int main(int argc, char ** argv)
{
try {
const std::string description =
"Parallel Empirical Variogram: A massively parallel solution "
"for computing empirical variograms of large-scale data.";
const std::string input_desc = "Path to input file";
const std::string output_desc = "Path to output file. Defaults to stdout.";
const std::string timing_desc = "Path to timing file. "
"Defaults to not writing timing information.";
const std::string repl_desc = "Desired replication factor. Defaults to 1.";
const std::string bins_desc = "Number of distance bins to use for empirical variogram. "
"Defaults to 15.";
const std::string format_desc = "Prints information about output formats instead of performing computations.";
// TODO: Improve input file descriptions to describe output formats
TCLAP::CmdLine cmd(description, ' ', "0.1");
TCLAP::ValueArg<std::string> input_file_arg("i", "input", input_desc, true, "", "path");
TCLAP::ValueArg<std::string> output_file_arg("o", "output", output_desc, false, "", "path");
TCLAP::ValueArg<std::string> timing_file_arg("t", "timing", timing_desc, false, "", "path");
// TCLAP::ValueArg<int> repl_factor_arg("c", "replication", repl_desc,
// false, 1, &positive_constraint);
TCLAP::ValueArg<int> num_bins_arg("b", "bins", bins_desc, false, 15, &positive_constraint);
TCLAP::SwitchArg format_arg("p", "print-formats", format_desc, false);
// We only want EITHER input file or format
cmd.xorAdd(input_file_arg, format_arg);
cmd.add(output_file_arg);
cmd.add(timing_file_arg);
// cmd.add(repl_factor_arg);
cmd.add(num_bins_arg);
cmd.parse(argc, argv);
const int num_bins = num_bins_arg.getValue();
// const int replication_factor = repl_factor_arg.getValue();
const std::string input_path = input_file_arg.getValue();
const std::string output_path = output_file_arg.getValue();
const std::string timing_path = timing_file_arg.getValue();
MPI_Init(&argc, &argv);
int rank, ranks;
MPI_Comm_rank(MPI_COMM_WORLD, &rank);
MPI_Comm_size(MPI_COMM_WORLD, &ranks);
if (format_arg.isSet())
{
if (rank == 0)
print_formats();
MPI_Finalize();
return 0;
}
pev::parallel_options options(ranks, 1);
if (rank == 0)
{
std::cout << "Using " << options.active_processor_count() << " of "
<< ranks << " processors with replication factor " << options.replication_factor()
<< std::endl;
}
pev::variogram_data variogram = pev::empirical_variogram_parallel(input_path, num_bins);
if (rank == 0)
{
std::ofstream outfile;
if (output_file_arg.isSet())
outfile.open(output_path);
// Print to stdout if output file not supplied
std::ostream & out = output_file_arg.isSet()
? outfile
: std::cout;
pev::print_variogram(out, variogram);
if (timing_file_arg.isSet())
{
std::ofstream outfile;
// Append to timing file if path is given
if (timing_path != "-")
outfile.open(timing_path, std::ios_base::out | std::ios_base::app);
std::ostream & out = timing_path == "-"
? std::cout
: outfile;
pev::print_timing_info(out, variogram);
}
}
MPI_Finalize();
return 0;
} catch (const TCLAP::ArgException & e)
{
std::cerr << "ERROR: " << e.what() << " for argument " << e.argId() << std::endl;
return -1;
}
}
int main (int argc, char *argv[]) {
/* Program Option parsing */
int num_frames = 1;
int threads = 1;
bool disable_output = false;
bool colourise_quantiser = false;
bool colourise_padding = false;
bool colourise_unpadded = false;
bool quartersize = false;
bool verbose = false;
std::string input_filename;
std::string output_filename;
try {
TCLAP::CmdLine cmd("VC2 HQ profile Decoder Example\n"
"All input files must be vc2 streams\n"
"All output files will be yuv422p10le\n", '=', "0.1", true);
TCLAP::SwitchArg verbose_arg ("v", "verbose", "verbose mode", cmd, false);
TCLAP::ValueArg<int> num_frames_arg ("n", "num-frames", "Number of frames to decode", false, 1, "integer", cmd);
TCLAP::ValueArg<int> num_threads_arg ("t", "threads", "Number of threads", false, 1, "integer", cmd);
TCLAP::SwitchArg disable_output_args ("d", "disable-output", "disable output", cmd, false);
TCLAP::SwitchArg colourise_quantiser_args("q", "colourise-quantiser", "colourise based on quantiser levels", cmd, false);
TCLAP::SwitchArg colourise_padding_args ("p", "colourise-padding", "colourise based on padding levels", cmd, false);
TCLAP::SwitchArg colourise_unpadded_args ("u", "colourise-unpadded", "colourise based on lack of padding", cmd, false);
TCLAP::UnlabeledValueArg<std::string> input_file_arg("input_file", "encoded input file", true, "", "string", cmd);
TCLAP::UnlabeledValueArg<std::string> output_file_arg("output_file", "output file (defaults to input file + .yuv)", false, "", "string", cmd);
cmd.parse( argc, argv );
num_frames = num_frames_arg.getValue();
threads = num_threads_arg.getValue();
disable_output = disable_output_args.getValue();
colourise_quantiser = colourise_quantiser_args.getValue();
colourise_padding = colourise_padding_args.getValue();
colourise_unpadded = colourise_unpadded_args.getValue();
verbose = verbose_arg.getValue();
input_filename = input_file_arg.getValue();
output_filename = output_file_arg.getValue();
} catch (TCLAP::ArgException &e) {
std::cerr << "error: " << e.error() << " for arg " << e.argId() << std::endl; return 1;
}
if (output_filename == "")
output_filename = input_filename + ".yuv";
/* Variables for storing input and output buffers */
char *idata;
size_t input_length = 0;
uint16_t **odata = new uint16_t*[num_frames]; // Output frame pointers
uint16_t ***opics = new uint16_t**[num_frames*2]; // Output picture pointers (which may be different for interlaced sequences)
for (int i = 0; i < num_frames * 2; i++) {
opics[i] = new uint16_t*[3];
}
int ostride[3];
int num_frames_decoded_from_sequence = 0;
int total_frames_decoded = 0;
/* Initialise decoder */
vc2decode_init();
VC2DecoderHandle decoder = vc2decode_create();
/* Configure decoder */
{
VC2DecoderParamsUser params;
memset((void *)¶ms, 0, sizeof(params));
params.threads = threads;
params.colourise_quantiser = colourise_quantiser;
params.colourise_padding = colourise_padding;
params.colourise_unpadded = colourise_unpadded;
/* QuarterSize is only really sensible for HD */
if (quartersize) {
params.partial_decode = true;
params.partial_decode_width = 1920/2;
params.partial_decode_height = 1080/2;
params.partial_decode_offset_x = 0;
params.partial_decode_offset_y = 540;
}
{
VC2DecoderResult r = vc2decode_set_parameters(decoder, params);
if (r != VC2DECODER_OK) {
printf("Parameter Error: %d\n", r);
return 1;
}
}
}
/* Read input data */
{
FILE *f = FOPEN(input_filename.c_str(), "rb");
if (!f) {
fprintf(stderr, "Could not open: %s\n", input_filename.c_str());
perror("Invalid input file: ");
return 1;
}
int r = fseek(f, 0L, SEEK_END);
if (r < 0) {
perror("Error seeking in input file");
return 1;
}
input_length = ftell(f);
rewind(f);
idata = (char *)malloc(input_length);
size_t s = fread(idata, 1, input_length, f);
if (s != input_length) {
if (ferror(f))
fprintf(stderr, "Error reading file\n");
else
fprintf(stderr, "Improper Length on Input File\n");
return 1;
}
fclose(f);
printf("Read %lubytes of input data\n", input_length);
}
/* Set the output fmt to something invalid to start with */
VC2DecoderOutputFormat fmt;
fmt.width = 0;
fmt.height = 0;
fmt.signal_range = 0;
fmt.source_sampling = 0;
fmt.frame_rate_numer = 0;
fmt.frame_rate_denom = 0;
fmt.interlaced = 0;
ostride[0] = 0;
ostride[1] = 0;
ostride[2] = 0;
for (int i = 0; i < num_frames; i++) {
odata[i] = NULL;
}
/* Begin the main program loop */
int err = 0;
uint64_t time_taken = 0;
while (total_frames_decoded < num_frames) {
VC2DecoderResult r;
/* Reset to the start of the input data */
char *id = idata;
char *iend = idata + input_length;
/* We are at the start of a sequence, so synchronise */
r = vc2decode_synchronise(decoder, &id, iend - id, true);
if (r != VC2DECODER_OK_RECONFIGURED) {
fprintf(stderr, "Error synchronising to sequence");
err = 1;
break;
}
/* We have synchronised, so get the output format */
VC2DecoderOutputFormat new_fmt;
r = vc2decode_get_output_format(decoder, &new_fmt);
if (r != VC2DECODER_OK) {
fprintf(stderr, "Output format error: %d\n", r);
err = 1;
break;
}
/* Reconfigure output buffers if this format doesn't
match the current one */
if (fmt.width != new_fmt.width ||
fmt.height != new_fmt.height ||
fmt.interlaced != new_fmt.interlaced) {
fmt = new_fmt;
for (int i = 0; i < num_frames; i++) {
if (odata[i])
free(odata[i]);
odata[i] = (uint16_t *)malloc(fmt.width*fmt.height*2*sizeof(uint16_t));
}
if (fmt.interlaced) {
ostride[0] = fmt.width*2;
ostride[1] = fmt.width;
ostride[2] = fmt.width;
for (int i = 0; i < num_frames; i++) {
opics[2*i + 0][0] = odata[i];
opics[2*i + 0][1] = odata[i] + fmt.width*fmt.height;
opics[2*i + 0][2] = odata[i] + fmt.width*fmt.height + fmt.width*fmt.height/2;
opics[2*i + 1][0] = odata[i] + fmt.width;
opics[2*i + 1][1] = odata[i] + fmt.width*fmt.height + fmt.width/2;
opics[2*i + 1][2] = odata[i] + fmt.width*fmt.height + fmt.width*fmt.height/2 + fmt.width/2;
}
} else {
ostride[0] = fmt.width;
ostride[1] = fmt.width/2;
ostride[2] = fmt.width/2;
for (int i = 0; i < num_frames; i++) {
opics[i][0] = odata[i];
opics[i][1] = odata[i] + fmt.width*fmt.height;
opics[i][2] = odata[i] + fmt.width*fmt.height + fmt.width*fmt.height/2;
}
}
}
/* Decode Pictures from stream */
int picture = 0;
struct timespec start, end;
clock_gettime(CLOCK_MONOTONIC, &start);
while (total_frames_decoded < num_frames) {
r = vc2decode_decode_one_picture(decoder, &id, iend - id, opics[picture], ostride, true);
/* If End of Sequence break from loop */
if (r == VC2DECODER_OK_EOS) {
break;
}
/* If a picture has been sucessfully decoded continue */
if (r == VC2DECODER_OK_PICTURE) {
if (!fmt.interlaced || (picture%2))
total_frames_decoded++;
picture++;
if (id == NULL) {
fprintf(stderr, "Premature end of stream\n");
break;
}
continue;
}
/* If a reconfiguration has been triggered this is an invalid sequence, flag that */
if (r == VC2DECODER_OK_RECONFIGURED) {
fprintf(stderr, "Warning: this sequence changes parameters part way through\n");
break;
}
/* Otherwise we have an error */
if (r < 0) {
fprintf(stderr, "Error decoding:\n");
fprintf(stderr, " %s\n", VC2DecoderErrorString[-r]);
err=1;
break;
}
fprintf(stderr, "Unknown Return value: %d\n", r);
break;
}
clock_gettime(CLOCK_MONOTONIC, &end);
/* We have reached the end of a sequence, or have decoded enough frames */
/*If an error has occurred bail */
if (err)
break;
/*If no pictures were decoded bail */
if (picture == 0) {
fprintf(stderr, "No pictures in sequence!\n");
err = 1;
break;
}
/* Otherwise update the record of how many decoded frames have been recorded */
int num_frames_decoded_from_this_sequence = (fmt.interlaced)?(picture/2):(picture);
if (num_frames_decoded_from_this_sequence > num_frames_decoded_from_sequence)
num_frames_decoded_from_sequence = num_frames_decoded_from_this_sequence;
/* And update the record of time taken to decode */
time_taken += (end.tv_sec*1000000000 + end.tv_nsec) - (start.tv_sec*1000000000 + start.tv_nsec);
}
/* If there has been no error print the speed */
if (!err) {
printf("--------------------------------------------------\n");
printf(" %d frames decoded in %5.3fs\n", total_frames_decoded, time_taken/1000000000.0);
printf(" %5.3ffps\n", total_frames_decoded*1000000000.0/time_taken);
printf("--------------------------------------------------\n");
}
/* If there has been no error get decoder statistics */
if (!err) {
VC2DecoderSequenceInfo info;
vc2decode_sequence_info(decoder, &info);
print_sequence_info(info, verbose);
}
/*
If there is no error and output is enabled write the output to a file
This code actually performs an endianess swap and some bit shifting on the output
data to put it into the desired format (10-bit data in the high order bits of 16-bit
words in network byte order).
*/
if (!err && !disable_output) {
try {
FILE *of = FOPEN(output_filename.c_str(), "wb");
for (int i = 0; i < num_frames_decoded_from_sequence; i++) {
for (int y = 0; y < fmt.height; y++) {
size_t s = fwrite(&odata[i][(y*fmt.width)], 1, fmt.width*sizeof(uint16_t), of);
if (s != fmt.width*sizeof(uint16_t))
throw std::runtime_error("Writing Error");
}
for (int y = 0; y < fmt.height*2; y++) {
size_t s = fwrite(&odata[i][(fmt.height*fmt.width + y*fmt.width/2)], 1, fmt.width/2*sizeof(uint16_t), of);
if (s != fmt.width/2*sizeof(uint16_t))
throw std::runtime_error("Writing Error");
}
}
printf("Wrote out %d frames\n", num_frames_decoded_from_sequence);
fclose(of);
} catch(...) {
fprintf(stderr, "Error writing output\n");
err = 1;
}
}
/* Destroy the decoder */
vc2decode_destroy(decoder);
/* Deallocate buffers */
free(idata);
for (int i = 0; i < num_frames; i++)
if (odata[i])
free(odata[i]);
delete[] odata;
for (int i = 0; i < num_frames * 2; i++)
delete[] opics[i];
delete[] opics;
return err;
}
