Status

Build is ok; passed all existing tests. Tested on Nvidia GTX 1070 on Linux.

Performance is good (up to 27x speedup) on large tiles, but poor on small tiles.

Algorithms

Tile size is too small: Average working area of a composite function is 64x64 pixels (16KB), which means that runtime overhead of memory transfer b/w a host and a device is significant.
Need more testing: Unit tests for OpenCL components, more tests for composite ops.

Optimize CompositeOpAlphaDarken32 for small tiles
- [ ] Pre-allocate pinned memory
- [ ] Allocate tiles in GPU memory and keep it consistent with CPU.
Run on Intel OpenCL for CPU and GPU.

Name		Name	Last commit message	Last commit date
Latest commit History 47,214 Commits
.github		.github
3rdparty		3rdparty
benchmarks		benchmarks
build-tools		build-tools
cmake		cmake
interfaces		interfaces
krita		krita
libs		libs
packaging		packaging
pics		pics
plugins		plugins
sdk		sdk
winquirks		winquirks
.arcconfig		.arcconfig
.gitattributes		.gitattributes
.gitignore		.gitignore
AUTHORS		AUTHORS
CMakeLists.txt		CMakeLists.txt
CODE_OF_CONDUCT.md		CODE_OF_CONDUCT.md
COPYING		COPYING
COPYING-CMAKE-SCRIPTS		COPYING-CMAKE-SCRIPTS
CTestConfig.cmake		CTestConfig.cmake
HACKING		HACKING
KoAlwaysInline.h		KoAlwaysInline.h
KoConfig.h.cmake		KoConfig.h.cmake
Mainpage.dox		Mainpage.dox
Messages.sh		Messages.sh
README.org		README.org
action_i18n.pl		action_i18n.pl
config-gsl.h.cmake		config-gsl.h.cmake
config-hash-table-implementaion.h.cmake		config-hash-table-implementaion.h.cmake
config-hide-safe-asserts.h.cmake		config-hide-safe-asserts.h.cmake
config-limit-long-tests.h.cmake		config-limit-long-tests.h.cmake
config-ocio.h.cmake		config-ocio.h.cmake
config-powf.h.cmake		config-powf.h.cmake
config-qtmultimedia.h.cmake		config-qtmultimedia.h.cmake
config-vc.h.cmake		config-vc.h.cmake
config_convolution.h.cmake		config_convolution.h.cmake
extracti18n.pl		extracti18n.pl
kundo2_aware_xgettext.sh		kundo2_aware_xgettext.sh