apgsearch v3.1

[apg]

Is there any limit to the number of soups? I've been running apgsearch for Day & Night [b3678s34678].
The haul of 10000 soups is uploaded, but the haul of 100000000 soups is not.

Not number of soups, but rather file-size: there's a 1-megabyte limit on all uploaded hauls.

[Apple Bottom]

Is there any limit to the number of soups? I've been running apgsearch for Day & Night [b3678s34678].
The haul of 10000 soups is uploaded, but the haul of 100,000,000 soups is not.

There's a haul size limit of 1 MiB. Day & Night generates very large hauls; the practical limit for C1 is ~8,000,000 soups, and much less for the higher symmetries.

This is also why, when investigating rules you've not investigated before, it's generally a good idea to do test runs to gauge the size of the generated hauls (and the search speed, obviously). When choosing the final haul size I generally aim for a time of about 2 hours between hauls, subject of course to the size additional constraints.

Speaking of which, there's also some minimums imposed: hauls that reach neither 10k soups nor 250k objects are also rejected. Combined with the 1 MiB size limit, this renders some rules unsearchable on Catagolue.

[David]

My processor is Intel Core i7-6700, it has 4 cores and 8 threads.
How many threads should I parallelize Apgmera on, "-p 4" or "-p 8"?

[Apple Bottom]

My processor is Intel Core i7-6700, it has 4 cores and 8 threads.
How many threads should I parallelize Apgmera on, "-p 4" or "-p 8"?

Eight will eke out some extra performance -- not quite double, but enough to be worth it. On my Core i7-3770K, I get ~1,450 soups/sec*instance if running eight instances (11,600 soups/sec total), and ~2,200 soups/sec*instance if running four (8,800 soups/sec total).

[Apple Bottom]

Interesting things seem to be afoot. A new symmetry, SS, showed up, with hauls being generated by a tool called "hs", and sample soups looking like this:

Code: Select all

#C -1E 5E 15E 13O 3E -15O -1E 49O -15O 5O 11O -9E -15O
x = 1792, y = 1792, rule = B3/S23
boo$obbo$boo9$
10b3o$10bo$11bo126$
141b3o$141bo$142bo126$
274b3o$274bo$275bo126$
402b2o$401b2o$403bo126$
524b3o$524bo$525bo126$
644b2o$643b2o$645bo126$
778b3o$778bo$779bo126$
932b2o$931b2o$933bo126$
1028b2o$1027b2o$1029bo126$
1166b2o$1165b2o$1167bo126$
1297b2o$1296b2o$1298bo126$
1414b3o$1414bo$1415bo126$
1540b2o$1539b2o$1541bo126$!

These seem to be based on firing slow glider salvos (hence also "SS") at distant objects and seeing what evolves. Can't wait to hear more!

[muzik]

Here's another one, not mine obviously:

Code: Select all

x = 1792, y = 1792, rule = B3/S23
oo$oo9$
14b3o$14bo$15bo126$
149b3o$149bo$150bo126$
270b3o$270bo$271bo126$
415b3o$415bo$416bo126$
526b2o$525b2o$527bo126$
655b2o$654b2o$656bo126$
784b2o$783b2o$785bo126$
914b3o$914bo$915bo126$
1036b2o$1035b2o$1037bo126$
1184b3o$1184bo$1185bo126$
1300b2o$1299b2o$1301bo126$
1407b2o$1406b2o$1408bo126$
1557b3o$1557bo$1558bo126$!

[BlinkerSpawn]

I'm pretty sure somebody just modified an existing salvo-search script to upload found objects to Catagolue (a pretty economical move considering it's already a well-fleshed-out way to store data about SL occurences) and the results are showing up in the tabulations.

[Apple Bottom]

I'm pretty sure somebody just modified an existing salvo-search script to upload found objects to Catagolue (a pretty economical move considering it's already a well-fleshed-out way to store data about SL occurences) and the results are showing up in the tabulations.

That somebody being Calcyman himself, mind -- and there was some server-side support needed, too, e.g. for sample soup (re)construction.

Hope we'll get a code drop for the new tool some time!

[Apple Bottom]

BTW, I noticed a discrepancy in the figures for the B3/S23/SS stats. The census currently says:

The following census was compiled from 926 committed hauls containing 456 028 192 614 objects generated from 3 732 136 786 soups.

However, the textcensus contains fewer objects:

Code: Select all

$ cat b3s23.SS.txt | perl -nE '@a=split /,/; ($b=$a[1])=~s/\"//g; $sum+=$b; END {say $sum}'
455951534486
$ 

[Apple Bottom]

Apologies for triple-posting, but I noticed that Catagolue's reached half a million submitted hauls now. This one (submitted by benetnasch85 and verified by goldtiger997) is the 500,000th one, if I'm not mistaken. Congratulations!

[muzik]

Professor and cloverleaf interchange still aren't named on Catagolue yet.

[benetnasch85]

Hear ye, hear ye! Apple Bottom has contributed 10 trillion objects to Catagolue!

[Apple Bottom]

Hear ye, hear ye! Apple Bottom has contributed 10 trillion objects to Catagolue!

Aww shucks, sugarcube. 'taint nothin'.

[muzik]

And yet those two still lifes still aren't named...

[drc]

What if we made a version of apgsearch that searches big sparse life soups (something like 1000x1000)? It sounds really dumb but it's merely a suggestion.

[muzik]

What if we made a version of apgsearch that searches big sparse life soups (something like 1000x1000)? It sounds really dumb but it's merely a suggestion.

Seems interesting.

[fluffykitty]

How sparse?

[Rhombic]

How sparse?

Sparse enough. Beyond a certain limit, lowering the probability of a live cell doesn't alter the distribution of objects; in fact, it just has more blank space in between them.

[fluffykitty]

Catagolue bug:
Link to page

[apg]

Catagolue bug:
Link to page

Are you referring to the z-order of the bars in the chart? That's beyond my jurisdiction; you should be complaining to the creators of Highcharts instead.

[fluffykitty]

Okay. On an unrelated note, we only got about 45B objects yesterday. When will the new apgsearch release bring us back to 1T/day?

[apg]

Okay. On an unrelated note, we only got about 45B objects yesterday. When will the new apgsearch release bring us back to 1T/day?

The 1T/day was when I was running apgmera on roughly 150 or 200 cores in total. The easiest way to get back up to that is for me to resume deploying on many cores -- there isn't really any possibility of attaining a significant speedup just using CPUs.

[fluffykitty]

How did you get 200 cores to run apgmera?

[fluffykitty]

How do I check the lifetime contribution of someone to B3S23C1?

[apg]

How did you get 200 cores to run apgmera?

By having ssh access to about 60 machines of varying sizes (mostly quad-core machines).

How do I check the lifetime contribution of someone to B3S23C1?

View the HTML source of https://catagolue.appspot.com/census/b3s23/C1/