Subversion Repositories shark

Rev

Rev 2 | Go to most recent revision | Details | Compare with Previous | Last modification | View Log | RSS feed

Rev Author Line No. Line
2 pj 1
The following are a number of ideas for future work that we have
2
thought of, or which have been suggested to us.  Let us know
3
(fftw@theory.lcs.mit.edu) if you have other proposals, or if there is
4
something that you want to work on.
5
 
6
* Implement some sort of Prime Factor algorithm (Temperton's?)  (PFA is
7
now used in the codelets.)
8
 
9
* Try the Winograd blocks for the base cases. (We now use Rader's
10
algorithm for prime size codelets.)
11
 
12
* Try on-the-fly generation of twiddle factors, to save space and
13
cache. (Done.  However, not yet enabled in the standard distribution.
14
The codelet generator is capable of generating code that either loads
15
or computes the twiddle factors, and the FFTW C code supports both
16
ways.  We do not have enough experimental numbers to determine which
17
way is faster, however)
18
 
19
* Since we now have "strided wisdom," it would be nice to keep the
20
stride into account when planning 1D transform recursively.  We should
21
eliminate the planner table altogether, and just use the wisdom table
22
for planning.
23
 
24
* Implement fast DCT and DST codes (cosine and sine transforms);
25
equivalently, implement fast algorithms for transforms of real/even
26
and real/odd data.  There are two parts to this: (i) modify the
27
codelet generator to output hard-coded transforms of small sizes [this
28
is done], and (ii) figure out & implement a recursive framework for
29
combining these codelets to achieve transforms of general lengths.
30
(Once this is done, implement multi-dimensional transforms, etcetera.)
31
 
32
* Implement a library of convolution routines, windowing, filters,
33
etcetera based on FFTW.  As DSP isn't our field (we are interested in
34
FFTs for other reasons), this sort of thing is probably best left to
35
others.  Let us know if you're interested in writing such a thing,
36
though, and we'll be happy to link to your site and give you feedback.
37
 
38
* Generate multi-dimensional codelets for use in two/three-dimensional
39
transforms.  (i.e. implement what is sometimes called a "vector-radix"
40
algorithm.)  There are potential cache benefits to this.
41
 
42
* Take advantage of the vector instructions on the Pentium-III and
43
forthcoming PowerPC architectures.  (Coming from the old Cray vector
44
supercomputers and the horrible coding they encouraged, this seems
45
suspiciously like a giant step backwards in computer architectures...)
46
We'd like to see better gcc support before we do anything along these
47
lines, though.
48
 
49
* In rfftw, implement a fast O(n lg n) algorithm for prime sizes and
50
large prime factors (currently, only the complex FFTW has fast
51
algorithms for prime sizes).  The basic problem is that we don't know
52
of any such algorithm specialized for real data; suggestions and/or
53
references are welcome.
54
 
55
* In the MPI transforms, implement a parallel 1D transform for real
56
data (i.e. rfftw_mpi).  (Currently, there are only parallel 1D
57
transforms for complex data in the MPI code.)
58
 
59
* In the MPI transforms, implement more sophisticated (i.e. faster)
60
in-place and out-of-place transpose routines for the in-process
61
transposes (used as subroutines by the distributed transpose).  The
62
current routines are quite simplistic, although it is not clear how
63
much they hurt performance.