csa/tomo/approximate/ufo-filters.git - Unnamed repository; edit this file 'description' to name the repository.

	Commit message (Collapse)	Author	Age	Files	Lines
*	Don't force non-dash task names	Tomas Farago	2020-08-14	2	-5/+2
\|
*	Fix undefined CL_TARGET_OPENCL_VERSION warning	Tomas Farago	2020-06-12	9	-0/+20
\| \| \| \| \| \|	This is hardcoded to 120 (= OpenCL 1.2) because nvidia does not support later standards until 3.0 comes around and we do not make use of later features anyway.
*	med-mad-reject-2d: fix index overflow in sorting	Tomas Farago	2020-03-19	1	-2/+2
\|
*	Fix #166: propagate OpenCL errors if possible	Matthias Vogelgesang	2018-06-20	3	-15/+15
\|
*	Fix #165: use current get_kernel API	Matthias Vogelgesang	2018-06-20	1	-1/+1
\|
*	Use updated _get_requisition API	Matthias Vogelgesang	2018-06-20	4	-6/+8
\|
*	Use new _get_kernel API	Matthias Vogelgesang	2018-06-19	3	-9/+6
\|
*	stat-monitor: clean up and removal of dead code	Matthias Vogelgesang	2017-10-18	1	-477/+389
\|
*	contrib: compile with common code	Matthias Vogelgesang	2017-10-18	1	-1/+1
\|
*	Move contrib filters and allow builds with meson	Matthias Vogelgesang	2017-10-18	19	-801/+108
\|
*	Correct computation erros in stat-monitor filter	Serge Cohen	2017-01-25	2	-34/+84
\| \| \| \| \| \| \| \| \| \| \|	The previous version of stat-monitor had a few important bugs that are now solved : * Was producing boggus statistics when the number of pixels was lower than the inferred number of work items to use. * The mean and var computed were based on a very limited number of pixels. Indeed the computation was done on all pixels but a final reduction step was mistakenly using a small number of those * The min/max were wrong as soon as the number of pixel was too low.
*	Fix med-mad-reject corners	Serge Cohen	2017-01-23	1	-43/+145
\| \| \| \|	In the previous kernel version, corner of each frame would end up having a value of 0.0f. Now the value is more likely to be unchanged and if changed the new value is one coming from the surrounding pixels.
*	Include documentation for contributed work	Matthias Vogelgesang	2017-01-17	1	-145/+0
\|
*	First version of contributed filters documentation	Serge Cohen	2017-01-17	1	-0/+145
\|
*	Trace processed frame number in stat-monitor	Serge Cohen	2017-01-17	1	-2/+31
\| \| \| \|	... as well as indexing stat-monitor tasks.
*	Use macros pixel access macros in ocl_1liner.cl	Serge Cohen	2017-01-17	2	-5/+20
\| \| \| \| \|	Implements macros in ocl_1liner.cl as more direct shortcuts to current (work-item) pixel of output and input(s).
*	Remove default output to stdout of ocl1-liner and stat-monitor	Serge Cohen	2017-01-17	2	-3/+47
\|
*	Fixups and cleanup	Serge Cohen	2017-01-17	5	-22/+31
\| \| \| \| \| \| \| \| \| \| \| \|	* Making sure /contrib/ is used in singular form. * Small correction to avoid some warning during compilation (unused variables, asprintf on Linux, cl_long printf). * Correcting the type of box-size parameter of med-mad-reject-2d. * Correcting bugs in ocl-1liner (generation of 'in_X' input parameters, comma after 'out' parameter)'. * Fixing up the ‘quiet’ property of stat-monitor. * Fixing memory double-release in ocl-1liner. * Fixing semantic of the kernel source code buffer free.
*	Inherit CMake settings from the root project	Matthias Vogelgesang	2017-01-10	3	-9/+0
\| \| \| \|	Otherwise we have to set the policy options each and every time.
*	Include limited contrib/ directory	Serge Cohen	2017-01-10	18	-0/+3311
	Cmake integration included a first contributed filter(s), building contrib requires the CMake option : -DWITH_CONTRIBS:BOOL=ON (this one is OFF by default). Added 4 new contributed filters : * Median-MAD rejection in 3D 3x3x3 box size * Median-MAD rejection in 2D but variable box size * OpenCL 1-liner and multi-input filter * Monitoring statistics of an image stream (min, max, mean, var)