Mercurial > hg > octave-lojdl
changeset 4809:b60be7678bdc
[project @ 2004-03-02 17:40:08 by jwe]
author | jwe |
---|---|
date | Tue, 02 Mar 2004 17:40:08 +0000 |
parents | a9ec0ce18568 |
children | 72a6d410a14a |
files | liboctave/oct-fftw.cc |
diffstat | 1 files changed, 89 insertions(+), 46 deletions(-) [+] |
line wrap: on
line diff
--- a/liboctave/oct-fftw.cc +++ b/liboctave/oct-fftw.cc @@ -31,20 +31,23 @@ #include "oct-fftw.h" #include "quit.h" -// Helper class to create and cache fftw plans for both 1d and 2d. This -// implementation uses FFTW_ESTIMATE to create the plans, which in theory -// is suboptimal, but provides quite reasonable performance. +// Helper class to create and cache fftw plans for both 1d and +// 2d. This implementation uses FFTW_ESTIMATE to create the plans, +// which in theory is suboptimal, but provides quite reasonable +// performance. // Also note that if FFTW_ESTIMATE is not used the planner in FFTW3 -// destroys the input and output arrays. So with the form of the -// current code we definitely want FFTW_ESTIMATE!! However, we use -// any wsidom that is available, either in a FFTW3 system wide file -// or as supplied by the user. +// destroys the input and output arrays. So with the form of the +// current code we definitely want FFTW_ESTIMATE!! However, we use any +// wsidom that is available, either in a FFTW3 system wide file or as +// supplied by the user. -// XXX FIXME XXX If we can ensure 16 byte alignment in Array<T> (<T> *data) -// the FFTW3 can use SIMD instructions for further acceleration. +// XXX FIXME XXX -- if we can ensure 16 byte alignment in Array<T> +// (<T> *data) the FFTW3 can use SIMD instructions for further +// acceleration. -// Note that it is profitable to store the FFTW3 plans, for small ffts +// Note that it is profitable to store the FFTW3 plans, for small +// ffts. class octave_fftw_planner @@ -65,22 +68,47 @@ int plan_flags; + // XXX FIXME XXX -- perhaps this should be split into two classes? + // Plan for fft and ifft of complex values fftw_plan plan[2]; - int d[2]; // dist - int s[2]; // stride - int r[2]; // rank - int h[2]; // howmany - dim_vector n[2]; // dims + + // dist + int d[2]; + + // stride + int s[2]; + + // rank + int r[2]; + + // howmany + int h[2]; + + // dims + dim_vector n[2]; + bool simd_align[2]; + // Plan for fft of real values fftw_plan rplan; - int rd; // dist - int rs; // stride - int rr; // rank - int rh; // howmany - dim_vector rn; // dims + + // dist + int rd; + + // stride + int rs; + + // rank + int rr; + + // howmany + int rh; + + // dims + dim_vector rn; + bool rsimd_align; }; @@ -98,7 +126,7 @@ rsimd_align = false; rn = dim_vector (); - // If we have a system wide wisdom file, import it + // If we have a system wide wisdom file, import it. fftw_import_system_wisdom (); } @@ -116,21 +144,25 @@ bool create_new_plan = false; bool ioalign = CHECK_SIMD_ALIGNMENT (in) && CHECK_SIMD_ALIGNMENT (out); - // Don't create a new plan if we have a non SIMD plan already - // but can do SIMD. This prevents endlessly recreating plans - // if we change the alignment + // Don't create a new plan if we have a non SIMD plan already but + // can do SIMD. This prevents endlessly recreating plans if we + // change the alignment. + if (plan[which] == 0 || d[which] != dist || s[which] != stride || r[which] != rank || h[which] != howmany || ((ioalign != simd_align[which]) ? !ioalign : false)) create_new_plan = true; else - // We still might not have the same shape of array - for (int i = 0; i < rank; i++) - if (dims(i) != n[which](i)) - { - create_new_plan = true; - break; - } + { + // We still might not have the same shape of array. + + for (int i = 0; i < rank; i++) + if (dims(i) != n[which](i)) + { + create_new_plan = true; + break; + } + } if (create_new_plan) { @@ -149,8 +181,10 @@ if (*cur_plan_p) fftw_destroy_plan (*cur_plan_p); - // Note reversal of dimensions for column major storage in FFTW + // Note reversal of dimensions for column major storage in FFTW. + OCTAVE_LOCAL_BUFFER (int, tmp, rank); + for (int i = 0, j = rank-1; i < rank; i++, j--) tmp[i] = dims(j); @@ -176,20 +210,24 @@ bool create_new_plan = false; bool ioalign = CHECK_SIMD_ALIGNMENT (in) && CHECK_SIMD_ALIGNMENT (out); - // Don't create a new plan if we have a non SIMD plan already - // but can do SIMD. This prevents endlessly recreating plans - // if we change the alignment + // Don't create a new plan if we have a non SIMD plan already but + // can do SIMD. This prevents endlessly recreating plans if we + // change the alignment. + if (rplan == 0 || rd != dist || rs != stride || rr != rank || rh != howmany || ((ioalign != rsimd_align) ? !ioalign : false)) create_new_plan = true; else - // We still might not have the same shape of array - for (int i = 0; i < rank; i++) - if (dims(i) != rn(i)) - { - create_new_plan = true; - break; - } + { + // We still might not have the same shape of array. + + for (int i = 0; i < rank; i++) + if (dims(i) != rn(i)) + { + create_new_plan = true; + break; + } + } if (create_new_plan) { @@ -208,8 +246,10 @@ if (*cur_plan_p) fftw_destroy_plan (*cur_plan_p); - // Note reversal of dimensions for column major storage in FFTW + // Note reversal of dimensions for column major storage in FFTW. + OCTAVE_LOCAL_BUFFER (int, tmp, rank); + for (int i = 0, j = rank-1; i < rank; i++, j--) tmp[i] = dims(j); @@ -318,7 +358,8 @@ fftw_execute_dft_r2c (plan, (const_cast<double *>(in)), reinterpret_cast<fftw_complex *> (out)); - // Need to create other half of the transform + // Need to create other half of the transform. + convert_packcomplex_1d (out, nsamples, npts, stride, dist); return 0; @@ -372,7 +413,8 @@ dist *= dv(i); // Fool with the position of the start of the output matrix, so that - // creating other half of the matrix won't cause cache problems + // creating other half of the matrix won't cause cache problems. + int offset = (dv.numel () / dv(0)) * ((dv(0) - 1) / 2); fftw_plan plan = fftw_planner.create_plan (rank, dv, 1, 1, dist, @@ -381,7 +423,8 @@ fftw_execute_dft_r2c (plan, (const_cast<double *>(in)), reinterpret_cast<fftw_complex *> (out+ offset)); - // Need to create other half of the transform + // Need to create other half of the transform. + convert_packcomplex_Nd (out, dv); return 0;