int FilterSeqsCommand::filterSequences() {
try {
numSeqs = 0;
for (int s = 0; s < fastafileNames.size(); s++) {
for (int i = 0; i < lines.size(); i++) { delete lines[i]; } lines.clear();
map<string, string> variables;
variables["[filename]"] = outputDir + m->getRootName(m->getSimpleName(fastafileNames[s]));
string filteredFasta = getOutputFileName("fasta", variables);
vector<unsigned long long> positions;
if (savedPositions.size() != 0) { positions = savedPositions[s]; }
else {
#if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) || (__linux__) || (__unix__) || (__unix)
positions = m->divideFile(fastafileNames[s], processors);
#else
if(processors != 1){
int numFastaSeqs = 0;
positions = m->setFilePosFasta(fastafileNames[s], numFastaSeqs);
if (numFastaSeqs < processors) { processors = numFastaSeqs; }
}
#endif
}
#if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) || (__linux__) || (__unix__) || (__unix)
//vector<unsigned long long> positions = m->divideFile(fastafileNames[s], processors);
for (int i = 0; i < (positions.size()-1); i++) {
lines.push_back(new linePair(positions[i], positions[(i+1)]));
}
if(processors == 1){
int numFastaSeqs = driverRunFilter(filter, filteredFasta, fastafileNames[s], lines[0]);
numSeqs += numFastaSeqs;
}else{
int numFastaSeqs = createProcessesRunFilter(filter, fastafileNames[s], filteredFasta);
numSeqs += numFastaSeqs;
}
if (m->control_pressed) { return 1; }
#else
if(processors == 1){
lines.push_back(new linePair(0, 1000));
int numFastaSeqs = driverRunFilter(filter, filteredFasta, fastafileNames[s], lines[0]);
numSeqs += numFastaSeqs;
}else {
int numFastaSeqs = positions.size()-1;
//positions = m->setFilePosFasta(fastafileNames[s], numFastaSeqs);
//figure out how many sequences you have to process
int numSeqsPerProcessor = numFastaSeqs / processors;
for (int i = 0; i < processors; i++) {
int startIndex = i * numSeqsPerProcessor;
if(i == (processors - 1)){ numSeqsPerProcessor = numFastaSeqs - i * numSeqsPerProcessor; }
lines.push_back(new linePair(positions[startIndex], numSeqsPerProcessor));
}
numFastaSeqs = createProcessesRunFilter(filter, fastafileNames[s], filteredFasta);
numSeqs += numFastaSeqs;
}
if (m->control_pressed) { return 1; }
#endif
outputNames.push_back(filteredFasta); outputTypes["fasta"].push_back(filteredFasta);
}
return 0;
}
catch(exception& e) {
m->errorOut(e, "FilterSeqsCommand", "filterSequences");
exit(1);
}
}
int FilterSeqsCommand::createProcessesRunFilter(string F, string filename, string filteredFastaName) {
try {
int process = 1;
int num = 0;
processIDS.clear();
bool recalc = false;
#if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) || (__linux__) || (__unix__) || (__unix)
//loop through and create all the processes you want
while (process != processors) {
pid_t 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 + m->mothurGetpid(process) + ".temp";
num = driverRunFilter(F, filteredFasta, filename, lines[process]);
//pass numSeqs to parent
ofstream out;
string tempFile = filename + m->mothurGetpid(process) + ".num.temp";
m->openOutputFile(tempFile, out);
out << num << endl;
out.close();
exit(0);
}else {
m->mothurOut("[ERROR]: unable to spawn the number of processes you requested, reducing number to " + toString(process) + "\n"); processors = process;
for (int i = 0; i < processIDS.size(); i++) { kill (processIDS[i], SIGINT); }
//wait to die
for (int i=0;i<processIDS.size();i++) {
int temp = processIDS[i];
wait(&temp);
}
m->control_pressed = false;
for (int i=0;i<processIDS.size();i++) {
m->mothurRemove(filename + (toString(processIDS[i]) + ".temp"));
m->mothurRemove(filename + (toString(processIDS[i]) + ".num.temp"));
}
recalc = true;
break;
}
}
if (recalc) {
//test line, also set recalc to true.
//for (int i = 0; i < processIDS.size(); i++) { kill (processIDS[i], SIGINT); } for (int i=0;i<processIDS.size();i++) { int temp = processIDS[i]; wait(&temp); } m->control_pressed = false; for (int i=0;i<processIDS.size();i++) {m->mothurRemove(filename + (toString(processIDS[i]) + ".temp"));m->mothurRemove(filename + (toString(processIDS[i]) + ".num.temp"));}processors=3; m->mothurOut("[ERROR]: unable to spawn the number of processes you requested, reducing number to " + toString(processors) + "\n");
//redo file divide
for (int i = 0; i < lines.size(); i++) { delete lines[i]; } lines.clear();
vector<unsigned long long> positions = m->divideFile(filename, processors);
for (int i = 0; i < (positions.size()-1); i++) { lines.push_back(new linePair(positions[i], positions[(i+1)])); }
num = 0;
processIDS.resize(0);
process = 1;
//loop through and create all the processes you want
while (process != processors) {
pid_t 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 + m->mothurGetpid(process) + ".temp";
num = driverRunFilter(F, filteredFasta, filename, lines[process]);
//pass numSeqs to parent
ofstream out;
string tempFile = filename + m->mothurGetpid(process) + ".num.temp";
m->openOutputFile(tempFile, out);
out << num << endl;
out.close();
exit(0);
}else {
m->mothurOut("[ERROR]: unable to spawn the necessary processes."); m->mothurOutEndLine();
for (int i = 0; i < processIDS.size(); i++) { kill (processIDS[i], SIGINT); }
exit(0);
}
}
}
num = driverRunFilter(F, filteredFastaName, filename, lines[0]);
//force parent to wait until all the processes are done
for (int i=0;i<processIDS.size();i++) {
int temp = processIDS[i];
wait(&temp);
}
for (int i = 0; i < processIDS.size(); i++) {
ifstream in;
string tempFile = filename + toString(processIDS[i]) + ".num.temp";
m->openInputFile(tempFile, in);
if (!in.eof()) { int tempNum = 0; in >> tempNum; num += tempNum; }
in.close(); m->mothurRemove(tempFile);
m->appendFiles((filename + toString(processIDS[i]) + ".temp"), filteredFastaName);
m->mothurRemove((filename + toString(processIDS[i]) + ".temp"));
}
#else
//////////////////////////////////////////////////////////////////////////////////////////////////////
//Windows version shared memory, so be careful when passing variables through the filterData struct.
//Above fork() will clone, so memory is separate, but that's not the case with windows,
//Taking advantage of shared memory to allow both threads to add info to F.
//////////////////////////////////////////////////////////////////////////////////////////////////////
vector<filterRunData*> pDataArray;
DWORD dwThreadIdArray[processors-1];
HANDLE hThreadArray[processors-1];
//Create processor worker threads.
for( int i=0; i<processors-1; i++){
string extension = "";
if (i != 0) { extension = toString(i) + ".temp"; }
filterRunData* tempFilter = new filterRunData(filter, filename, (filteredFastaName + extension), m, lines[i]->start, lines[i]->end, alignmentLength, i);
pDataArray.push_back(tempFilter);
processIDS.push_back(i);
hThreadArray[i] = CreateThread(NULL, 0, MyRunFilterThreadFunction, pDataArray[i], 0, &dwThreadIdArray[i]);
}
num = driverRunFilter(F, (filteredFastaName + toString(processors-1) + ".temp"), filename, lines[processors-1]);
//Wait until all threads have terminated.
WaitForMultipleObjects(processors-1, hThreadArray, TRUE, INFINITE);
//Close all thread handles and free memory allocations.
for(int i=0; i < pDataArray.size(); i++){
num += pDataArray[i]->count;
if (pDataArray[i]->count != pDataArray[i]->end) {
m->mothurOut("[ERROR]: process " + toString(i) + " only processed " + toString(pDataArray[i]->count) + " of " + toString(pDataArray[i]->end) + " sequences assigned to it, quitting. \n"); m->control_pressed = true;
}
CloseHandle(hThreadArray[i]);
delete pDataArray[i];
}
for (int i = 1; i < processors; i++) {
m->appendFiles((filteredFastaName + toString(i) + ".temp"), filteredFastaName);
m->mothurRemove((filteredFastaName + toString(i) + ".temp"));
}
#endif
return num;
}
int FilterSeqsCommand::filterSequences() {
try {
numSeqs = 0;
for (int s = 0; s < fastafileNames.size(); s++) {
for (int i = 0; i < lines.size(); i++) { delete lines[i]; } lines.clear();
map<string, string> variables;
variables["[filename]"] = outputDir + m->getRootName(m->getSimpleName(fastafileNames[s]));
string filteredFasta = getOutputFileName("fasta", variables);
#ifdef USE_MPI
int pid, numSeqsPerProcessor, num;
int tag = 2001;
vector<unsigned long 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 inMPI;
int outMode=MPI_MODE_CREATE|MPI_MODE_WRONLY;
int inMode=MPI_MODE_RDONLY;
char outFilename[1024];
strcpy(outFilename, filteredFasta.c_str());
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);
if (m->control_pressed) { MPI_File_close(&inMPI); MPI_File_close(&outMPI); return 0; }
if (pid == 0) { //you are the root process
MPIPos = m->setFilePosFasta(fastafileNames[s], num); //fills MPIPos, returns numSeqs
numSeqs += num;
//send file positions to all processes
for(int i = 1; i < processors; i++) {
MPI_Send(&num, 1, MPI_INT, i, tag, MPI_COMM_WORLD);
MPI_Send(&MPIPos[0], (num+1), MPI_LONG, i, tag, MPI_COMM_WORLD);
}
//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[5];
MPI_Recv(buf, 5, MPI_CHAR, i, tag, MPI_COMM_WORLD, &status);
}
}else { //you are a child process
MPI_Recv(&num, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status);
MPIPos.resize(num+1);
numSeqs += num;
MPI_Recv(&MPIPos[0], (num+1), MPI_LONG, 0, tag, MPI_COMM_WORLD, &status);
//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[5];
strcpy(buf, "done");
//tell parent you are done.
MPI_Send(buf, 5, MPI_CHAR, 0, tag, MPI_COMM_WORLD);
}
MPI_File_close(&outMPI);
MPI_File_close(&inMPI);
MPI_Barrier(MPI_COMM_WORLD); //make everyone wait - just in case
#else
vector<unsigned long long> positions;
if (savedPositions.size() != 0) { positions = savedPositions[s]; }
else {
#if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) || (__linux__) || (__unix__) || (__unix)
positions = m->divideFile(fastafileNames[s], processors);
#else
if(processors != 1){
int numFastaSeqs = 0;
positions = m->setFilePosFasta(fastafileNames[s], numFastaSeqs);
if (positions.size() < processors) { processors = positions.size(); }
}
#endif
}
#if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) || (__linux__) || (__unix__) || (__unix)
//vector<unsigned long long> positions = m->divideFile(fastafileNames[s], processors);
for (int i = 0; i < (positions.size()-1); i++) {
lines.push_back(new linePair(positions[i], positions[(i+1)]));
}
if(processors == 1){
int numFastaSeqs = driverRunFilter(filter, filteredFasta, fastafileNames[s], lines[0]);
numSeqs += numFastaSeqs;
}else{
int numFastaSeqs = createProcessesRunFilter(filter, fastafileNames[s], filteredFasta);
numSeqs += numFastaSeqs;
}
if (m->control_pressed) { return 1; }
#else
if(processors == 1){
lines.push_back(new linePair(0, 1000));
int numFastaSeqs = driverRunFilter(filter, filteredFasta, fastafileNames[s], lines[0]);
numSeqs += numFastaSeqs;
}else {
int numFastaSeqs = positions.size()-1;
//positions = m->setFilePosFasta(fastafileNames[s], numFastaSeqs);
//figure out how many sequences you have to process
int numSeqsPerProcessor = numFastaSeqs / processors;
for (int i = 0; i < processors; i++) {
int startIndex = i * numSeqsPerProcessor;
if(i == (processors - 1)){ numSeqsPerProcessor = numFastaSeqs - i * numSeqsPerProcessor; }
lines.push_back(new linePair(positions[startIndex], numSeqsPerProcessor));
}
numFastaSeqs = createProcessesRunFilter(filter, fastafileNames[s], filteredFasta);
numSeqs += numFastaSeqs;
}
if (m->control_pressed) { return 1; }
#endif
#endif
outputNames.push_back(filteredFasta); outputTypes["fasta"].push_back(filteredFasta);
}
return 0;
}
catch(exception& e) {
m->errorOut(e, "FilterSeqsCommand", "filterSequences");
exit(1);
}
}
