SMP experts advice needed

[AlvaroBegue]

I used to fake similar errors (like "an hydrogen atom") whenever I had to write a very negative report on a paper I had to referee, to miguide the authors about my identity.

And I gather you are misspelling "misguide" on purpose to pretend to be from somewhere else, right?

[Don]

Are you French?

Yes, but what's that got to do with hyper-threading ?

Because of your name I had classified you in the back of my mind as German.

What gave it away was that you made a spelling mistake that is only natural to an English speaker with a heavy French accent, which does not pronounce the 'h': "an hyper-threading option".

I used to fake similar errors (like "an hydrogen atom") whenever I had to write a very negative report on a paper I had to referee, to miguide the authors about my identity.

So what is the deal with "an historic event" - I see his usage all the time and it bugs me.

Here is a snippet from a random article on the web that talks about good writing skills:

Summary

A historic is more common in both American and British English, but both usages are sufficiently common to be considered correct.

I guess being an anal programmer type I dislike arbitrary rules based on simply going with common use. I guess following this reasoning we will see "I aint got no sense" and other such phrases which ARE in common use these days become proper and correct.

[bob]

I'll trust your judgment

I never trust the judgement especially that of other people than me.

I have not done a full study, but I have many empirical anecdotes I found during my tests that show that testing with concurrency > 1 adds noise.

But the point is deeper than that. Unfortunately current ELO estimators do not take in account the different noise models: for bayes elo (but the others are the same) the info you can get from 1000 games is the same regardless if the games where hyper bullet 2" games ran on a computer while you were browsing and watching a movie at the same time or they were high quality, longer TC games on a dedicated machine.

The noise level is completely different in both cases but current ELO estimators, amazingly, do not take in account that and give the same LOS independently from the testing conditions (never understood why, probably because people working on them do not have a good background on information exchange over a noisy channel).

Depends on what you mean by "concurrency". If you mean an SMP search, of course it adds noise. But if you mean simultaneous games that are independent, that has been verified to be perfectly safe. I'd want to test this on a 6/8 and beyond cores per chip box, since there is a possible serious bottleneck. But on 1/2/4 cores per chip, with up to 8 cores per node, it works as reliable as just running one game per node...

[AlvaroBegue]

This seems like a good explanation:
http://oxforddictionaries.com/words/a-h ... oric-event

Summary: In the 18th and 19th centuries people often pronounced "historic" without the "h" sound (like we currently do with "honor", "hour", or "heir"). So perhaps "an historic event" is a remnant of those times.

[Sven]

I used to fake similar errors (like "an hydrogen atom") whenever I had to write a very negative report on a paper I had to referee, to miguide the authors about my identity.

And I gather you are misspelling "misguide" on purpose to pretend to be from somewhere else, right?

That is his keyboard, of course. You won't find many HGM posts where no letter is missing

Sven

[hgm]

Alas, missing letters is a much larger hazard than having extra letters. The keyboard of this laptop is really getting crappy. Especially the 'r' key is a poblem. Replacement of the laptop is long overdue... But I am so attached to it, that it is difficult to part. It feels like having to put down a pet...

Believe it or not, I have actually worn down the left shift key so much, that it now has a hole in it!

[syzygy]

I know that some of you are experts on SMP, and would like your advice on my testing methodology.

Are you asking about SMP and SMP implementations or about running test games concurrently? Those are two very different things.

Regarding your "N=8 CPU (i7)", this is most likely a 4-core CPU (with hyperthreading). It might make sense to run 8 test matches (i.e. 16 non-pondering single-threaded engines) in parallel, but running just 4 test matches (i.e. 8 non-pondering single-threaded engines) in parallel probably gives more reliable results. In any case, you shouldn't think of running 8 games concurrently with SMP enabled.

[ZirconiumX]

If in doubt, ask the guy who speaks it natively.

I always use a before h - although the sign in my city says 'Derby - An Historic City' - so go figure.

Back on topic, my strategy for testing is to have games in metric sh** tons - at ridiculously small TCs. Durandal is being tuned on games of 40/1 second + 0.05s per move - using four games in parallel. I'm thinking of switching to 16 games in parallel - 4 per core on a 2x Core 2 Duo box.

Matthew:out

[Don]

If in doubt, ask the guy who speaks it natively.

I always use a before h - although the sign in my city says 'Derby - An Historic City' - so go figure.

Back on topic, my strategy for testing is to have games in metric sh** tons - at ridiculously small TCs. Durandal is being tuned on games of 40/1 second + 0.05s per move - using four games in parallel. I'm thinking of switching to 16 games in parallel - 4 per core on a 2x Core 2 Duo box.

Matthew:out

Our system starts with fast time control games and progresses to longer time control games. We use some common sense because some changes have different scaling characteristics than others - so it happens that we will get a fantastic result at fast time controls and then it goes bad at long time controls.

Even if the result is not positive at the fast we will sometimes proceed if we have reason to believe the idea will scale. There are some types of changes that you are pretty much forced to test at long time controls and others that don't require it but we play it safe most of the time.