Prime95 30.8 (big P-1 changes, see post #551)

[chalsall]

Originally Posted by kruoli

Good news! Using the custom mprime build with hwloc 2.8.0, HWLOC_CPUKINDS_RANKING=coretype gave the correct architecture. HWLOC_CPUKINDS_MAXFREQ=adjust=0 still showed wrong results.

Please forgive me for this, but there is something to be said for experimentation. Useful to have multiple ways of measuring things. And, never implicitly trust your instruments.

I call it "Making friends with the Kit". 8^)

Two claim they are Jesus; one of them must be wrong. (Clearly Knopfler was not deep into cats...)

[chalsall]

Originally Posted by chalsall

(Clearly Knopfler was not deep into cats...)

A better recording.

Knopfler blows my mind. But then, that's not a difficult thing to do.

[Prime95]

Originally Posted by kruoli

Good news! Using the custom mprime build with hwloc 2.8.0, HWLOC_CPUKINDS_RANKING=coretype gave the correct architecture. HWLOC_CPUKINDS_MAXFREQ=adjust=0 still showed wrong results.

Is there C code to set the environment variable such that hwloc uses HWLOC_CPUKINDS_RANKING=coretype?

[kruoli]

Have you had a look at this? I would assume this has to be called before HWLOC get initialised.

[congsz]

One question: Is there any method that can let me to do P-1 stage 1 only, using v30.8b15? I want to perform stage 1 and stage 2 with the same exponent on different machines, so I wish there could be a "Stage 1 only" setup.

[James Heinrich]

Originally Posted by congsz

One question: Is there any method that can let me to do P-1 stage 1 only

You can manually specify bounds with Pminus1 worktodo lines instead of Pfactor:
Pminus1=1,2,<EXPONENT>,-1,<B1>,<B2>,<TF>

Or a lazy way would be in local.txt to set the Memory setting to low for most of the time and high for one minute, I haven't tried this but if you set the one-minute memory to the high amount of your Stage-2 machine then the bounds should be selected for the high-memory amount, but your stage-1 machine should (almost) never get to run it:
Memory=40960 during 0:00-0:01 else 128 (40GB for stage 2, but limit this stage1 machine to 128GB most of the day so it almost-never gets to run stage2).

Probably even better would be to set high memory limits on the stage1 machine (so the bounds get selected correctly) but then put LowMemWhileRunning=explorer in prime.txt so that the machine is always forced into low-memory mode (since explorer.exe is always running on Windows machines).

[lycorn]

I /think/ that using the Pminus1= syntax and choosing B1 = B2 will do.

[James Heinrich]

Originally Posted by lycorn

I /think/ that using the Pminus1= syntax and choosing B1 = B2 will do.

It will, but that assumes you know what B1 should be. PFactor conveniently auto-selects appropriate bounds for the exponent and available RAM. You can use that to see what bounds are auto-selected for a representative exponent and then use the same B1 bounds for similar exponents on the stage-1 machine. The stage2 machine can continue to use PFactor to auto-select bounds.
Naturally you'll need to copy the savefiles over from the stage1 machine to the stage2 machine before attempting stage2.

[congsz]

Originally Posted by James Heinrich

You can manually specify bounds with Pminus1 worktodo lines instead of Pfactor:
Pminus1=1,2,<EXPONENT>,-1,<B1>,<B2>,<TF>

Originally Posted by lycorn

I /think/ that using the Pminus1= syntax and choosing B1 = B2 will do.

This does not work. If I entered Pminus1 worktodo type with B1=B2, it would still say that B2=TBD at the beginning of Stage 1, instead of B2=B1.
I have been using Pminus1 with manually specified B1 and no Pfactor for 2 months.

Originally Posted by James Heinrich

Naturally you'll need to copy the savefiles over from the stage1 machine to the stage2 machine before attempting stage2.

Yes, this is exactly what I will do.

Originally Posted by James Heinrich

Or a lazy way would be in local.txt to set the Memory setting to low for most of the time and high for one minute, I haven't tried this but if you set the one-minute memory to the high amount of your Stage-2 machine then the bounds should be selected for the high-memory amount, but your stage-1 machine should (almost) never get to run it:
Memory=40960 during 0:00-0:01 else 128 (40GB for stage 2, but limit this stage1 machine to 128GB most of the day so it almost-never gets to run stage2).

Probably even better would be to set high memory limits on the stage1 machine (so the bounds get selected correctly) but then put LowMemWhileRunning=explorer in prime.txt so that the machine is always forced into low-memory mode (since explorer.exe is always running on Windows machines).

Thanks for advice, I will try it.

[axn]

Originally Posted by congsz

This does not work. If I entered Pminus1 worktodo type with B1=B2, it would still say that B2=TBD at the beginning of Stage 1, instead of B2=B1.
I have been using Pminus1 with manually specified B1 and no Pfactor for 2 months.

Get rid of the TF parameter.
For example Pminus1=1,2,947917,-1,50000000,50000000 is a current entry in my workodo and it does only the stage 1

[congsz]

Originally Posted by axn

Get rid of the TF parameter.
For example Pminus1=1,2,947917,-1,50000000,50000000 is a current entry in my workodo and it does only the stage 1

Done successfully. Thanks!