Optimizations of JT9 and JT65 decoders; change clock in timer routine.

Both decoders now have slightly better performance and faster
execution.  The rare "duplicate decodes" in JT9 were eliminated.
On Windows, at least, calls to f90 routine system_clock() do not
provide correct wall time increments.  Changed to using secnds()
instead.


git-svn-id: svn+ssh://svn.code.sf.net/p/wsjt/wsjt/branches/wsjtx@4571 ab8295b8-cf94-4d9e-aec4-7959e3be5d79

lib/decoder.f90

@@ -21,7 +21,6 @@ subroutine decoder(ss,id2)
   common/tracer/limtrace,lu
   save
 
-  call system_clock(iclock0,iclock_rate,iclock_max)           !###
   nfreqs0=0
   nfreqs1=0
   ndecodes0=0
@@ -84,13 +83,13 @@ subroutine decoder(ss,id2)
   dblim=db(864.0/nsps8) - 26.2
 
   do nqd=1,0,-1
-     limit=1000
-     ccflim=4.0
+     limit=5000
+     ccflim=3.0
      red2lim=1.6
      schklim=2.2
      if(ndepth.eq.2) then
         limit=10000
-        ccflim=3.5
+        ccflim=2.7
      endif
      if(ndepth.ge.3 .or. nqd.eq.1) then
         limit=100000
@@ -135,17 +134,15 @@ subroutine decoder(ss,id2)
                 freq,drift,schk,i1SoftSymbols)
            call timer('softsym ',1)
 
-!           write(71,3001) nqd,i,f,fpk,ccfred(i),red2(i),schk
-!3001       format(2i6,2f8.1,3f6.1)
-!           call flush(71)
-
-           if(schk.lt.schklim) cycle
+           sync=(syncpk+1)/4.0
+           if(maxval(i1SoftSymbols).eq.0) cycle
+           if(nqd.eq.1 .and. ((sync.lt.0.5) .or. (schk.lt.2.0))) cycle
+           if(nqd.ne.1 .and. ((sync.lt.1.0) .or. (schk.lt.schklim))) cycle
 
            call timer('decode9 ',0)
            call decode9(i1SoftSymbols,limit,nlim,msg)
            call timer('decode9 ',1)
 
-           sync=(syncpk+1)/4.0
            if(sync.lt.0.0 .or. snrdb.lt.dblim-2.0) sync=0.0
            nsync=sync
            if(nsync.gt.10) nsync=10
@@ -183,7 +180,7 @@ subroutine decoder(ss,id2)
      call jt65a(dd,npts65,newdat,nutc,nf1,nf2,nfqso,ntol65,nagain,ndecoded)
   endif
 
-!### JT65 is not yet producing info for nsynced, ndecoded.
+! JT65 is not yet producing info for nsynced, ndecoded.
 800 write(*,1010) nsynced,ndecoded
 1010 format('<DecodeFinished>',2i4)
   call flush(6)