Age | Commit message (Collapse) | Author | |
---|---|---|---|
2023-05-02 | Avoid unused variable if compiled without HAVE_SCHED_SETAFFINITY | Thomas White | |
2023-05-02 | indexamajig: Add --cpu-pin | Thomas White | |
Similar code was stripped out 3.5 years ago by 4f4e5d05269eee6b192c271f2f9730a1ad3a4fbd. However, it seems there are still systems where it improves performance a lot. | |||
2022-11-08 | indexamajig: Exit only once all workers received kEndOfStream | Thomas White | |
Previously, the whole program would exit if any worker saw kEndOfStream. Unfortunately, this happens quite often, due to data starvation (too many workers for the data rate) or just general slowness. Therefore we need a more robust criterion. | |||
2022-10-05 | indexamajig: Improve accuracy of speed reports | Thomas White | |
2022-07-07 | indexamajig: Remove separate waitpid() calls when cleaning up last processes | Thomas White | |
2022-07-06 | indexamajig: Add timeout for ZMQ socket | Thomas White | |
2022-07-06 | indexamajig: Add profiling and last_task info for ZMQ fetch | Thomas White | |
2022-06-22 | indexamajig: Wrap ZMQ parameters into separate structure | Thomas White | |
2022-06-22 | indexamajig: Wrap ASAP::O parameters up inside separate structure | Thomas White | |
2022-06-17 | ASAP::O: Simplify by removing stream switching logic | Thomas White | |
2022-06-16 | ASAP::O: Exit quicker (and with explanation) if setup fails | Thomas White | |
2022-06-02 | ASAP::O: Add last_task for ASAPO fetch | Thomas White | |
2022-06-02 | Add worker ID to profiling data | Thomas White | |
2022-06-02 | indexamajig: Add --asapo-stream | Thomas White | |
2022-06-02 | indexamajig: Track metadata (e.g. for ASAP::O) | Thomas White | |
2022-06-02 | ASAP::O: Add profiling bits | Thomas White | |
2022-06-02 | Generate placeholder filename/event earlier | Thomas White | |
Previously, a placeholder was put in the queue ("ZMQdata //" or "ASAPOdata //"), and replaced by image_read_data_block. Instead, the "final" placeholder can be put in the queue already. Note that ASAP::O, at least, will replace this placeholder with a filename delivered by the data transport. | |||
2022-06-02 | ASAP::O: Pass filename through | Thomas White | |
2022-06-02 | ASAP::O: Remove filesystem path | Thomas White | |
This is only really needed for offline processing. It's an added complication for now. Maybe we'll revisit it later and re-expose this, though. | |||
2022-06-02 | ASAP::O: Expose data source | Thomas White | |
2022-06-02 | ASAP::O: Simplify im_asapo_make_unique_group_id | Thomas White | |
2022-06-02 | ASAP::O: Don't send data to process_image if nothing came | Thomas White | |
2022-06-02 | ASAP::O: Show the uniquely-generated group ID | Thomas White | |
2022-06-02 | indexamajig: Don't do ASAP::O stuff unless asked for | Thomas White | |
2022-06-02 | indexamajig: Exit if unique ASAP::O group ID can't be created | Thomas White | |
2022-06-02 | indexamajig: Special cases for ASAP::O as for ZMQ | Thomas White | |
2022-06-02 | ASAP::O: Remove vestigial block | Thomas White | |
2022-06-02 | ASAP::O: Fix memory leak on error path | Thomas White | |
2022-06-02 | ASAP::O: Tidy up build process | Thomas White | |
This removes a lot of random #ifdefs from the source code. | |||
2022-06-02 | Fix build without ASAP::O | Thomas White | |
2022-06-02 | ASAP::O: Actually get the data | Thomas White | |
2022-06-02 | ASAP::O: Group ID stuff | Thomas White | |
2022-06-02 | ASAP::O guts | Thomas White | |
2022-05-04 | Write profile information in one call to write() | Thomas White | |
Unfortunately, the complicated series of nested printf()s doesn't work well from multiple processes. | |||
2022-05-04 | Add improved profiling | Thomas White | |
Improvements: - Recursive, so we can see how things are working at different levels - Everything is profiled, even if not explicitly labelled - No possibility of losing time - Detects mistakes if routines aren't wrapped correctly | |||
2022-05-03 | Strip out time-accounts | Thomas White | |
2022-03-02 | indexamajig: Return a specific error code (5) if no files could be processed | Thomas White | |
2021-12-17 | indexamajig: Avoid fclose(NULL) | Thomas White | |
When running online, there's no input file handle to close. | |||
2021-12-17 | indexamajig: Send sandbox ping while waiting for next frame | Thomas White | |
This avoids the process getting killed while waiting e.g. between runs in an online context. | |||
2021-12-07 | indexamajig: Generate per-frame profiling numbers | Thomas White | |
2021-12-07 | indexamajig: Shut down cleanly on SIGUSR1 | Thomas White | |
2021-09-07 | indexamajig: Avoid crash on NULL event | Thomas White | |
2021-07-23 | indexamajig: Fix error path when worker pipe is closed mid-chunk | Thomas White | |
2021-07-23 | try_status: Remove duplicate assignment | Thomas White | |
2021-07-21 | Add missing cleanup on error paths | Thomas White | |
2021-05-05 | indexamajig: Add --zmq-request | Thomas White | |
This (re-)adds the ability to get data via a request/reply socket. See afcb7b568947c for when it was removed. | |||
2021-04-23 | indexamajig: Generate a unique filename for ZMQ data | Thomas White | |
The unique filename is needed by the GUI for looking up results in a stream. Otherwise, the filename is "(null)" for everything and the lookup just returns the first chunk in the stream. The filename is generated based on the unique serial number for each chunk, and is therefore unique across one run of indexamajig regardless of the number of worker processes (-j). This might have to change in future to accommodate jobs run across multiple nodes, if there is any demand for looking at results in one big concatenated stream. This also changes the condition for deciding when to look for a 'real' file, to take into account that there is always a non-NULL filename. | |||
2021-04-23 | indexamajig: Mercilessly ignore short ZMQ messages | Thomas White | |
This is a hack to filter out what seems like "keepalive" messages. It's nasty, but better than contaminating the stream with spurious chunks. | |||
2021-04-23 | indexamajig: Add --zmq-subscribe | Thomas White | |
2021-04-20 | Fix memory handling semantics for ZMQ data | Thomas White | |