Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/TODO @ 82:d0c2a83c1364

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW 3.3.8 source, and a Linux build
author Chris Cannam
date Tue, 19 Nov 2019 14:52:55 +0000
parents
children
rev   line source
Chris@82 1 TODO before FFTW-$2\pi$:
Chris@82 2
Chris@82 3 * figure out how to autodetect NEON at runtime
Chris@82 4
Chris@82 5 * figure out the arm cycle counter business
Chris@82 6
Chris@82 7 * Wisdom: make it clear that it is specific to the exact fftw version
Chris@82 8 and configuration. Report error codes when reading wisdom. Maybe
Chris@82 9 have multiple system wisdom files, one per version?
Chris@82 10
Chris@82 11 * DCT/DST codelets? which kinds?
Chris@82 12
Chris@82 13 * investigate the addition-chain trig computation
Chris@82 14
Chris@82 15 * I can't believe that there isn't a closed form for the omega
Chris@82 16 array in Rader.
Chris@82 17
Chris@82 18 * convolution problem type(s)
Chris@82 19
Chris@82 20 * Explore the idea of having n < 0 in tensors, possibly to mean
Chris@82 21 inverse DFT.
Chris@82 22
Chris@82 23 * better estimator: possibly, let "other" cost be coef * n, where
Chris@82 24 coef is a per-solver constant determined via some big numerical
Chris@82 25 optimization/fit.
Chris@82 26
Chris@82 27 * vector radix, multidimensional codelets
Chris@82 28
Chris@82 29 * it may be a good idea to unify all those little loops that do
Chris@82 30 copying, (X[i], X[n-i]) <- (X[i] + X[n-i], X[i] - X[n-i]),
Chris@82 31 and multiplication of vectors by twiddle factors.
Chris@82 32
Chris@82 33 * Pruned FFTs (basically, a vecloop that skips zeros).
Chris@82 34
Chris@82 35 * Try FFTPACK-style back-and-forth (Stockham) FFT. (We tried this a
Chris@82 36 few years ago and it was slower, but perhaps matters have changed.)
Chris@82 37
Chris@82 38 * Generate assembly directly for more processors, or maybe fork gcc. =)
Chris@82 39
Chris@82 40 * ensure that threaded solvers generate (block_size % 4 == 0)
Chris@82 41 to allow SIMD to be used.
Chris@82 42
Chris@82 43 * memoize triggen.