Commit Graph

13 Commits

Author SHA1 Message Date
Bill Somerville
95926577ae Make Fortran profiling timer function a callback with a default null implementation
Groundwork for calling the decoders directly from C/C++ threads.

To  access   the  timer   module  timer_module   must  now   be  used.
Instrumented code need  only use the module function  'timer' which is
now a  procedure pointer that  is guaranteed to be  associated (unless
null()  is assigned  to it,  which should  not be  done). The  default
behaviour of 'timer' is to do nothing.

If a  Fortran program  wishes to  profile code it  should now  use the
timer_impl module  which contains a default  timer implementation. The
main program should call 'init_timer([filename])' before using 'timer'
or     calling     routines     that     are     instrumented.      If
'init_timer([filename])'.  If it is called  then an optional file name
may  be  provided  with  'timer.out'  being used  as  a  default.  The
procedure 'fini_timer()' may be called to close the file.

The default  timer implementation is  thread safe if used  with OpenMP
multi-threaded code  so long as  the OpenMP  thread team is  given the
copyin(/timer_private/) attribute  for correct operation.   The common
block /timer_private/ should  be included for OpenMP  use by including
the file 'timer_common.inc'.

The module 'lib/timer_C_wrapper.f90' provides  a Fortran wrapper along
with 'init' and 'fini' subroutines  which allow a C/C++ application to
call timer instrumented  Fortran code and for it  to receive callbacks
of 'timer()' subroutine invocations.  No C/C++ timer implementation is
provided at this stage.

git-svn-id: svn+ssh://svn.code.sf.net/p/wsjt/wsjt/branches/wsjtx@6320 ab8295b8-cf94-4d9e-aec4-7959e3be5d79
2015-12-27 15:40:57 +00:00
Joe Taylor
276b726b2d Remove calls to timer() around the "flatten" loop in filbig.
git-svn-id: svn+ssh://svn.code.sf.net/p/wsjt/wsjt/branches/wsjtx@5017 ab8295b8-cf94-4d9e-aec4-7959e3be5d79
2015-03-06 19:44:46 +00:00
Joe Taylor
93b729282b The "big FFTs" should be computed only once for each new set of Rx data.
This requires setting newdat=0 after the big FFT is computed.  In the OMP
code this must be done separately for each mode; so new variables newdat9 
and newdat65 have been defined.  Both are set to "newdat", the value
forwarded from the GUI, each time jt9[_omp][.exe] goes into action. 


git-svn-id: svn+ssh://svn.code.sf.net/p/wsjt/wsjt/branches/wsjtx@4946 ab8295b8-cf94-4d9e-aec4-7959e3be5d79
2015-02-08 15:07:31 +00:00
Joe Taylor
ec7bf0652a Another instance of variables needing to be saved explicitly between
calls to a subroutine.  I believe this fixes the known outstanding decode
issue.


git-svn-id: svn+ssh://svn.code.sf.net/p/wsjt/wsjt/branches/wsjtx@4941 ab8295b8-cf94-4d9e-aec4-7959e3be5d79
2015-02-06 18:44:45 +00:00
Bill Somerville
43397b9580 Only use >1 thread for the biggest FFTs
git-svn-id: svn+ssh://svn.code.sf.net/p/wsjt/wsjt/branches/wsjtx@4933 ab8295b8-cf94-4d9e-aec4-7959e3be5d79
2015-02-04 20:52:42 +00:00
Bill Somerville
8d8d64c07c Thread safety enhancements to jt9
Serialize thread unsafe FFTW3 calls

Serialize FFTW3 plan initializations in four2a

Serialize decoder results output to file/console

git-svn-id: svn+ssh://svn.code.sf.net/p/wsjt/wsjt/branches/wsjtx@4928 ab8295b8-cf94-4d9e-aec4-7959e3be5d79
2015-02-04 01:41:26 +00:00
Joe Taylor
8543283404 Add rfilt and df to the "save" list in filbig.f90.
git-svn-id: svn+ssh://svn.code.sf.net/p/wsjt/wsjt/branches/wsjtx@4927 ab8295b8-cf94-4d9e-aec4-7959e3be5d79
2015-02-04 00:37:33 +00:00
Joe Taylor
960da46d28 Many changes in program jt9[.exe] aimed at speeding up the decoders.
The long FFTs can now use the multi-threaded FFTW routines.
Subroutine decode9.f90 was renamed jt9fano.f90.
The JT9 decoder's top-level functions were removed from decoder.f90
and put into a separate subroutine decjt90.f90.
Subroutine decoder.f90 is now configured for possible use of OpenMP 
SECTIONS, with the JT9 and JT65 decoders running concurrently on
a multi-core machine.  Note, however, that this concurrent processing 
is not yet fully implemented.  Probably calls to timer need to be removed; 
some variables used in calls to jt65a and decjt9 may need to be 
declared PRIVATE in decoder; some sections probably need to be declared 
CRITICAL; probably some SAVE statements in downstream routines have
made them not thread-safe; etc., etc.  

I'm a neophyte at using OpenMP.  Comments, suggestions, and/or tests by
others will be welcome!



git-svn-id: svn+ssh://svn.code.sf.net/p/wsjt/wsjt/branches/wsjtx@4919 ab8295b8-cf94-4d9e-aec4-7959e3be5d79
2015-02-01 16:23:36 +00:00
Joe Taylor
647f809c97 Made a start at implementing an option to use multi-threaded FFTs.
New command-line option for jt9: [-m nthreads].  Default is nthreads=1.
Also refactored a loop in filbig.f90 that was taking far too much
time.


git-svn-id: svn+ssh://svn.code.sf.net/p/wsjt/wsjt/branches/wsjtx@4916 ab8295b8-cf94-4d9e-aec4-7959e3be5d79
2015-01-30 21:28:10 +00:00
Joe Taylor
175f1913b4 FFTW wisdom is now built into jt9[.exe].
New optional argument to jt9: -w patience
Default is patience = 1

Example timing measurements for 130610_2343.wav:

patience  plan   execute
          (s)      (s)
-----------------------------------------------
   0      0.01     1.25  FFTW_ESTIMATE
   1      0.69     1.25  FFTW_ESTIMATE_PATIENT
   2     16.97     1.15  FFTW_MEASURE
   3    390.88     1.15  FFTW_PATIENT

Conclusions, consistent with expectation based on past experience
with similar FFTs:
  - First decode (in each mode) with patience = 2 is slow.
  - Speed advantage of patience = 2 is small but measurable.
  - No measurable advantage in using patience > 2.

Present mainwindow.cpp has "-w 1" hard-wired.


git-svn-id: svn+ssh://svn.code.sf.net/p/wsjt/wsjt/branches/wsjtx@4610 ab8295b8-cf94-4d9e-aec4-7959e3be5d79
2014-11-20 18:48:53 +00:00
Joe Taylor
11487c69fa Correct the logic for negative-frequency index into new ca array.
git-svn-id: svn+ssh://svn.code.sf.net/p/wsjt/wsjt/branches/wsjtx@4609 ab8295b8-cf94-4d9e-aec4-7959e3be5d79
2014-11-19 20:17:15 +00:00
Joe Taylor
875dd0724b Several decoder optimizations, including implementation of FFTW wisdom.
Notes:
1. Equivalents of wisdom1.bat will be needed for *nix and OS X.  (The
   version now added to the source .../lib directory is an example only.)
2. Installers should offer to run the wisdom1[.bat] script at installation
   time.
3. wisdom1[.bat] and fftwf-wisdom[.exe] must be installed in .../bin directory 



git-svn-id: svn+ssh://svn.code.sf.net/p/wsjt/wsjt/branches/wsjtx@4607 ab8295b8-cf94-4d9e-aec4-7959e3be5d79
2014-11-19 17:23:57 +00:00
Joe Taylor
8fb483a200 Merging code for v1.1 back into the main wsjtx branch.
git-svn-id: svn+ssh://svn.code.sf.net/p/wsjt/wsjt/branches/wsjtx@3462 ab8295b8-cf94-4d9e-aec4-7959e3be5d79
2013-07-08 13:17:22 +00:00