Hi All,
sorry for having missed the telecon yesterday and for not having replied so
far. I am on 'kids-sitting' vacations. I know total convolver, or
improvements, is a solution, but I don't think it is the optimal one for
our observations from a computational viewpoint. I admit that optimal means
NOW timely, so it may be _the_ practical solution, but we may want
something else, more flexible.
I wrote yesterday just after seeing Chao-Lin reference to a Paris-Planck
colleague with whom I learned to use total convolver, -I was worried for
FSL-dipole contamination in Planck HFI some time ago-, who has finally run
last week the FSL for Planck all-sky in HFI. He knew eaxactly waht was
used. I am also in direct contact with the actual writer of the code (that
Clem was correctly pointing out, a significantly faster version of total
convolver) who is Gary Prezau here on Lab who worked with Reinecke. So we
have 'all' at home.
Just waiting for some confirmations and opinions from him about porting
(or using it directly) and my eclectic idea (neither harmonic, neither
Gaussian). T Better not having been on the telecon :)
I wanted to have all in hands, but I see there is some action here. and I
was not saying anything.
I'll let you know asap -today there is a special session on consistency
paper and Gary will be busy.
On Wed, Aug 7, 2013 at 11:42 AM, John Kovac <jmkovac(a)cfa.harvard.edu> wrote:
I expect Stefan would be as delighted as all of us if we were able to
include full 3-angle explicit convolution on the sphere as a pipeline
option, useful for at least some applications.
I think the standards for how computationally expensive we can afford
this to be are relaxed for systematics studies, of which we need only a
few realizations on limited tag subsets.
I hope Sergi or Chao-Lin are able to pursue this.
John
On 8/7/13 1:15 PM, Clem Pryke wrote:
This uses the "Total Convolver"
algorithm from Wandelt and Gorski:
http://arxiv.org/abs/astro-ph/0008227
I talked to Ben and his postdoc extensively once about this. I don't
remember
the details although I think there is another
algorithm they ended up
deciding
was even better.
It seems intuitively obvious that a high-resolution convolution of 4pi
with
4pi at arbitrary 3 angle relative rotation is
going to be expensive - I
am
happy to be proven wrong. My point on the telecon
was simply that if
such a
convolution is practical in the main pipeline we
should use it always -
although Stefan would presumably be upset - as he has spent much time on
the
various multi-Gauss style approximations ;-)
Clem
--
___________________________________________________________________
John Kovac jmkovac(a)cfa.harvard.edu
Associate Professor, Astronomy and Physics, Harvard University
160 Concord Ave rm 310, Cambridge MA 02138, 617-496-0611
_______________________________________________
Bicep2-list mailing list
Bicep2-list(a)lists.fas.harvard.edu
https://lists.fas.harvard.edu/mailman/listinfo/bicep2-list
--
Dr Sergi Hildebrandt Rafels
Jet Propulsion Laboratory ---------------- California Institute of
Technology
169-217 Cahill-383
4800 Oak Grove Drive 1200, E. California Bvd
Pasadena, CA, 91109 Pasadena, CA, 91125
MC/169-237 MC/367-17
1-818-354-0220 1-626-395-2147