What exponents correspond to FFT length changes

JuanTutors · 2006-11-11, 18:10

I know this question has been asked may times before, but for a while now I have been trying and can't find the answer. I want to know exactly when Prime95 switches from one FFT length to another, so that I can rearrange things in my worktodo.ini file.

Does anyone know the answer off hand, or maybe have a link to the old thread?

jinydu · 2006-11-11, 19:08

http://www.mersenne.org/status.htm

But I don't about exponents larger than 79.3 million.

PhilF · 2006-11-11, 19:31

I don't know about the lower ranges, but the point for switching from 1792K to 2048K FFT is about 34620000. Any exponent above that gets 2048K, and ones in the range of approximately 34560000 to 34620000 is in the "soft" crossover point, although every one of mine in that range has been tested with a 1792K FFT.

JuanTutors · 2006-11-11, 21:56

Quote:

Originally Posted by jinydu

http://www.mersenne.org/status.htm

But I don't about exponents larger than 79.3 million.

Is that table accurate? I posted because I wanted an exact answer so that I can do some rearranging of my worktodo.ini file, but could have sworn I saw somewhere that the point they switched from 1792K to 2048K was APPROXIMATELY where PhilF said.

Can we get yet another opinion on this question?

markr · 2006-11-11, 23:30

This thread has a post by garo quoting George comparing FFT crossovers in version 21 & 22.8, which I think was the last major change to FFT crossovers.

JuanTutors · 2006-11-12, 02:22

Quote:

Originally Posted by markr

This thread has a post by garo quoting George comparing FFT crossovers in version 21 & 22.8, which I think was the last major change to FFT crossovers.

Sweet. Thanks.

cheesehead · 2006-11-12, 10:11

dominicanpapi82,

For the "official" correspondences of exponent and FFT length, look in Prime95 source module mult.asm at tables jmptable, jmptablep, xjmptable, and xjmptablep.

2006-11-11, 18:10	#1
JuanTutors Mar 2004 2²×3³×5 Posts	What exponents correspond to FFT length changes I know this question has been asked may times before, but for a while now I have been trying and can't find the answer. I want to know exactly when Prime95 switches from one FFT length to another, so that I can rearrange things in my worktodo.ini file. Does anyone know the answer off hand, or maybe have a link to the old thread?

2006-11-12, 10:11	#7
cheesehead "Richard B. Woods" Aug 2002 Wisconsin USA 2²×3×641 Posts	dominicanpapi82, For the "official" correspondences of exponent and FFT length, look in Prime95 source module mult.asm at tables jmptable, jmptablep, xjmptable, and xjmptablep. Last fiddled with by cheesehead on 2006-11-12 at 10:27

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Help with FFT length	Gradient	Software	3	2013-12-16 01:53
Factorization length test?	siegert81	Math	17	2011-12-20 15:33
FFT Length	Samoflan	Information & Answers	8	2010-02-16 22:05
FFt length	mack	Information & Answers	1	2009-09-06 03:24
llr: FFT-length not monotone in n?	hhh	Software	4	2008-12-20 09:49

2006-11-11, 19:08	#2
jinydu Dec 2003 Hopefully Near M48 3336₈ Posts	http://www.mersenne.org/status.htm But I don't about exponents larger than 79.3 million.

2006-11-11, 19:31	#3
PhilF Feb 2005 Colorado 2×7×47 Posts	I don't know about the lower ranges, but the point for switching from 1792K to 2048K FFT is about 34620000. Any exponent above that gets 2048K, and ones in the range of approximately 34560000 to 34620000 is in the "soft" crossover point, although every one of mine in that range has been tested with a 1792K FFT.

2006-11-11, 23:30	#5
markr "Mark" Feb 2003 Sydney 3·191 Posts	This thread has a post by garo quoting George comparing FFT crossovers in version 21 & 22.8, which I think was the last major change to FFT crossovers.