+string FilterSeqsCommand::createFilter() {
+ try {
+ string filterString = "";
+ Filters F;
+
+ if (soft != 0) { F.setSoft(soft); }
+ if (trump != '*') { F.setTrump(trump); }
+
+ F.setLength(alignmentLength);
+
+ if(soft != 0 || isTrue(vertical)){
+ F.initialize();
+ }
+
+ if(hard.compare("") != 0) { F.doHard(hard); }
+ else { F.setFilter(string(alignmentLength, '1')); }
+
+ numSeqs = 0;
+ if(trump != '*' || isTrue(vertical) || soft != 0){
+ for (int s = 0; s < fastafileNames.size(); s++) {
+
+ for (int i = 0; i < lines.size(); i++) { delete lines[i]; } lines.clear();
+
+#ifdef USE_MPI
+ int pid, numSeqsPerProcessor, num;
+ int tag = 2001;
+ vector<long> MPIPos;
+
+ MPI_Status status;
+ MPI_File inMPI;
+ MPI_Comm_size(MPI_COMM_WORLD, &processors);
+ MPI_Comm_rank(MPI_COMM_WORLD, &pid);
+
+ //char* tempFileName = new char(fastafileNames[s].length());
+ //tempFileName = &(fastafileNames[s][0]);
+
+ char tempFileName[1024];
+ strcpy(tempFileName, fastafileNames[s].c_str());
+
+ MPI_File_open(MPI_COMM_WORLD, tempFileName, MPI_MODE_RDONLY, MPI_INFO_NULL, &inMPI); //comm, filename, mode, info, filepointer
+
+ if (m->control_pressed) { MPI_File_close(&inMPI); return 0; }
+
+ if (pid == 0) { //you are the root process
+ MPIPos = setFilePosFasta(fastafileNames[s], num); //fills MPIPos, returns numSeqs
+ numSeqs += num;
+
+ //send file positions to all processes
+ MPI_Bcast(&num, 1, MPI_INT, 0, MPI_COMM_WORLD); //send numSeqs
+ MPI_Bcast(&MPIPos[0], (num+1), MPI_LONG, 0, MPI_COMM_WORLD); //send file pos
+
+ //figure out how many sequences you have to do
+ numSeqsPerProcessor = num / processors;
+ int startIndex = pid * numSeqsPerProcessor;
+ if(pid == (processors - 1)){ numSeqsPerProcessor = num - pid * numSeqsPerProcessor; }
+
+
+ //do your part
+ MPICreateFilter(startIndex, numSeqsPerProcessor, F, inMPI, MPIPos);
+
+ if (m->control_pressed) { MPI_File_close(&inMPI); return 0; }
+
+ }else { //i am the child process
+
+ MPI_Bcast(&num, 1, MPI_INT, 0, MPI_COMM_WORLD); //get numSeqs
+ MPIPos.resize(num+1);
+ numSeqs += num;
+ MPI_Bcast(&MPIPos[0], (num+1), MPI_LONG, 0, MPI_COMM_WORLD); //get file positions
+
+ //figure out how many sequences you have to align
+ numSeqsPerProcessor = num / processors;
+ int startIndex = pid * numSeqsPerProcessor;
+ if(pid == (processors - 1)){ numSeqsPerProcessor = num - pid * numSeqsPerProcessor; }
+
+
+ //do your part
+ MPICreateFilter(startIndex, numSeqsPerProcessor, F, inMPI, MPIPos);
+
+ if (m->control_pressed) { MPI_File_close(&inMPI); return 0; }
+ }
+
+ MPI_File_close(&inMPI);
+
+#else
+ #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
+ if(processors == 1){
+ ifstream inFASTA;
+ openInputFile(fastafileNames[s], inFASTA);
+ int numFastaSeqs=count(istreambuf_iterator<char>(inFASTA),istreambuf_iterator<char>(), '>');
+ inFASTA.close();
+
+ numSeqs += numFastaSeqs;
+
+ lines.push_back(new linePair(0, numFastaSeqs));
+
+ driverCreateFilter(F, fastafileNames[s], lines[0]);
+ }else{
+ setLines(fastafileNames[s]);
+ createProcessesCreateFilter(F, fastafileNames[s]);
+ }
+
+ if (m->control_pressed) { return filterString; }
+ #else
+ ifstream inFASTA;
+ openInputFile(fastafileNames[s], inFASTA);
+ int numFastaSeqs=count(istreambuf_iterator<char>(inFASTA),istreambuf_iterator<char>(), '>');
+ inFASTA.close();
+
+ numSeqs += numFastaSeqs;
+
+ lines.push_back(new linePair(0, numFastaSeqs));
+
+ driverCreateFilter(F, fastafileNames[s], lines[0]);
+ if (m->control_pressed) { return filterString; }
+ #endif
+#endif
+
+ }
+ }
+
+
+#ifdef USE_MPI
+ int pid;
+ int Atag = 1; int Ttag = 2; int Ctag = 3; int Gtag = 4; int Gaptag = 5;
+ MPI_Status status;
+
+ MPI_Comm_rank(MPI_COMM_WORLD, &pid);
+
+ if(trump != '*' || isTrue(vertical) || soft != 0){
+
+ if (pid == 0) { //only one process should output the filter
+
+ vector<int> temp; temp.resize(alignmentLength+1);
+
+ //get the frequencies from the child processes
+ for(int i = 0; i < ((processors-1)*5); i++) {
+ MPI_Recv(&temp[0], (alignmentLength+1), MPI_INT, MPI_ANY_SOURCE, 2001, MPI_COMM_WORLD, &status);
+ int receiveTag = temp[temp.size()-1]; //child process added a int to the end to indicate what letter count this is for
+
+ if (receiveTag == Atag) { //you are recieveing the A frequencies
+ for (int k = 0; k < alignmentLength; k++) { F.a[k] += temp[k]; }
+ }else if (receiveTag == Ttag) { //you are recieveing the T frequencies
+ for (int k = 0; k < alignmentLength; k++) { F.t[k] += temp[k]; }
+ }else if (receiveTag == Ctag) { //you are recieveing the C frequencies
+ for (int k = 0; k < alignmentLength; k++) { F.c[k] += temp[k]; }
+ }else if (receiveTag == Gtag) { //you are recieveing the G frequencies
+ for (int k = 0; k < alignmentLength; k++) { F.g[k] += temp[k]; }
+ }else if (receiveTag == Gaptag) { //you are recieveing the gap frequencies
+ for (int k = 0; k < alignmentLength; k++) { F.gap[k] += temp[k]; }
+ }
+ }
+ }else{
+
+ //send my fequency counts
+ F.a.push_back(Atag);
+ int ierr = MPI_Send(&(F.a[0]), (alignmentLength+1), MPI_INT, 0, 2001, MPI_COMM_WORLD);
+ F.t.push_back(Ttag);
+ ierr = MPI_Send (&(F.t[0]), (alignmentLength+1), MPI_INT, 0, 2001, MPI_COMM_WORLD);
+ F.c.push_back(Ctag);
+ ierr = MPI_Send(&(F.c[0]), (alignmentLength+1), MPI_INT, 0, 2001, MPI_COMM_WORLD);
+ F.g.push_back(Gtag);
+ ierr = MPI_Send(&(F.g[0]), (alignmentLength+1), MPI_INT, 0, 2001, MPI_COMM_WORLD);
+ F.gap.push_back(Gaptag);
+ ierr = MPI_Send(&(F.gap[0]), (alignmentLength+1), MPI_INT, 0, 2001, MPI_COMM_WORLD);
+ }
+
+ }
+
+ if (pid == 0) { //only one process should output the filter
+#endif
+ F.setNumSeqs(numSeqs);
+
+ if(isTrue(vertical) == 1) { F.doVertical(); }
+ if(soft != 0) { F.doSoft(); }
+
+ filterString = F.getFilter();
+
+#ifdef USE_MPI
+ //send filter string to kids
+ MPI_Bcast(&filterString[0], alignmentLength, MPI_CHAR, 0, MPI_COMM_WORLD);
+ }else{
+ //recieve filterString
+ char* tempBuf = new char[alignmentLength];
+ MPI_Bcast(tempBuf, alignmentLength, MPI_CHAR, 0, MPI_COMM_WORLD);
+
+ filterString = tempBuf;
+ if (filterString.length() > alignmentLength) { filterString = filterString.substr(0, alignmentLength); }
+ delete tempBuf;
+ }