FFTW 2.1.5

Use of Healpix, camb, CLASS, cosmomc, compilers, etc.
Post Reply
William Jones
Posts: 7
Joined: February 24 2005
Affiliation: Princeton
Contact:

FFTW 2.1.5

Post by William Jones » July 30 2005

Hi Folks,

Has anybody ever run into problems with seg faults in FFTW when handling moderately large arrays? Im having a weird time with it recently - getting seg faults when passing

rfftw_create_plan((int)nsamp,FFTW_REAL_TO_COMPLEX,FFTW_ESTIMATE);

with nsmap >= 2^19 . Its not a memory thing, and it's repeatable on at least three different platforms. Its definately isolated to the create_plan call. hoping not to have to port to v3 ...

Thanks,

-Bill

Gil Holder
Posts: 27
Joined: September 25 2004
Affiliation: McGill University

FFTW 2.1.5

Post by Gil Holder » July 30 2005

I have had no problems on 2048x2048 maps doing complex transforms, single precision, using some version of fftw2.x (unfortunately I don't have ready access to the value of x). Is your whole array 2^19 elements? If so, that is very strange.

William Jones
Posts: 7
Joined: February 24 2005
Affiliation: Princeton
Contact:

FFTW 2.1.5

Post by William Jones » August 01 2005

Ah - apologies to Frigo and Johnson. Was a problem in my makefile, not properly linking some of the rfft libraries properly.. Thanks though, Gil. -bill

Post Reply