try {
if (abort == true) { return 0; }
- vector<string> outputNames;
ifstream inFASTA;
openInputFile(fastafileNames[0], inFASTA);
inFASTA.close();
////////////create filter/////////////////
+ m->mothurOut("Creating Filter... "); m->mothurOutEndLine();
filter = createFilter();
+ m->mothurOutEndLine(); m->mothurOutEndLine();
+
+ if (m->control_pressed) { return 0; }
+
+ #ifdef USE_MPI
+ int pid;
+ MPI_Comm_rank(MPI_COMM_WORLD, &pid);
+
+ if (pid == 0) { //only one process should output the filter
+ #endif
+
ofstream outFilter;
string filterFile = outputDir + filterFileName + ".filter";
outFilter.close();
outputNames.push_back(filterFile);
+ #ifdef USE_MPI
+ }
+ #endif
////////////run filter/////////////////
- numSeqs = 0;
- for (int i = 0; i < fastafileNames.size(); i++) {
- ifstream in;
- openInputFile(fastafileNames[i], in);
- string filteredFasta = outputDir + getRootName(getSimpleName(fastafileNames[i])) + "filter.fasta";
- ofstream outFASTA;
- openOutputFile(filteredFasta, outFASTA);
- outputNames.push_back(filteredFasta);
-
-
- while(!in.eof()){
- if (m->control_pressed) { in.close(); outFASTA.close(); for(int i = 0; i < outputNames.size(); i++) { remove(outputNames[i].c_str()); } return 0; }
-
- Sequence seq(in);
- if (seq.getName() != "") {
- string align = seq.getAligned();
- string filterSeq = "";
-
- for(int j=0;j<alignmentLength;j++){
- if(filter[j] == '1'){
- filterSeq += align[j];
- }
- }
-
- outFASTA << '>' << seq.getName() << endl << filterSeq << endl;
- numSeqs++;
- }
- gobble(in);
- }
- outFASTA.close();
- in.close();
- }
+ m->mothurOut("Running Filter... "); m->mothurOutEndLine();
+
+ filterSequences();
+ m->mothurOutEndLine(); m->mothurOutEndLine();
+
int filteredLength = 0;
for(int i=0;i<alignmentLength;i++){
if(filter[i] == '1'){ filteredLength++; }
}
}
/**************************************************************************************/
-string FilterSeqsCommand::createFilter() {
+int FilterSeqsCommand::filterSequences() {
try {
- string filterString = "";
+ numSeqs = 0;
+
+ for (int s = 0; s < fastafileNames.size(); s++) {
+
+ for (int i = 0; i < lines.size(); i++) { delete lines[i]; } lines.clear();
+
+ string filteredFasta = outputDir + getRootName(getSimpleName(fastafileNames[s])) + "filter.fasta";
+#ifdef USE_MPI
+ int pid, start, end, numSeqsPerProcessor, num;
+ int tag = 2001;
+ vector<long>MPIPos;
+
+ MPI_Status status;
+ MPI_Comm_size(MPI_COMM_WORLD, &processors); //set processors to the number of mpi processes running
+ MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
+
+ MPI_File outMPI;
+ MPI_File tempMPI;
+ MPI_File inMPI;
+ int outMode=MPI_MODE_CREATE|MPI_MODE_WRONLY;
+ int inMode=MPI_MODE_RDONLY;
+
+ //char* outFilename = new char[filteredFasta.length()];
+ //memcpy(outFilename, filteredFasta.c_str(), filteredFasta.length());
+
+ char outFilename[1024];
+ strcpy(outFilename, filteredFasta.c_str());
+
+ //char* inFileName = new char[fastafileNames[s].length()];
+ //memcpy(inFileName, fastafileNames[s].c_str(), fastafileNames[s].length());
+
+ char inFileName[1024];
+ strcpy(inFileName, fastafileNames[s].c_str());
+
+ MPI_File_open(MPI_COMM_WORLD, inFileName, inMode, MPI_INFO_NULL, &inMPI); //comm, filename, mode, info, filepointer
+ MPI_File_open(MPI_COMM_WORLD, outFilename, outMode, MPI_INFO_NULL, &outMPI);
+
+ //delete inFileName;
+ //delete outFilename;
+
+ if (m->control_pressed) { MPI_File_close(&inMPI); MPI_File_close(&outMPI); 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
+ driverMPIRun(startIndex, numSeqsPerProcessor, inMPI, outMPI, MPIPos);
+
+ if (m->control_pressed) { MPI_File_close(&inMPI); MPI_File_close(&outMPI); return 0; }
+
+ //wait on chidren
+ for(int i = 1; i < processors; i++) {
+ char buf[4];
+ MPI_Recv(buf, 4, MPI_CHAR, i, tag, MPI_COMM_WORLD, &status);
+ }
+
+ }else { //you are a child process
+ MPI_Bcast(&num, 1, MPI_INT, 0, MPI_COMM_WORLD); //get numSeqs
+ numSeqs += num;
+ MPIPos.resize(num+1);
+ 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; }
+
+
+ //align your part
+ driverMPIRun(startIndex, numSeqsPerProcessor, inMPI, outMPI, MPIPos);
+
+ if (m->control_pressed) { MPI_File_close(&inMPI); MPI_File_close(&outMPI); return 0; }
+
+ char buf[4];
+ strcpy(buf, "done");
+
+ //tell parent you are done.
+ MPI_Send(buf, 4, MPI_CHAR, 0, tag, MPI_COMM_WORLD);
+ }
+
+ MPI_File_close(&outMPI);
+ 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();
+
+ lines.push_back(new linePair(0, numFastaSeqs));
+
+ numSeqs += numFastaSeqs;
+
+ driverRunFilter(filter, filteredFasta, fastafileNames[s], lines[0]);
+ }else{
+ setLines(fastafileNames[s]);
+ createProcessesRunFilter(filter, fastafileNames[s]);
+
+ rename((fastafileNames[s] + toString(processIDS[0]) + ".temp").c_str(), filteredFasta.c_str());
+
+ //append fasta files
+ for(int i=1;i<processors;i++){
+ appendFiles((fastafileNames[s] + toString(processIDS[i]) + ".temp"), filteredFasta);
+ remove((fastafileNames[s] + toString(processIDS[i]) + ".temp").c_str());
+ }
+ }
+
+ if (m->control_pressed) { return 1; }
+ #else
+ ifstream inFASTA;
+ openInputFile(fastafileNames[s], inFASTA);
+ int numFastaSeqs=count(istreambuf_iterator<char>(inFASTA),istreambuf_iterator<char>(), '>');
+ inFASTA.close();
+
+ lines.push_back(new linePair(0, numFastaSeqs));
+
+ numSeqs += numFastaSeqs;
+
+ driverRunFilter(filter, filteredFasta, fastafileNames[s], lines[0]);
+
+ if (m->control_pressed) { return 1; }
+ #endif
+#endif
+ outputNames.push_back(filteredFasta);
+ }
+
+ return 0;
+ }
+ catch(exception& e) {
+ m->errorOut(e, "FilterSeqsCommand", "filterSequences");
+ exit(1);
+ }
+}
+#ifdef USE_MPI
+/**************************************************************************************/
+int FilterSeqsCommand::driverMPIRun(int start, int num, MPI_File& inMPI, MPI_File& outMPI, vector<long>& MPIPos) {
+ try {
+ string outputString = "";
+ int count = 0;
+ MPI_Status status;
+
+ for(int i=0;i<num;i++){
+
+ if (m->control_pressed) { return 0; }
+
+ //read next sequence
+ int length = MPIPos[start+i+1] - MPIPos[start+i];
+ char* buf4 = new char[length];
+ MPI_File_read_at(inMPI, MPIPos[start+i], buf4, length, MPI_CHAR, &status);
+
+ string tempBuf = buf4;
+ if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length); }
+ istringstream iss (tempBuf,istringstream::in);
+ delete buf4;
+
+ Sequence seq(iss); gobble(iss);
+
+ if (seq.getName() != "") {
+ string align = seq.getAligned();
+ string filterSeq = "";
+
+ for(int j=0;j<alignmentLength;j++){
+ if(filter[j] == '1'){
+ filterSeq += align[j];
+ }
+ }
+
+ count++;
+ outputString += ">" + seq.getName() + "\n" + filterSeq + "\n";
+
+ if(count % 10 == 0){ //output to file
+ //send results to parent
+ int length = outputString.length();
+ char* buf = new char[length];
+ memcpy(buf, outputString.c_str(), length);
+
+ MPI_File_write_shared(outMPI, buf, length, MPI_CHAR, &status);
+ outputString = "";
+ delete buf;
+ }
+
+ }
+
+ if((i+1) % 100 == 0){ cout << (i+1) << endl; m->mothurOutJustToLog(toString(i+1) + "\n"); }
+ }
+
+ if(outputString != ""){ //output to file
+ //send results to parent
+ int length = outputString.length();
+ char* buf = new char[length];
+ memcpy(buf, outputString.c_str(), length);
+
+ MPI_File_write_shared(outMPI, buf, length, MPI_CHAR, &status);
+ outputString = "";
+ delete buf;
+ }
+
+ if((num) % 100 != 0){ cout << (num) << endl; m->mothurOutJustToLog(toString(num) + "\n"); }
+
+ return 0;
+ }
+ catch(exception& e) {
+ m->errorOut(e, "FilterSeqsCommand", "driverRunFilter");
+ exit(1);
+ }
+}
+#endif
+/**************************************************************************************/
+int FilterSeqsCommand::driverRunFilter(string F, string outputFilename, string inputFilename, linePair* line) {
+ try {
+ ofstream out;
+ openOutputFile(outputFilename, out);
+
+ ifstream in;
+ openInputFile(inputFilename, in);
+
+ in.seekg(line->start);
+
+ for(int i=0;i<line->num;i++){
+
+ if (m->control_pressed) { in.close(); out.close(); return 0; }
+
+ Sequence seq(in);
+ if (seq.getName() != "") {
+ string align = seq.getAligned();
+ string filterSeq = "";
+
+ for(int j=0;j<alignmentLength;j++){
+ if(filter[j] == '1'){
+ filterSeq += align[j];
+ }
+ }
+
+ out << '>' << seq.getName() << endl << filterSeq << endl;
+ }
+ gobble(in);
+ }
+ out.close();
+ in.close();
+
+ return 0;
+ }
+ catch(exception& e) {
+ m->errorOut(e, "FilterSeqsCommand", "driverRunFilter");
+ exit(1);
+ }
+}
+/**************************************************************************************************/
+
+int FilterSeqsCommand::createProcessesRunFilter(string F, string filename) {
+ try {
+#if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
+ int process = 0;
+ int exitCommand = 1;
+ processIDS.clear();
+
+ //loop through and create all the processes you want
+ while (process != processors) {
+ int pid = fork();
+
+ if (pid > 0) {
+ processIDS.push_back(pid); //create map from line number to pid so you can append files in correct order later
+ process++;
+ }else if (pid == 0){
+ string filteredFasta = filename + toString(getpid()) + ".temp";
+ driverRunFilter(F, filteredFasta, filename, lines[process]);
+ exit(0);
+ }else { m->mothurOut("unable to spawn the necessary processes."); m->mothurOutEndLine(); exit(0); }
+ }
+
+ //force parent to wait until all the processes are done
+ for (int i=0;i<processors;i++) {
+ int temp = processIDS[i];
+ wait(&temp);
+ }
+
+ return exitCommand;
+#endif
+ }
+ catch(exception& e) {
+ m->errorOut(e, "FilterSeqsCommand", "createProcessesRunFilter");
+ exit(1);
+ }
+}
+/**************************************************************************************/
+string FilterSeqsCommand::createFilter() {
+ try {
+ string filterString = "";
Filters F;
if (soft != 0) { F.setSoft(soft); }
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, rc, ierr;
- int Atag = 1; int Ttag = 2; int Ctag = 3; int Gtag = 4; int Gaptag = 5;
+ int pid, numSeqsPerProcessor, num;
int tag = 2001;
+ vector<long> MPIPos;
MPI_Status status;
- MPI_File in;
- rc = MPI_Comm_size(MPI_COMM_WORLD, &processors);
- rc = MPI_Comm_rank(MPI_COMM_WORLD, &pid);
-
+ 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 = 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
- MPI_File_open(MPI_COMM_WORLD, tempFileName, MPI_MODE_RDONLY, MPI_INFO_NULL, &in); //comm, filename, mode, info, filepointer
+ if (m->control_pressed) { MPI_File_close(&inMPI); return 0; }
if (pid == 0) { //you are the root process
- setLines(fastafileNames[s]);
+ MPIPos = setFilePosFasta(fastafileNames[s], num); //fills MPIPos, returns numSeqs
+ numSeqs += num;
- for (int j = 0; j < lines.size(); j++) { //each process
- if (j != 0) { //don't send to yourself
- MPI_Send(&lines[j]->start, 1, MPI_INT, j, tag, MPI_COMM_WORLD); //start position in file
- MPI_Send(&lines[j]->numSeqs, 1, MPI_INT, j, tag, MPI_COMM_WORLD); //how many sequences we are sending
- MPI_Send(&bufferSizes[j], 1, MPI_INT, j, tag, MPI_COMM_WORLD); //how bytes for the read
- }
- }
-
- char buf[bufferSizes[0]];
- MPI_File_read_at(in, 0, buf, bufferSizes[0], MPI_CHAR, &status);
-
- MPICreateFilter(F, buf);
-
- vector<int> temp; temp.resize(alignmentLength+1);
+ //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; }
- //get the frequencies from the child processes
- for(int i = 0; i < ((processors-1)*5); i++) {
- cout << "i = " << i << endl;
- //vector<int> trial; trial.resize(10);
- //cout << "trials size = " << trial.size() << endl;
- //int ierr = MPI_Recv(&trial[0], 10, MPI_INT, MPI_ANY_SOURCE, tag, MPI_COMM_WORLD, &status);
- int ierr = MPI_Recv(&temp[0], (alignmentLength+1), MPI_INT, MPI_ANY_SOURCE, tag, MPI_COMM_WORLD, &status);
- cout << "recieved something" << endl;
- //for (int g = 0; g < trial.size(); g++) { cout << trial[g] << '\t'; } cout << endl;
- int receiveTag = temp[temp.size()-1]; //child process added a int to the end to indicate what letter count this is for
- cout << "reciveve tag = " << receiveTag << endl;
- for (int k = 0; k < alignmentLength; k++) { cout << k << '\t' << temp[k] << endl; }
- cout << "done " << endl << endl;
-
- int sender = status.MPI_SOURCE;
-
- 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]; }
- }
-
- m->mothurOut("receive tag = " + toString(receiveTag) + " " + toString(sender) + " is complete."); m->mothurOutEndLine();
- }
-
+
+ //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
- int startPos, numLines, bufferSize;
- ierr = MPI_Recv(&startPos, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status);
- ierr = MPI_Recv(&numLines, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status);
- ierr = MPI_Recv(&bufferSize, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status);
-
- //send freqs
- char buf2[bufferSize];
- MPI_File_read_at( in, startPos, buf2, bufferSize, MPI_CHAR, &status);
-
- MPICreateFilter(F, buf2);
-
- //send my fequency counts
- F.a.push_back(Atag);
- int ierr = MPI_Send(&(F.a[0]), (alignmentLength+1), MPI_INT, 0, tag, MPI_COMM_WORLD);
- F.t.push_back(Ttag);
- ierr = MPI_Send (&(F.t[0]), (alignmentLength+1), MPI_INT, 0, tag, MPI_COMM_WORLD);
- F.c.push_back(Ctag);
- ierr = MPI_Send(&(F.c[0]), (alignmentLength+1), MPI_INT, 0, tag, MPI_COMM_WORLD);
- F.g.push_back(Gtag);
- ierr = MPI_Send(&(F.g[0]), (alignmentLength+1), MPI_INT, 0, tag, MPI_COMM_WORLD);
- F.gap.push_back(Gaptag);
- ierr = MPI_Send(&(F.gap[0]), (alignmentLength+1), MPI_INT, 0, tag, MPI_COMM_WORLD);
+
+ 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_Barrier(MPI_COMM_WORLD);
+ MPI_File_close(&inMPI);
#else
#if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
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);
lines.push_back(new linePair(0, numFastaSeqs));
- driverCreateFilter(F, lines[0], fastafileNames[s]);
+ driverCreateFilter(F, fastafileNames[s], lines[0]);
+ if (m->control_pressed) { return filterString; }
#endif
#endif
}
- cout << "made it here, numSeqs = " << numSeqs << endl;
+#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(); }
-//cout << "Filter String = " << F.getFilter() << endl;
+
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;
+ }
+
+ MPI_Barrier(MPI_COMM_WORLD);
+#endif
+
+
return filterString;
}
catch(exception& e) {
in.seekg(line->start);
- for(int i=0;i<line->numSeqs;i++){
+ for(int i=0;i<line->num;i++){
if (m->control_pressed) { in.close(); return 1; }
Sequence seq(in);
if (seq.getName() != "") {
+ if (seq.getAligned().length() != alignmentLength) { m->mothurOut("Sequences are not all the same length, please correct."); m->mothurOutEndLine(); m->control_pressed = true; }
+
if(trump != '*'){ F.doTrump(seq); }
if(isTrue(vertical) || soft != 0){ F.getFreqs(seq); }
cout.flush();
}
//report progress
- if((line->numSeqs) % 100 != 0){ m->mothurOut(toString(line->numSeqs)); m->mothurOutEndLine(); }
+ if((line->num) % 100 != 0){ m->mothurOut(toString(line->num)); m->mothurOutEndLine(); }
in.close();
exit(1);
}
}
+#ifdef USE_MPI
/**************************************************************************************/
-int FilterSeqsCommand::MPICreateFilter(Filters& F, string input) {
+int FilterSeqsCommand::MPICreateFilter(int start, int num, Filters& F, MPI_File& inMPI, vector<long>& MPIPos) {
try {
- vector<string> seqStrings;
- parseBuffer(input, seqStrings);
+ MPI_Status status;
+ int pid;
+ MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
- for(int i=0;i<seqStrings.size();i++){
-
- if (m->control_pressed) { return 1; }
+ for(int i=0;i<num;i++){
- Sequence seq("", seqStrings[i]);
+ if (m->control_pressed) { return 0; }
+
+ //read next sequence
+ int length = MPIPos[start+i+1] - MPIPos[start+i];
+
+ char* buf4 = new char[length];
+ MPI_File_read_at(inMPI, MPIPos[start+i], buf4, length, MPI_CHAR, &status);
+
+ string tempBuf = buf4;
+ if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length); }
+ istringstream iss (tempBuf,istringstream::in);
+ delete buf4;
+
+ Sequence seq(iss);
+
+ if (seq.getAligned().length() != alignmentLength) { cout << "Alignment length is " << alignmentLength << " and sequence " << seq.getName() << " has length " << seq.getAligned().length() << ", please correct." << endl; exit(1); }
if(trump != '*'){ F.doTrump(seq); }
if(isTrue(vertical) || soft != 0){ F.getFreqs(seq); }
cout.flush();
//report progress
- if((i+1) % 100 == 0){ m->mothurOut(toString(i+1)); m->mothurOutEndLine(); }
+ if((i+1) % 100 == 0){ cout << (i+1) << endl; m->mothurOutJustToLog(toString(i+1) + "\n"); }
}
//report progress
- if((seqStrings.size()) % 100 != 0){ m->mothurOut(toString(seqStrings.size())); m->mothurOutEndLine(); }
+ if((num) % 100 != 0){ cout << num << endl; m->mothurOutJustToLog(toString(num) + "\n"); }
return 0;
}
exit(1);
}
}
-
+#endif
/**************************************************************************************************/
int FilterSeqsCommand::createProcessesCreateFilter(Filters& F, string filename) {
#if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
int process = 0;
int exitCommand = 1;
- vector<int> processIDS;
+ processIDS.clear();
//loop through and create all the processes you want
while (process != processors) {
int FilterSeqsCommand::setLines(string filename) {
try {
- vector<int> positions;
- map<int, int> buf;
+
+ vector<long int> positions;
bufferSizes.clear();
- int pid;
- MPI_Comm_rank(MPI_COMM_WORLD, &pid);
-
ifstream inFASTA;
openInputFile(filename, inFASTA);
string input;
- int numbuf = 0;
while(!inFASTA.eof()){
input = getline(inFASTA);
if (input.length() != 0) {
- numbuf += input.length();
- if(input[0] == '>'){ long int pos = inFASTA.tellg(); positions.push_back(pos - input.length() - 1); buf[(pos - input.length() - 1)] = numbuf; }
+ if(input[0] == '>'){ long int pos = inFASTA.tellg(); positions.push_back(pos - input.length() - 1); }
}
}
-
inFASTA.close();
int numFastaSeqs = positions.size();
-
+
+ FILE * pFile;
+ long size;
+
+ //get num bytes in file
+ pFile = fopen (filename.c_str(),"rb");
+ if (pFile==NULL) perror ("Error opening file");
+ else{
+ fseek (pFile, 0, SEEK_END);
+ size=ftell (pFile);
+ fclose (pFile);
+ }
+
numSeqs += numFastaSeqs;
int numSeqsPerProcessor = numFastaSeqs / processors;
long int startPos = positions[ i * numSeqsPerProcessor ];
if(i == processors - 1){
numSeqsPerProcessor = numFastaSeqs - i * numSeqsPerProcessor;
- bufferSizes.push_back(numbuf-buf[startPos]);
+ bufferSizes.push_back(size - startPos);
}else{
- int myEnd = positions[ (i+1) * numSeqsPerProcessor ];
- bufferSizes.push_back(buf[myEnd]-buf[startPos]);
+ long int myEnd = positions[ (i+1) * numSeqsPerProcessor ];
+ bufferSizes.push_back(myEnd-startPos);
}
lines.push_back(new linePair(startPos, numSeqsPerProcessor));
}
exit(1);
}
}
-/**************************************************************************************************/
-int FilterSeqsCommand::parseBuffer(string file, vector<string>& seqs) {
- try {
-
- istringstream iss (file,istringstream::in);
- string name, seqstring;
-
- while (iss) {
-
- if (m->control_pressed) { return 0; }
-
- Sequence seq(iss); gobble(iss);
-
- if (seq.getName() != "") {
- seqs.push_back(seq.getAligned());
- }
- }
-
- return 0;
- }
- catch(exception& e) {
- m->errorOut(e, "FilterSeqsCommand", "parseBuffer");
- exit(1);
- }
-}
/**************************************************************************************/
projects / mothur.git / blobdiff