mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Computer Science & Computational Number Theory (https://www.mersenneforum.org/forumdisplay.php?f=116)
-   -   Wilson-prime search practicalities (https://www.mersenneforum.org/showthread.php?t=16028)

 fivemack 2011-09-11 16:47

I ran 111G-112G to get a speed and memory use measurement for interval size 1e7 for wilsontest compiled against gmp-5.0, and then 112G-113G to get the same measurement for wilsontest2.c compiled against mpir-2.4.0)

 Jeff Gilchrist 2011-09-11 17:18

[QUOTE=fivemack;271462]I ran 111G-112G to get a speed and memory use measurement for interval size 1e7 for wilsontest compiled against gmp-5.0, and then 112G-113G to get the same measurement for wilsontest2.c compiled against mpir-2.4.0)[/QUOTE]

And what did you find for speed and memory difference? I have seen a big difference in speed compared to gmp5 and mpir.

Jeff.

 fivemack 2011-09-11 18:02

ah yes, results often come in useful

111e9 .. 112e9 interval 1e7
[code]
Testing p=111000000049...111508672567, p==1 mod 3, time=343 sec.
Testing p=111000000029...111999999989, p==5 mod 12, time=19874 sec.
Testing p=111000000071...111999999983, p==11 mod 12, time=55544 sec.
Testing p=111508672579...111999999961, p==1 mod 3, time=124224 sec.
Done the st=111000000000-en=112000000000 interval. Time=143094 sec.
126835.38user 1555.64system 39:44:53elapsed 89%CPU (0avgtext+0avgdata 4027728maxresident)k
[/code]

112e9..113e9, new code + mpir, interval 1e7
[code]
Testing p=112000000289...112999999877, p==5 mod 12, time=237 sec.
Testing p=112000000003...112508982601, p==1 mod 3, time=25044 sec.
Testing p=112000000031...112999999859, p==11 mod 12, time=42104 sec.
Testing p=112508982637...112999999951, p==1 mod 3, time=95729 sec.
Done the st=112000000000-en=113000000000 interval. Time=112854 sec.
99689.07user 2899.91system 31:20:53elapsed 90%CPU (0avgtext+0avgdata 4067952maxresident)k
[/code]

So the new code runs in 80% of the time and uses essentially the same amount of memory. Note that the 'maxresident' figures are overstated by precisely a factor four - actually both processes use about a gigabyte.

 Jeff Gilchrist 2011-09-11 18:18

I will also reserve 144e9 to 160e9.

 MrRepunit 2011-09-12 07:30

Finished my range 1E10 to 5E10.
Found a few more near Wilson primes:

35525054743 -1+26p
24664241321 -1+46p
28737804211 -1-58p
44034466379 -1+39p

Also reserving 16E10 to 20E10.

 Jeff Gilchrist 2011-09-12 13:51

Thanks rajula for maintaining your reservation page, it is very helpful, and the new green blocks for completed work make it easier to read. You are fast too, sometimes the pages are updated almost immediately. :smile:

Now looking at an i5 based system comparing a 1e9 range to a 4e9 range.

1e9 range = 54219 sec
4e9 range = 189560 sec (47390 sec/1e9 range)

Some savings.

Jeff.

 rajula 2011-09-12 15:09

[QUOTE=Jeff Gilchrist;271534]Thanks rajula for maintaining your reservation page, it is very helpful, and the new green blocks for completed work make it easier to read. You are fast too, sometimes the pages are updated almost immediately. :smile:
[/QUOTE]

You are welcome. I am happy to maintain the list and I try to make updates as fast as I can. I do check the forums quite frequently... Sometimes I am away for a week or so, so expect some delay from time to time. :smile:

 Jeff Gilchrist 2011-09-13 00:12

I just finished my 80e9 to 100e9 range with nothing new that hasn't already been reported.

I will now reserve 200e9 to 250e9.

Jeff.

 fivemack 2011-09-13 08:24

120G to 144G done, nothing of interest found.

 Jeff Gilchrist 2011-09-16 17:57

113e9 to 120e9 finished, nothing found.

 Jeff Gilchrist 2011-09-17 10:20

144e9 to 160e9 finished, nothing found.

All times are UTC. The time now is 16:13.