Using Windows too.You don't get this?

No, just lines like this[CODE]Factoring 94 digits: 8772020681400041406139719532523827989347735295630997567630483178615199289699145983221184823767
Factoring 99 digits: 406446645948796774593047520388671951245474083870251239639663116682068731148323403166959701433763497
Factoring 110 digits: 16963768298673486259853213930114780568753753263992185490587800693417143230848722608478513597730619360083719679
*****
Factoring 110 digits: 72228390425379024300405517434120605282375707524211565380444859107430996912124117966407928585011537909152546561
Factoring 94 digits: 8577633976311037508351102280685805345589768882226505987751744262859358952824781170511355692019
Factoring 93 digits: 160795562913696562292163019122589479514888016229400671788152153623275827160823554361153233537
Factoring 106 digits: 2790938948167199797827175671693565714597806753511796615241396782206297849595331071389455254356554215512217[/CODE]Turned out i had an old yafu version in this directory. Using 1.32 i does work.

I have tweaked the params with
[QUOTE]open(YAFU, 'yafu ecm('.$composite.',200) -B1ecm 250000 -threads 4|')[/QUOTE]the first 15000 numbers are running - hmmm that`s 3.000.000 curves - help me :D

at the moment 56% are factored

btw nice script :)

This may not be very helpful overall. Other people factoring small numbers won't know which numbers have been ECMed, so will have to run ECM against every number before they switch to QS/SNFS/GNFS. Which wastes effort if it has already been ECMed.

It would be more helpful to concentrate on factoring the 1000 lowest numbers that yafu@home skips over. (It would be even more helpful if they didn't skip them.) Or generate Primo certificates for the smaller PRPs.

Just my opinions.

Chris (using SNFS to split the easier numbers in the first 10,000)

with 20 curves @ B1=100r3, I *check* about 25-30 65-108 digits numbers by minute. That's about 2 seconds of work by numbers. Even if one wanted to factorise every number, 2 second is a really small amount of time compared to a ecm-test &gnfs factorisation of a 105-110 digit number, right?

[QUOTE=firejuggler;308164]with 20 curves @ B1=100r3, I *check* about 25-30 65-108 digits numbers by minute[/QUOTE]
So checking 25-30 65-108 digits numbers is a minute of wasted time.
yafu@home, as well as "Other people factoring small numbers" have to do ecm anyway.

Just switched off all my helpers that fully factored "small numbers" for a very long time, now focusing on aliquotS <115digits in size.

During this minute, I get around 13% of factored composite. Meaning around 3 or 4 composite factored by minute.
I don't spend much time on each composite, but I use a sledgehammer.
If it break, it's good, if not, too bad. I hit the next composite, in the hope it break quickly.
pratical case : 8635 composite tested ,1070 factored
best scenario : 30 tested by min, 288 minutes , 4H and 48 minutes
worst scenario: 25 tested by min, 346 minutes, 5H 45 minutes.

If I tweak the test a bit more, I 'll get more factored . But the goal here is to hit fast, not accuratelly, nor persistently.

This looks like people have gone crazy running curves. My intention with writing this script was to run a tiny bit of ecm on each number such that there isn't much work duplicated if someone runs a full ecm on this number. Some people here seem to be running a large part of t30.
I personally like the fact that by doing curves at 25 digit level I can occassionally find 30-35 digit factors. It might be worth running a worker with P-1 for a while.

[QUOTE=henryzz;308562]This looks like people have gone crazy running curves.[/QUOTE]Using Dario's applet, I knocked off some of the lowest composites with no known factors.

If anyone is spending a lot of time doing this then it is probably better to pick composites at random and factor them completely.

[QUOTE=lorgix;308624]If anyone is spending a lot of time doing this then it is probably better to pick composites at random and factor them completely.[/QUOTE]
Exactly. I was just frustrated with having >30% of small composites factorable within 10 seconds.