Commit Graph

5720 Commits

Author SHA1 Message Date
Joe Taylor
109fc57861 Merge branch 'develop' of bitbucket.org:k1jt/wsjtx into develop 2020-03-17 14:40:45 -04:00
Joe Taylor
aff8229fef For non-FT8 modes, reset m_nDecodes to 0 after receiving <DecodeFinished>. 2020-03-17 14:39:15 -04:00
Steven Franke
0f2793dcbe Update calls to bpdecode174_91. 2020-03-17 12:18:10 -05:00
Steven Franke
8f7221825f First cut at setting up Fast/Normal/Deep parameters. 2020-03-17 08:51:37 -05:00
Joe Taylor
a2c3907b75 Fix a bug that affected jt9 running from command line. 2020-03-16 13:47:41 -04:00
Joe Taylor
5e94de71fd Use tseq rather than ipc1 as test for early bailouts. 2020-03-16 12:11:56 -04:00
Joe Taylor
8b8cadcf3a Fix some comments and do not compare ints with floats. 2020-03-16 11:08:09 -04:00
Joe Taylor
d6cc789f40 Merge branch 'develop' of bitbucket.org:k1jt/wsjtx into develop 2020-03-16 10:54:25 -04:00
Bill Somerville
5024cc4e2d
Add missed file to repo 2020-03-16 14:39:42 +00:00
Joe Taylor
fea2d43489 Merge branch 'develop' of bitbucket.org:k1jt/wsjtx into develop 2020-03-16 10:35:08 -04:00
Joe Taylor
2653626d45 Replace the changes I removed by executing 'git reset HEAD~'. 2020-03-16 10:27:22 -04:00
Bill Somerville
488c2e8066
Cleanup and refactor Fortran shared memory usage
also added some missing locking for the ipc(1) value in ft8_decode().
2020-03-16 13:58:38 +00:00
Joe Taylor
3962aa3c43 Replace hijacked ss(1:3,1) with a separate ipc(3) array. 2020-03-15 14:25:56 -04:00
Bill Somerville
46638f6f07
Add shared memory synchronization to make access safe 2020-03-15 00:25:35 +00:00
Joe Taylor
649a79dfc1 New IPC trial: use shmem instead of .start .stop .quit files. 2020-03-14 17:06:54 -04:00
Bill Somerville
523e9a1a07
Enhanced process control for jt9 executable
The wsjtx  process creates control  files .start, .stop, or  .quit and
the jt9  process deletes  them.  This  is intended  to avoid  any race
conditions that get the processes out of sync.
2020-03-14 00:01:54 +00:00
Joe Taylor
ebf4952c7c Auto-clear of hung decoder status. 2020-03-13 14:25:12 -04:00
Joe Taylor
8151816c84 Change formats of diagnostic writes so that GUI and Fortran output can be sorted together. 2020-03-13 11:25:22 -04:00
Joe Taylor
fedb72a38c Set m_nDecodes=0 only after the 50-buffer decode pass. 2020-03-12 16:52:19 -04:00
K9AN
f7d7c073c8 Fix a bug. 2020-03-12 15:36:03 -05:00
Joe Taylor
afe0262ec9 Previous commit message should have mentioned number of decodes is now displayed on status bar. 2020-03-12 15:11:20 -04:00
Joe Taylor
8e7685ea36 Do not alolow bailouts when data were read from disk. 2020-03-12 15:07:56 -04:00
Joe Taylor
0d535f24fb Move the AA and CC diagnostic writes from jt9a() to decoder(). Add ndecodes to BB and CC. 2020-03-12 14:31:02 -04:00
Joe Taylor
f30754bdf5 Remove superfluous part of if() test. 2020-03-12 13:53:57 -04:00
Joe Taylor
68c942a70e Correct the 47-buffer bailout place. Change bailout times 45->46 and 48->49. 2020-03-12 13:00:42 -04:00
Joe Taylor
6e308f0018 More attempts to diagnose or fix the FT8 "hung decoder" issue.
On the Fortran side:
 - For the nzhsym=41 activation, bail out if m_ihsym reaches 45
 - For the nzhsym=47 activation, bail out if m_ihsym reaches 48
 - Change the format and content of what's written to fort.71
 - Change msdelay from 10 to 1

In mainwindow.cpp:
 - change format and content of what's written to qDebug
 - always start FT8 decoder at m_ihsym = 41, 47, and 50
2020-03-12 12:33:34 -04:00
Bill Somerville
6a7263dc94
Add escape route from retries of IPC errors 2020-03-11 22:38:41 +00:00
Bill Somerville
c58a690bf1
Add error checking to jt9 process control by .lock/.quit files 2020-03-11 22:26:41 +00:00
Joe Taylor
4d9c9e08e2 If 47-buffer call takes too long, start the 50-buffer call late. 2020-03-11 16:03:51 -04:00
Joe Taylor
d69cb49e9a Diagnostic writes to fort.71 in jt9a.f90. 2020-03-11 14:02:19 -04:00
Joe Taylor
199ff3dd3d Experimental work-around for FT8 decoding on slow computers. Important changes:
- don't start function decode() is decoder is already busy
  - send updated m_ihsym to jt9[.exe] via ss(1,1).  Bill won't like this!
  - jt9 bails out of the 41-buffer pass at m_ihsym=45.
2020-03-10 16:26:42 -04:00
Joe Taylor
592f4eea71 Remove a qDebug() call. 2020-03-09 15:29:57 -04:00
Steven Franke
cf5aab0088 Allow up to 500 pre-candidates. 2020-03-09 14:03:53 -05:00
Steven Franke
d57b84be27 More changes to increase ft8 decoder yield. 2020-03-08 12:03:52 -05:00
Steven Franke
b130479249 Make jt9 work with the 41/47/50 scheme. 2020-03-06 19:03:09 -06:00
Steven Franke
83d87e9654 Fix an apparent typo in ft8_decode. 2020-03-06 19:02:41 -06:00
Steven Franke
fa071324e5 subtractft8: Delete unused variable, i1 2020-03-06 13:21:04 -06:00
Steven Franke
3467ee7924 Make Normal decode the same as Deep except no DT refinement at nearly=47. 2020-03-06 11:54:28 -06:00
Steven Franke
f2266c3c54 1. subtractft8: on the last call to sqf, don't need to do the long FFT.
2. mainwindow.cpp: based on the time when <DecodeFinished> is received, calculate
mswait to try to place the killFile time at 0.75*m_TRperiod.
2020-03-06 11:06:44 -06:00
Joe Taylor
f367b6d6b9 Temporary version: decode call at 47 buffers only does subtractions. 2020-03-06 10:38:24 -05:00
Joe Taylor
431ec50212 Progress toward decoding at 41/47/50 buffers. Not yet complete! 2020-03-05 15:37:31 -05:00
Steven Franke
6773768586 Clean up the timestamp adjustment logic, and make the threshold 0.5*TRperiod. 2020-03-05 11:25:49 -06:00
Steven Franke
cb687c7034 Merge branch 'develop' of bitbucket.org:k1jt/wsjtx into develop 2020-03-05 11:17:03 -06:00
Steven Franke
07c753e284 Make ALL.TXT Rx timestamp equal to the beginning of the TR period, even if the decode comes in after the start of the next TR period. Make Tx timestamp the actual time that Tx started. 2020-03-05 11:16:56 -06:00
Joe Taylor
cd9efc54ea Correct the reported value "ndecoded" for FT8 mode. 2020-03-03 15:18:25 -05:00
Joe Taylor
02eb4fe90b Use a table lookup for exp(i*phi) in gen_ft8wave.f90. 2020-03-03 12:40:27 -05:00
Steven Franke
761ee1cd85 Merge branch 'develop' of bitbucket.org:k1jt/wsjtx into develop 2020-02-29 13:47:09 -06:00
Steven Franke
037805ca8f Improve dt estimation in ft8b so that dt refinement procedure in subtractft8.f90 can be done with 4 calls to sqf instead of 7. 2020-02-29 13:46:59 -06:00
Joe Taylor
6c1efac576 Once more, to get the <DecodeFinished> behavior right for realtime monitoring as well as data from disk. 2020-02-28 15:45:51 -05:00
Joe Taylor
8e7a56634b Set nearly=41; don't write <DecodeFinished> after the early decode pass. 2020-02-28 15:03:39 -05:00