summaryrefslogtreecommitdiff
AgeCommit message (Collapse)Author
2010-02-16Turbo version using zero-copy where possibleMartin Lucina
Not going on master since this is all kind of complicated, but it's useful to have around as a stress test
2010-02-03Correct RGBA mask used in SDL_CreateRGBSurfaceFrom()Martin Lucina
This fixes the funky colours :-)
2010-02-03Fix multicast transportsMartin Lucina
2010-02-01Set ZMQ_RATE for PGM transports to 100MbpsMartin Lucina
2010-01-30Don't spew messages on frame conv failuresMartin Lucina
2010-01-30Add architecture description, make ESC quit the appMartin Lucina
2010-01-30Remove useless backslashes from string literalMartin Lucina
2010-01-30Code and comments cleanupMartin Lucina
2010-01-30Signal handling take 2, use zmq_poll in main threadMartin Lucina
Signal handling now uses an inproc transport to send a message to the main thread instead of acting on SIGINT and other termination signals and using a global variable to signal alarm expiration (yuck). This also changes the main/receiver thread architecture which now uses zmq_poll() and gives us the additional advantage of being able to poll for GUI events regularly. This solves the problem of the receive mode UI freezing when the sender has gone away.
2010-01-30Implement synchronous signal handlingMartin Lucina
This is to demonstrate how SIGINT, etc. can be handled properly even if the application thread is blocked on zmq_recv().
2010-01-30Add primitive frame rate limitingMartin Lucina
Doesn't work quite as well as I would like it to, but until 0MQ has queue limits implemented it's useful if you want to test a large video stream on e.g. 802.11g wifi. Also added proper option handling using getopt().
2010-01-30Try to fix printf warnings with uint64_tMartin Lucina
2010-01-30Implement bandwidth estimationMartin Lucina
Raw video streams require large amounts of bandwidth. For example a 640x480 RGB24 stream at 15 frames per second will consume around 100 Mbps. This code adds a bandwidth and fps display in the window title on the receiving side so that you can see the bandwidth requirements for the stream being sent.
2010-01-29Add snide comment about cheap camerasMartin Lucina
2010-01-29Change the inproc:// connection logicMartin Lucina
An inproc:// endpoint must exist before we can connect to it. This change removes the silly sleep(1) hack to wait for the sender thread to come online and instead does the zmq_bind() in the main thread and the zmq_connect() in the sender thread.
2010-01-29Clarified installation of deps on Debian/UbuntuMartin Lucina
2010-01-29Clarify dependenciesMartin Lucina
2010-01-29Initial commitMartin Lucina