8 4096, 16384 ; Read 4KB (so jumping back immediately) in timeit.asm, because there was a problem with UASM saying it could not build an EVEX from this for some reason. I wanted to mention this specifically since I am not fully sure for what this is used; a few test runs on AVX512 numbers went smoothly.

Congrats on getting this to compile. I have tried 2-3 times and failed. Is there any chance you could write up exactly how you did it so that others can do it in the future?

Originally Posted by kruoli

For some reason, the build is around 15 % slower on the same exponent when comparing to version 3.8.23 as the 32 bit build but running an LLR test instead of PRP.

If you force a LLR test using this version(ErrorChecking=0 in the ini file) then the speed on my system matches 3.8.23 64 bit. I think the speed difference is LLR vs PRP.

[rogue]

Jean, how can I use llr so that will run on a designated CPU? I thought that it supported Affinity= in the llr.ini, but that is apparently not the case, at least with Windows 10.