Brent tables reservations - Page 18

R.D. Silverman · 2013-11-14, 19:16

Quote:

Originally Posted by jcrombie

You can sieve on two sides -- the 'algebraic' side and the 'rational' side. If your .poly file has "type: snfs" then the factMsieve.pl script will sieve on the 'rational' side. If your .poly file has "type: gnfs" then it sieves on the 'algebraic' side. I believe the .py script is just a straight port from the perl script, so this should also apply there.

<snip>

I hate it when someone who is ignorant of the subject insists on posting
misinformation.

NFS sieves on BOTH SIDES. Didn't you read my prior post on this subject?
You can read, can't you???? Within this very thread I already said that it sieves on BOTH sides.

The only question is: On which side do we sieve with respect to the full
lattice and on which side do we sieve with respect to the special-q
sub-lattice? But both sides get sieved.

Quote:

Hope that helps a bit

.

About as much as a toothache.

bsquared · 2013-11-14, 19:44

Quote:

Originally Posted by R.D. Silverman

The only question is: On which side do we sieve with respect to the full
lattice and on which side do we sieve with respect to the special-q
sub-lattice? But both sides get sieved.

You are talking about different things. The OP asked about how to most effectively use the tools and you answered with an introduction to how to implement a sieve. That is fine, and the resulting discussion is good, but I see no need to jump on someone's case when they are trying to answer the OP's tool usage question.

R.D. Silverman · 2013-11-14, 22:05

Quote:

Originally Posted by bsquared

You are talking about different things. The OP asked about how to most effectively use the tools and you answered with an introduction to how to implement a sieve. That is fine, and the resulting discussion is good, but I see no need to jump on someone's case when they are trying to answer the OP's tool usage question.

You need to learn to read as well. The OP asked about how
small to make the starting value for special_q. I replied
with what I do. The OP then said he did not understand my reply. (#176)
The rest followed. The OP's first post showed that he knew how to
use the tool to set the starting q. The question was about its value.

The reply I criticized was NOT about tool usage. It contained
a blatantly false assertion about sieving on just one side.

Such assertions make my job much harder, because I have
to spend time to correct the misinformation. And this especially
applies when the misinformation directly contradicts something that
I said just a few posts earlier.

YuL · 2013-11-14, 22:14

Quote:

Originally Posted by R.D. Silverman

huh? One allocates space for the table. If there are more primes
to store than space allows one gets an overflow. Why is this a
mystery? One can also get an "overflow" as the table starts to get full.
If a collision occurs it has to be resolved. There are a variety of ways
to do this. But once the table starts to get full one spends TOO MUCH
TIME resolving the *()!#&*@#&* collisions. It isn't worth it.

Yes. I could re-code to dynamically allocate more space for the table
as it starts to get full. I never thought it worth the effort.

OK, now I see what you mean.

Quote:

Originally Posted by R.D. Silverman

Perhaps you might want to read "Development of the Number Field Sieve",
by Lenstra et.al.?

I read about that book (in this forum) I'm looking for it.

Quote:

Originally Posted by jcrombie

...
If your .poly file has "type: snfs" then the factMsieve.pl script will sieve on the 'rational' side. If your .poly file has "type: gnfs" then it sieves on the 'algebraic' side. I believe the .py script is just a straight port from the perl script, so this should also apply there.
...

No. It goes like this:

Code:

  # factmsieve.py line 1672
  siever_option = '-r' if lats_p['lss'] else '-a'
  siever_side = 'rational' if lats_p['lss'] else 'algebraic'

If lss is set in the poly file => 'rational', otherwise 'algebraic'.
However I have no idea what 'lss' is.

A lot of things were said in the previous posts thank you all,
I think I have to do my homework now :)
I understand the theory behind NFS is far from being easy, but I'm willing
to learn.
However on the practical side I think there is a lack of information
for someone who would like to start digging into the details
instead of just running the script (which is pretty easy thanks to this
excellent page).

bsquared · 2013-11-14, 22:57

Quote:

Originally Posted by R.D. Silverman

The reply I criticized was NOT about tool usage. It contained
a blatantly false assertion about sieving on just one side.

Where did he say "only one side"? I believe it is colloquial to say "algebraic side" when one means "special-q on the algebraic side". Nowhere can I see him say that only one side is sieved.

R.D. Silverman · 2013-11-14, 23:04

Quote:

Originally Posted by bsquared

Where did he say "only one side"? I believe it is colloquial to say "algebraic side" when one means "special-q on the algebraic side". Nowhere can I see him say that only one side is sieved.

I quote:

"You can sieve on two sides -- the 'algebraic' side and the 'rational' side. If your .poly file has "type: snfs" then the factMsieve.pl script will sieve on the 'rational' side. If your .poly file has "type: gnfs" then it sieves on the 'algebraic' side."

This is clear. The post says that it sieves on one side depending on
how the type is set.

This is wrong. It sieves on both sides. "applying special_q" does NOT
equal "sieve".

jcrombie · 2013-11-15, 04:52

Quote:

Originally Posted by R.D. Silverman

I quote:
This is wrong. It sieves on both sides. "applying special_q" does NOT
equal "sieve".

Thanks for pointing that out. I'll update my language from now on.

jcrombie · 2013-11-16, 03:00

Quote:

Originally Posted by YuL

No. It goes like this:

Code:

  # factmsieve.py line 1672
  siever_option = '-r' if lats_p['lss'] else '-a'
  siever_side = 'rational' if lats_p['lss'] else 'algebraic'

If lss is set in the poly file => 'rational', otherwise 'algebraic'.
However I have no idea what 'lss' is.

My guess: lss = 'lattice sieve side'.

Look around a bit and you will see that it does actually mimic the perl script with regards to "type: snfs" vs. "type: gnfs". (It also apparently has an extra feature that allows the user to override the side

)

Quote:

Originally Posted by YuL

However on the practical side I think there is a lack of information
for someone who would like to start digging into the details
instead of just running the script (which is pretty easy thanks to this
excellent page).

Yes, sad but true.

YuL · 2013-11-16, 13:40

Quote:

Originally Posted by jcrombie

Yes, sad but true.

There is a tex document in subfolder ggnfs-doc of ggnfs written by Chris Monico,
I converted it to pdf, it is a very interesting reading. Although it is way
out of date there is a lot of useful information (see section 10 for info
about the parameters in the poly file).

Here is the pdf:

chris2be8 · 2013-11-16, 16:51

Back on topic I'm reserving:
53^128+1 C182
76^124+1 C156
96^113+1 C156

(The last two by GNFS).

Chris

chris2be8 · 2013-11-28, 16:55

Time to reserve:
53^143+1
89^113+1

Chris

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Vanishing Fermat Quotients and Brent's List	Zeta-Flux	Factoring	74	2010-04-07 22:03
Brent-Suyama extension of P-1 factoring	S485122	Math	1	2009-08-23 15:21
Brent-Montgomery-te Riele numbers	FactorEyes	Factoring	23	2008-02-22 00:36
brent suyama extension in P-1	bsquared	Factoring	9	2007-05-18 19:24
Brent's p-1 - How to deal with memory problems?	jhillenb	Factoring	4	2005-01-11 23:50

2013-11-16, 16:51	#197
chris2be8 Sep 2009 3²·233 Posts	Back on topic I'm reserving: 53^128+1 C182 76^124+1 C156 96^113+1 C156 (The last two by GNFS). Chris

2013-11-28, 16:55	#198
chris2be8 Sep 2009 3²×233 Posts	Time to reserve: 53^143+1 89^113+1 Chris