ufo-core - Debian dgit repo for package ufo-core

	Commit message (Collapse)	Author	Age
*	Fix #126: abort execution without resources	Matthias Vogelgesang	2017-10-13
\| \| \| \| \|	This breaks the API because we had to add an error variable to the base_scheduler_run() call!
*	Unref buffers allocated by fixed scheduler	Matthias Vogelgesang	2016-05-13
\|
*	Use same reductor policy for both schedulers	Matthias Vogelgesang	2015-06-10
\|
*	Remove UfoArchGraph	Matthias Vogelgesang	2015-04-21
\| \| \| \|	Remote node handling moved to the UfoResources object.
*	Remove ufo_base_scheduler_get_gpu_nodes()	Matthias Vogelgesang	2015-03-06
\| \| \| \|	All the information is already available.
*	Add sink task mode for ... sinks	Matthias Vogelgesang	2015-03-04
\| \| \| \| \|	At the moment this is purely semantic and does not change any behaviour. In terms of scheduling a sink behaves like a processor that has no output data.
*	Fix #78: reset num-processed before execution	Matthias Vogelgesang	2015-02-23
\|
*	Fix #74: increase processed items in FixedScheduler	Matthias Vogelgesang	2015-02-17
\|
*	Buf fix for propagate metadata	Alexandre Lewkowicz	2014-11-25
\|
*	Fix #60: propagate meta data from inputs to output	Matthias Vogelgesang	2014-11-25
\|
*	Improve docstrings	Matthias Vogelgesang	2014-10-30
\|
*	Improve debug message	Matthias Vogelgesang	2014-10-27
\|
*	Allow Python threads	Matthias Vogelgesang	2014-10-23
\|
*	Add UfoBaseScheduler::set_gpu_nodes	Matthias Vogelgesang	2014-10-23
\| \| \| \|	This allows the user to specify which GPUs to use from a particular arch graph.
*	Retire UfoConfig infrastructure	Matthias Vogelgesang	2014-10-23
\|
*	Decouple config from scheduler	Matthias Vogelgesang	2014-10-23
\|
*	Fix #48: use PyGILState_Ensure/Release for GIL	Matthias Vogelgesang	2014-10-23
\| \| \| \| \|	For some reason Py_BEGIN/END_ALLOW_THREADS is not working but directly releasing the GIL is ...
*	Resize output buffer if size changed	Matthias Vogelgesang	2014-10-21
\|
*	Call ufo_task_get_requisition for each output	Matthias Vogelgesang	2014-10-21
\| \| \| \|	This is necessary because the size might change between iterations.
*	Re-use last input buffers until ever filter finished	Matthias Vogelgesang	2014-10-21
\|
*	Add fixed scheduler	Matthias Vogelgesang	2014-10-16
	Unlike the other schedulers this one does not try to expand the graph but leaves parallelisation decisions to the user. However, the new approach also allows arbitrary number of outputs which makes broadcasting data straightforward.