ChangeLog ($Id$) Version date what has been changed TODO: missing quartet handling missing quartet counting in parallel case DONE: 5.2 14.07.2004 - Version 5.2 !!! NEW RELEASE =========== 5.2.rc17 14.07.2004 - Manual sources moved out of the package - numerous changes in the manual 5.2.rc16 12.07.2004 - Makefile.am: double distclean rule removed - fflush added before mygets() - because of Windows - manual spell-checked - '-consmrel' fixed - usage message extended 5.2.rc15 09.07.2004 - many changes to the manual 5.2.rc14 01.07.2004 - test suite checked, notice about results added 5.2.rc13 01.07.2004 - a number of small fixes (alloc and dealloc in epe) - convergence output for user branches fixed - G-T rate output precision fixed - tests checked and fixed 5.2.rc12 30.06.2004 - branch length convergence output corrected for user set branch lengths (i.e. no estimate) - avergedist updated: output of min/max/var/std.dev. - tstvratio and yrtsratio - no output for GTR - some future cmdline flags added to help - -prefix, -sub50, -randseed moved to usage output - -randseed=# added - release date changed in code 5.2.rc11 18.06.2004 - flags to set epsilon values added for debugging (dependent on USE_ADJUSTABLE_EPS compiler switch) 19.06.2004 - quartet percent computation fixed: quartet sums moved qinfomatr[0] -> qinfomatr[8] qinfomatr[0]=missing qs. (release 5.2.rc9) - character state output added to R/Q matrix output - average computation of dist-matrix changed: mean=average, variance, std deviation TODO: proper output / use Korbis S.E. method 25.06.2004 - GTR model ready, menu order HKY, TN, GTR, (SH), HKY 29.06.2004 - many changes to the manual 5.2.rc10 11.06.2004 - fixed no quartets in parameter estimtion for <4 seqs 15.06.2004 - computation of Bayesian weights moved into subroutine loglkl2weight. - use subroutine loglkl2weight in parallel code -> fixing division by zero bug 5.2.rc9 11.06.2004 - sequence number check for <4 and >257 added <4 - no quartet methods available >257 - no quartet puzzling available - counting for missing quartets added for parallel message checking (TODO: handling and checking) - output of root branch name added for clock analysis (sequence name or "internal branch") 5.2.rc8 04.05.2004 - tree tests move to treetest.c/treetest.h - comments added - typo corrected (suset -> subset) 08.06.2004 - ELW: continue counting p, when significantly worse - 1sKH/2sKH: restrict tests to 3 digits (0.00) (test behaviour: seems not to use some trees (KH), see above - used, but too low variance) 5.2.rc7 12.02.2004 - corrected typos in usage/README 20.02.2004 - cmdline: prefix flag fixed 22.02.2004 - '-wsr/-wsl' flags fixed: output file and file name output 5.2.rc6 25.11.2003 - 'FPRINTF...' changed to 'fprintf(STDOUT,' - strcpy added in openfile-routines 5.2.rc5 25.11.2003 - filename length chack added - typo FILENAMELENTH corrected 5.2.rc4 19.11.2003 - minor corrections: int/uli comparisons - SPRNG added to src/sprng - src/sprng autoconf/automake setup - Parallel Parameter Estimation (JPDC, Ekki 1) 5.2.rc3 07.11.2003 - WAG Matrix code/docu corrected (Goldman) TODO: remove additional output for printrmat TODO: make PEPS* smaller to have better accuracy in parameter estimation. 5.2.rc2 01.09.2003 - bug from 5.1.pl29 found and eliminated alloc size wrong: worked if Maxsubset<=Maxspc TODO: remove out-commented debug messages 5.2.rc1 20.08.2003 - first release candidate for tree-puzzle 5.2 CODE FREEZE =========== 5.1.pl30 20.08.2003 - naming convention changed for manual name input to use FILENAME.puzzle, ... 5.1.pl29 23.07.2003 - new compiler warnings fixed - index error naming fixed - debugging of matrix based permutation (not finished) 5.1.pl28 01.07.2003 - cmdline flags to switch off ml branch lengths (-notreelh), tree tests (-notreetest) 5.1.pl27 24.06.2003 - output quartet support for incorporated split in report file 5.1.pl26 06.06.2003 - menu order of tree reconstruction methods altered to QP, user tree, consensus, dist (no tree) (last two switched) 5.1.pl25 05.06.2003 - small changes to fix compiler warnings 5.1.pl24 04.06.2003 - changes from benchmark-branch merged: debug output removed - checktime fixed: sumtodo <-> done for quartets 5.1.pl23 26.05.2003 - puzzle.c: CONSENSUS -> QUARTPUZ, allocated unnecessary memory - additional malloc-checks added (if buffer == NULL) 01.06.2003 - mlmode conditions with readable constants (ml2.c,ml.h) - garbage collection in tree2bipart/subtree2bipart 5.1.pl22 16.05.2003 - fixing Array Bound Read in fputid (bug had no effect, but was read before check) - garbage collection completed in p-step algorithms and node-label (free_tree/internalnode), trueIDtmp (allpstep), - small fix in test script and rebuild of check-files 5.1.pl21 14.05.2003 - fixing free of all allocations in puzzling step (mrca, recur, split, orig) 5.1.pl20 05.05.2003 - small fix in recursive Alg. (recur.c) - debug output removed - cross check removed from (recur.c,mrca.c) 5.1.pl19 27.04.2003 - recursive Alg. (recur.c) fixed: edge cluster has not to be reconstructed after adding new leaves only updated. - debug output removed from recur.c, split.c - MRCA-based (mrca.c) finished 5.1.pl18 26.04.2003 - recursive (recur.c) finished TODO: remove debug output - MRCA (mrca.c) startet 5.1.pl17 25.04.2003 - split-based (split.c) finished TODO: remove debug output - recursive (recur.c) startet 5.1.pl16 22.04.2003 - orig pstep alg: updating of edgemap outsourced to updateedgemap_orig - writeOTU_orig, writetree_orig, resetedgeinfo_orig, minimumedgeinfo_orig: _orig removed, since independent of algorithm type - onepstep: _orig added (onepstep_orig) - time check/status output moved to checktime routine. added to ml step of parallel version - split.c temporarily added to release 5.1.pl15 18.04.2003 - trueID dependent procedures 'de-trueID-ed' trueID saved in the tips of the tree. 5.1.pl14 10.04.2003 - consensus stuff moved to consensus.h/.c - unique tree sorting moved to treesort.h/.c - changes to pstep-split.h/.c - changed names of procedures using trueID to *_trueID 5.1.pl13 07.04.2003 - subsetmatr.[hc] added to process missing data analyses. (old version, functions in puzzle2) - deglobalized: guessdatatype, readsequencefile, getsizesites, initid, getseqs removed obsolete: getdataset - moved: output of taxa after input - computation of Maxbrnch moved from readsequencefile to inputandinit - readsubsetfile, permutetaxa_ss added, '-rssm' works - empty quartets fixed in checkquartet missing data -> OK, HALT error otherwise 5.1.pl12 04.04.2003 - tests: check-qp-clock regenerated, changes in Makefile.am and template-test - obsolete stuff removed: global trueID_orig + alloc, inittree_orig/freetree_orig from PP_SendDoPermutBlock PP_slave_do_puzzling (ppuzzle), global _orig variables (pstep), makenewsplitentries (old), kh_test (old) - column variable in tree output routines deglobalized - copyright headers added to pstep.[hc] - pstep copied to pstep-{split | mrca | recursive} - implementation of pstep-split.[ch] started (done: structure/init, todo: remove _orig stuff) 5.1.pl11 03.04.2003 - tests/Makefile.am: 'build-puzzle failed' fixed - some changes in test script template - cleanup: removed obsolete routines - puzzle module: makenewsplitentriesnew - ppuzzle module: PP_do_puzzling, PP_SendPermut, PP_RecvPermut, PP_SendSplits, PP_RecvSplits - deglobalized: makenewsplitentries, makepart, computebiparts, copysplit - parallel puzzling step based on allpstep, onepstep - time check/status output added to parallel version (inaccurate since current slave progress not known) - TIMECHECK_INTERVAL for setting time interval of status output in ML/puzzling step - Header when printing 1 PAM rate matrix (-printrmat) - final pstep copied to pstep-{split | mrca | recursive} 5.1.pl10 02.04.2003 - tests/Makefile.am fixed (TODO: build-puzzle.test) - Dates/Version adjusted: April 2003/TREE-PUZZLE 5.2 - Dates/Version fixed in doc/*.html, doc/*.tex - number added to outgroup(menu)/sequence(start) output - Time measurement only done after each completed intermediate tree to reduce the the number from O(Numtrial*n^4) to O(Numtrial) -> less overhead, but 15min met less accurate. - quartet support output for splits added in writeoutputfile for later use (TODO: '#if0'ed) - (sequential) puzzling step moved from puzzle1.c to pstep.c: allpstep, onepstep TODO: parallel puzzling step in ppuzzle.c (works but with old PP_slave_do_puzzling) 5.1.pl9 31.03.2003 - consensus construction added to menu - support value computation fixed for phylip tree, wrong divisor used in consensus - test directory/Makefile.am/tests created (make check) 5.1.pl8 30.03.2003 - puzzling step proc's: most variables 'deglobalized' 5.1.pl7 26.03.2003 - original puzzling step out-sourced to pstep-orig.c and pstep-orig.h pstep.c, pstep.h (and above) added to Makefile.am 5.1.pl6 25.03.2003 - '-usebranch' option added fr later use - 'SONJA' added for external parameter estimation 5.1.pl5 24.03.2003 - ml branch length estimation added to consensus option (todo: plausibility check - is tree file available) - output for reportfile fixed - output of likelihood to clocklike tree in treefile added/fixed 5.1.pl4 21.03.2003 - consensus option added to compute usertree consensus (only consensus is computed and printed to stdin) - usertreefile PREFIX fixed 5.1.pl3 20.03.2003 - TP_MAX_EXP_DIFF=b-a>0 introduced, such that (1.0+exp(a-b) == 1.0) to avoid the computation as well as possible FPE, when a-b gets very small. - flag '-prefix=XXX' added: use 'XXX' as filename prefix instead of 'Infilename' 5.1.pl2 05.12.2002 - PP_Finalized before exit in parallel programs. 5.1.pl1 05.12.2002 - FPE in Brent's algorithm fixed: ()/0, when x=w=v. 5.1 05.11.2002 - Version 5.1 !!! NEW RELEASE =========== 5.1.rc9 02.11.2002 - CPBI datasets (EF.phy, EF.3trees) added to data - manual updates (statistics, Acknowledgments) - latex2html added to Makefile.am - Date changed from September to November 2002 5.1.rc8 24.10.2002 - manual updated: quartet statistics, gap/ambiguous character statistics, 1sKH, 2sKH, SH test, ELW. - fixes in manual (not finished: XXX) 5.1.rc7 24.10.2002 - printversion: ALPHA compiler warning fixed - HHH only for unimplemented stuff 5.1.rc6 15.09.2002 - small fixes: compiler warnings fixed - debug output removed - 2sKH moved to the end of test output (1sKH, SH, ELW, 2sHK) - save site rates ('-wsr') added 5.1.rc5 15.09.2002 - small fixes: printf -> PRINTF 5.1.rc4 10.09.2002 - site likelihoods output to file added (PHYLIP-like) TODO: binary - in preparation:-wslb, -wparam, -usebranch, -rssm, -consensus 5.1.rc3 10.09.2002 - SH test, ELW test added. - one sided KH test, based on pairwise SH added - fflush added to utree output. 5.1.rc2 06.09.2002 - likelihood mapping: simplex labels moved to upper triangle, removed at lower right triangle 5.1.rc1 02.09.2002 - first release candidate for tree-puzzle 5.1 - Changes to the LaTeX manual. CODE FREEZE =========== 5.0.pl30 02.09.2002 - usertree CR bug changed, now accepts trees also in Mac and DOS format (CR or CR+LF). - Q-Support not used in this release (set to FALSE) 5.0.pl29 01.09.2002 - comment added, obsolete/unused functions removed 5.0.pl28 30.08.2002 - Version-no./Copyright changed in the file headers 5.0.pl27 29.08.2002 - Quartet support for inner edges added. Q-Support is computed and added to consensus tree. [todo: Q-Support output in final tree (fputphylogeny)] 5.0.pl26 15.08.2002 - Changes for CodeWarrior added (thanx to Andrew) 5.0.pl25 05.08.2002 - Ambiguous Characters Statistics changed: sum added 5.0.pl24 17.05.2002 - Quartet statistics changed; badqs and badtaxon removed 5.0.pl23 17.05.2002 - find all congruent splits below 50% - output congruent splits below 50% to puzzle file 5.0.pl22 16.05.2002 - cmdline flag '-sub50' added: computation of consensus with splits below 50%, until incongruence or ambiguity 5.0.pl21 13.05.2002 - badtaxon moved from ml.h to puzzle.h and ... - quartet statistics extended from to unresoved (bad) quartets to also fully and partly resolved quartets. - TODO: remove badqs and badtaxon from sources obsolete by fullresqs, partresqs, unresqs, qinfomatr 5.0.pl20 07.05.2002 - a few comments added - '\0' delimited species name array Namestr added for better access. Copied from undelimited Identif after read in. - check for name match while reading in quartets added. exit if mismatch. 5.0.pl19 07.05.2002 - gaps and ambiguous characters added to puzzle report - abort when ambiguous characters = 100% - warning removed from translatedataset, because it ambiguity depends on datatype - might change in menu 5.0.pl18 03.04.2002 - src/test* removed from configure.ac and source tree 5.0.pl17 26.03.2002 - fflush added to evaluatetree after printing tree to treefile 5.0.pl16 20.03.2002 - files adapted for CVS (Id tag added to files) - manual converted from HTML to LaTeX, pdflatex added to Makefile.am 5.0.pl15 19.03.2002 - parameters added (translatedatas,getseqs,getdataset) - all malloc substituted by calloc. 5.0.pl14 15.03.2002 - counting gaps and ambiguous characters in sequence. print warning if > 50 percent. 5.0.pl13 29.02.2002 - exchange O(n^2) chooser by one using O(n) - ml3.c renamed to mlparam.c to collect routines for parameter estimation. - estimateparametersnotree, estimateparameterstree moved to mlparam.c/ml.h - optimseqevolparamsq/optimseqevolparamst renamed to optimseqevolparamsquart/optimseqevolparamstree 5.0.pl12 20.12.2001 - migration to automake 1.5/autoconf 2.52-minor changes - configure changed to use src/config.h - $(srcdir) added to allow build in other directories 5.0.pl11 19.12.2001 - alignment input fixed to accept also Mac format (CR) - removed debug message from 5.0.pl10 - dependencies on Makefile added 5.0.pl10 18.12.2001 - root root search bug, the kept root fixed after first root search. Was introduced in 5.0.pl5. 5.0.pl9 18.12.2001 - possible number of charcters per node in tree strings enlarged from 23/2 to 50. CHAR_PER_TREENODE introduced - typo basian, fixed to Bayesian 5.0.pl8 18.12.2001 - added parameter -achim - fixed doubly opened treefile - adopted to new autoconf/automake (e.g. EXEEXT added) 5.0.pl7 14.11.2001 - computation of rate categories moved to appropriate place in workflow. in some cases categories were computed without likelihoods from tree ready. consequently the output RATE HETEROGENEITY parameter had to move further to the end. - comments added to code - cc added to MPICC (CRAY MPI compiler) 5.0.pl6 20.08.2001 - inputfile name output in the puzzle report for interactive case corrected. was: infile 5.0.pl5 18.08.2001 - minor clock bug corrected, that set the location of the clocked root to first sequence. 5.0.pl4 19.07.2001 - VT matrix corrected, printratematrix added - minor compile time warnings fixed 5.0.pl3 05.07.2001 - radixsort renamed to tp_radixsort to avoid name clashes on Mac OS X - manual replaced by corrected version 5.0.pl2 09.01.2001 - LIBS moved to the end of the linker statement in the Makefile 5.0.pl1 15.11.2000 - Maxspc casted to double for log10 - 200% bug fixed by adding fabs(), log-likelihood differences should not become negative 5.0 26.08.2000 - changes to manual, Makefile.in - cpREV hidden by -DCPREV flag - chi2test, quartio included into source code files - generic scr/Makefile.generic - src/makefile.com for VAX - AUTHORS, README, ChangeLog updated - INSTALL checked 27.08.2000 - test code excluded - '-randseed#' added for debugging purposes - ./data added to autoconf/automake - warning output if cmdline parameter unknown 11.10.2000 - fixed output of rate categories of sites before computing them - check whether rate categories were computed by 1st user tree or NJ tree fixed in the output 12.10.2000 - invariant site model normalization fixed CODE FREEZE =========== 5.0.a33 15.08.2000 - changes for autoconf/automake 5.0.a32 01.08.2000 - a FPE error fixed (badq == 0) - small error in -bestq fixed - fflush's added at several places 5.0.a31 01.08.2000 - comments added to tree structure sorting puzzle2.c - changes in configure.in, Makefile.in 5.0.a30 23.07.2000 - some debugging in checkquart - changed to autoconf 5.0.a29 13.07.2000 - some debugging in checkquart 5.0.a28 13.07.2000 - use best quartet topology option (-bestq) implemented 5.0.a27 13.07.2000 - further developement to checkquart - ascii/binary quartet values (-wqla/-wqlb) - typo correction 5.0.a26 11.07.2000 - fflush at all checktimer - further developement at checkquart - possibility to write quartet values to file (-wqlh) 5.0.a25 06.07.2000 - fflush at checktimer 5.0.a24 02.07.2000 - further debugging of checkquart 5.0.a23 02.07.2000 - further developement to checkquart 5.0.a22 29.06.2000 - checkquart added to makefile - bad quartet stats added after reading in *.allquarts 5.0.a21 27.06.2000 - site pattern statistics implemented and added to SEQUENCE ALIGNMENT section in puzzle report 5.0.a20 26.06.2000 - cpREV45 implemented 5.0.a19 26.06.2000 - for debugging purposes: typo "MPE" changed to "FPE" - fflush(stdout) added in chi2test 5.0.a18 20.06.2000 - checkquart implemented 5.0.a17 19.06.2000 - FPRINTF(STDOUTFILE and STDOUT definition changed and moved; fputid/fputid10 writes to STDOUT instead of stdout - ppuzzle checks slaves enough slave-processes - numquarts, num2quart, quart2num moved from ppuzzle.c to puzzle1.c - read/writeallquart implemented (undocumented feature) to be used by -wqf/-rqf at comandline -wqf = write quartet file (infilename.allquart) after quartet evaluation -rqf = read quartet file (infilename.allquart), no quartet evaluation, unless -wqf is used as well, then quartets are written and read in - '-h' option at comandline -> printusage 5.0.a16 31.05.2000 - chi2test bug fixed - WAG matrix added, model choice adopted 13.06.2000 - date set to June 2000 - author order changed to Schmidt, Strimmer, Vingron, v.Haeseler - CPU time output stopped, due to overflow errors 16.06.2000 - sequence composition chi2test moved before parameter output. - output of chi2test and bad quartet statistics split, to do the chi2test output earlier. 5.0.a15 02.05.2000 - Names changed back from TREE-PUZZLE to PUZZLE 09.05.2000 - and to TREE-PUZZLE again ;-) 5.0.a14 13.03.2000 - Changes to the manual. - Executable names changed to (p)treepuzzle. (changes in the makefiles) 15.03.2000 - Output of parameters after estimation added. 5.0.a13 18.02.2000 - ALPHA version number removed from the code 5.0.a12 18.02.2000 - CPU time measurement problems fixed for case where clock_t is an unsigned type. 5.0.a11 17.02.2000 - time measure problems (CPU/wallclock) fixed not all features in addtimes are used at the moment. - unnecessary and unused routines removed fron source code. 5.0.a10 20.01.2000 - Name changes from PUZZLE to TREE-PUZZLE - Chi2-fit model guessing for VT model added - little model printing bug fixed 5.0.a9 22.12.1999 - VT Model incorporated (Mueller, Vingron (2000) JCB, to appear). - TODO: Chi2-fit model guessing for VT model 5.0.a8 21.12.1999 - 'sys/times.h' and 'sys/types.h' removed from puzzle.h. They were neither ANSI conform nor necessary, but occured in the SUN man pages. - Definition and call of writetimesstat eliminated from the sequention version by compiler switched, and not just the function body as before. - '-O4' canged to '-O' to be more generic. 5.0.a7 21.12.1999 - Macro constants introduced for data_optn (NUCLEOTIDE, AMINOACID, BINARY) - round robbing of datatype and AA model option changed in menu to make adjustment of the model possible by a determined sequence of letters: 'd': Auto -> Nucleotides -> Amino acids -> Binary states -> Auto ('m' && data_optn == AMINOACID): Auto -> Dayhoff -> JTT -> mtREV24 -> BLOSUM62 -> Auto - manual.html adjusted 5.0.a6 20.12.1999 - new manual.html added 5.0.a5 07.12.1999 - output bug fixed (bestrates were written before they were computed) 5.0.a4 02.12.1999 - header file inclusion ajusted: added: #include<string.h> changed from: #include "ppuzzle.h" to: #ifdef PARALLEL # include "ppuzzle.h" #endif 5.0.a3 27.11.1999 - '-h' comandline option removed, because of problems with MPICH under LINUX - new memory leaks of 5.0.a2 closed in PP_Finalize 5.0.a2 27.11.1999 - Cleanup of the source code - Measurement of CPU time added - Parallel load statistics added (quartets, trees, time) to puzzle report. - Cleanup debug messages - Comments "[...]" are removed from usertrees now. - single quotes will only be printed arount species names if -DUSEQUOTES is set at compiletime. - tree likelihood is printed infront of a tree as a comment, [ lh=-xx.xxxxx ](...); 5.0.a1 26.11.1999 - Cleanup of the directories - Copyright changes - Version changes VERSION CHANGE ============== 4.1.a26 25.11.1999 - Makefile made universal for pauzzle and ppuzzle - lines not needed removed from puzzle.h 4.1.a25 19.11.1999 - Output file prefixes for distances, trees, and puzzlereport changed in user trees analysis case to user tree file name - Temporary output of likelihood to treefile added 4.1.a24 11.11.1999 - Output of puzzling step trees changed ptorder: [ orderno # % ID #UniqTopos #Steps ]PHYLIP pstep: chunk #InChunk sum ID #UniqTopos #Steps - preliminary leap frog RNG implemented, i.e. uses the rand4 in the usual way in the sequential case. If run in parallel all rand4 are initialized with the same seed and started with PP_Myid-th random number. after that each process uses the every PP_NumProcs-th random number to make sure that these unique. 4.1.a23 08.11.1999 - output of sequential and parallel version to *.pstep made identical 4.1.a22 05.11.1999 - two different puzzle step tree outputs intruduced and added to the menu ("[ 1. 35 ](...);": - ordered unique tree list -> *.ptorder Format: "[ 1. 35 ]" (Ordernumber, Amount) - chronological tree list -> *.pstep Format: "[ 1. 35 ]" (Chunknumber, Amount in chunk) (the last is a problem in parallel, because the come in chunks, as scheduled) - debugged the output 4.1.a21 04.11.1999 - Makefile adjustments for other Plattforms - pstep tree output changed. unique treestructures printed to *.pstep file with a leading comment containing an order number and the ammount padded with blanks (e.g. "[ 1. 356 ]('mouse'..."). output is done right before writing the puzzle file. - controlled MPI finish to the Quit menu option added 4.1.a20 03.11.1999 - some garbage collection (free) added - makefile adjusted, OFLAGS for optimization added (ppuzzle/MPICH has problems with -O, so the ppuzzle is created without optimization) Some minor changes in the makefiles - still to do: garbage collection from 'internalnode' in master process 4.1.a19 13.10.1999 - adding the output of standardized (i.e. sorted) puzzling step trees. Those are printed to the standard output at the moment. (Routines to sort and print the trees implemented) 14.10.1999 - routines for printing the sorted trees to a string. needed to send them between Master and Worker, and to have a unique key to sort and count the trees. 21.10.1999 - counting of sorted trees implemented by doubly linked list, sort routine, print to stdout 25.10.1999 - change place of writing distances to file right after distances have been computed. - output of puzzling step trees now with true name, not numbers 02.11.1999 - parallel counting and sending of puzzling step trees - some parallel sending bugs fixed 4.1.a18 14.09.1999 - adding possibility to specify input file at command line, this specifies also the output filenames (puzzle output: *.puzzle; treefile: *.tree; distances: *.dist; Triangel EPS: *.eps; unresolved: *.qlist; puzzling step trees: *.pstep) If an unexisting name is given, one has to reenter the right name, but the wrong one is used as prefix. 15.09.1999 - sending back of bad quartets from slaves added - bug in quart2num fixed (not used before; was shifted by 1) - first version of a README added ;-) 4.1.a17 03.08.1999 - Recv-Error in receiving DoPuzzleBlock fixed - double freeing of same MPI_Datatype fixed - changing of scheduling algorithm to smaller chunks in gss -> sgss 13.09.1999 - bug fixed in optimization routine in ml2.c: boundary check added 4.1.a16 12.07.1999 - slight changes in verbosity levels - changed all printf to FPRINTF(STDOUTFILE to change easily from stdout to a file. 4.1.a15 08.07.1999 - scheduler for both parallel parts - several small changes 4.1.a14 25.06.1999 - computation of tree parallel, scheduler dependent, sending all biparts in one message instead of one by one - several small changes since a13 in sched.c, et al. 4.1.a13 10.06.1999 - computation of tree parallel (chunk = #trees/#slaves) - scheduling schemes implemented for minimum chunk sizes 4.1.a12 07.06.1999 - computation of quartets properly parallel - scheduling implemented - counting of quartets by slave ajusted - TODO: sending of bad quartets (array + list) - distinction between '1st user tree' and 'NJ tree' in result output removed again 4.1.a11 28.05.1999 - PP_SendDoQuartBlock, PP_RecvDoQuartBlock, PP_SendQuartBlock, PP_RecvQuartBlock - mallocquartets() changed from global to local variables to be more flexible - Quartet computation moved to slave (badquartet handling missing: output, badquartet vector); - distinction between '1st user tree' and 'NJ tree' added in result output (puzzle1.c around l.1756) 4.1.a10 20.05.1999 - num2quart, numquarts, quart2num introduced - parallel init/finalize, quartets computed on master and slave, compared -> equal -> all necessary parameter exported 4.1.a9 19.05.1999 - 'dvector forg' removed from onepamratematrix cmdline, because it's not used in the function. 4.1.a8 18.05.1999 - add _GAMMA_ (not necessary) to gamma.h and _PUZZLE_ to puzzle.h to avoid dublicate includes, possible due to ppuzzle.h - ppuzzle added to makefile and to check - 1st parallel version but no slave computations only sending parameters and done signals. 4.1.a7 18.05.1999 - export reevaluation of tree and evaluation of usertrees to evaluatetree. 4.1.a6 17.05.1999 - -DNEWFORLOOP added to fixed.src, because the changed for loop structure changes the sequence of randomized quartets during likelihood mapping - change 'int main()' to 'int main(argc, argv)' - export more functionalities from main: memcleanup(), inputandinit(&argc, &argv) - grouping if's (excluding eachother) together in switch() - split treereavaluation and 1st usertree, evaluate all usertrees together (TODO: both, treereavaluation and usertrees in one loop) - MAKE CHECK added to ./makefile 4.1.a5 16.05.1999 - adding ´dvector Brnlength´ to lslength cmdline to reduce globality of Brnlength. (Later better to *Tree) 4.1.a4 11.05.1999 - structure of for loops changed in computeallquartets and recon_tree, so that the quarted addresses are in one contigous sequence (for a<b<c<d). This reduces swapping on low memory machines and is easier to parallelize. 4.1.a3 07.05.1999 - export likelihood mapping from main to map_lklhd() - export tree reconstuction from main to recon_tree() 4.1.a2 06.05.1999 - moving all quartet_(a)lklhd calls to a general purpose routine compute_quartlklhds() 4.1.a1 06.05.1999 - Macro constants for typ_optn and puzzlemode for better readability. - Ring counter for typ_optn and puzzlemode in the menu realized with mod (%) instead of IF. - Possibility added to switch to a fixed random seed (-1) or/and fixed 'random' integers (0) by using -DFIXEDRANDSEED and -DFIXEDINTRAND options.