Throttle questions
 

Hi,

What is the difference between prime.txt and local.txt?

I'm adding a Throttle=75 line to prime.txt as per undoc.txt but wondered if there's any reason I might add it to local? I'm running mprime/Prime95 on 6 different computers, and I only want the 75% throttling to occur on one of my machines.

Also - in undoc.txt it says that:

[QUOTE]Some options in prime.txt can be configured to have different values
at different times of the day using this syntax:
Option=setting
where setting is defined as
value OR value during list-of-times else setting
At present, only Memory, MaxHighMemWorkers, PauseWhileRunning, LowMemWhileRunni$
and PauseCheckInterval support this during/else syntax.[/QUOTE]

Is the support for this feature strictly limited to those variables as stated? It would be handy to have it for the Throttle feature too.

Thanks.

Twenty years ago, I thought it would be handy to able to copy your prime95 setup from one computer to another by copying everything in the directory except local.txt and worktodo.txt. Probably not the best design decision I've ever made.

Rather than using throttle, how about using fewer than all cores. If you have a quad core, using three cores would be 75%.

[QUOTE=Prime95;477548]Twenty years ago, I thought it would be handy to able to copy your prime95 setup from one computer to another by copying everything in the directory except local.txt and worktodo.txt. Probably not the best design decision I've ever made.

Rather than using throttle, how about using fewer than all cores. If you have a quad core, using three cores would be 75%.[/QUOTE]

That's true, thanks.

Altering the number of cores isn't something which can be set to vary automatically at different times of day, is it?

If you do use 3 cores, lock Prime95 to cores 1, 2 and 3. This allows your OS and apps to use the “primary” core 0.

(Your apps and stuff will still use cores 1, 2 and 3, but leaving core 0 open will improve your interaction (latency?) with the system. At least it does for our computer and workload.)

PS - Make sure your cores are real cores. We have an AMD 880K that is advertised as having 4 cores, but it shares some sort of math unit between pairs of cores, so in reality it is a dual core CPU.

[QUOTE=Xyzzy;477582]Make sure your cores are real cores. We have an AMD 880K that is advertised as having 4 cores, but it shares some sort of math unit between pairs of cores, so in reality it is a 2 core CPU.[/QUOTE]

I have an AMD A8-7650K Radeon R7, 10 Compute Cores 4C+6G. Any idea how I can check this?

[QUOTE=Xyzzy;477582]If you do use 3 cores, lock Prime95 to cores 1, 2 and 3. This allows your OS and apps to use the “primary” core 0.[/QUOTE]

Is this in the undoc.txt? I don't remember seeing it.

[QUOTE=lukerichards;477585]Is this in the undoc.txt? I don't remember seeing it.[/QUOTE]

On a 4 core CPU (assuming hyperthreading is disabled), the local.txt you'd have something like:

[CODE]
WorkerThreads=1
CoresPerTest=3

[Worker #1]
Affinity=1,2,3
[/CODE]

If you have hyperthreading enabled, your 4 cores will show up as 8, but you want to set your affinity to just the "physical" core. Windows alternates between physical/HT cores in the numbering:
[CODE]
WorkerThreads=1
CoresPerTest=3

[Worker #1]
Affinity=2,4,6
[/CODE]

I use "physical" core loosely... a hyperthreaded core has 2 pipelines (or whatever you'd like to call it) that only has one FPU. So while it may appear to be 2 cpus in the OS, the fact that there's only one FPU between them means there's really no benefit to LL tests which rely entirely on FP calculations.

[QUOTE=Madpoo;477633]On a 4 core CPU (assuming hyperthreading is disabled), the local.txt you'd have something like:

[CODE]
WorkerThreads=1
CoresPerTest=3

[Worker #1]
Affinity=1,2,3
[/CODE]

If you have hyperthreading enabled, your 4 cores will show up as 8, but you want to set your affinity to just the "physical" core. Windows alternates between physical/HT cores in the numbering:
[CODE];
WorkerThreads=1
CoresPerTest=3

[Worker #1]
Affinity=2,4,6
[/CODE]

I use "physical" core loosely... a hyperthreaded core has 2 pipelines (or whatever you'd like to call it) that only has one FPU. So while it may appear to be 2 cpus in the OS, the fact that there's only one FPU between them means there's really no benefit to LL tests which rely entirely on FP calculations.[/QUOTE]

[url]http://www.mersenneforum.org/showpost.php?p=477589&postcount=1[/url] user runs on lubuntu