-
Notifications
You must be signed in to change notification settings - Fork 0
FififiJamie/cuj2k-32bpc-bugFix
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Folders and files
Name | Name | Last commit message | Last commit date | |
---|---|---|---|---|
Repository files navigation
This is a bug fix version of cuj2k. Original code (taken from following repo) has a bug when encoding 32bits per channel images. Having to say that this bug fix is not nessisary since jpeg2000 allows arbitrary numbers of channels hence you could divide your channels into multiple subchannels to bypass this bug. If you only need to work with original form (not as a shared lib), just apply the changes in tier2.c. The following text are from the original readme: -------------------------------------------------------------------- CUJ2K 1.1 - JPEG2000 Encoder on CUDA http://cuj2k.sourceforge.net/ Copyright (c) 2009 Norbert Fuerst, Martin Heide, Armin Weiss, Simon Papandreou, Ana Balevic Compilation =========== Windows: ------------------------------------------------------------ - use src_encoder\gpu_vc2005.sln for Visual Studio 2005 - use src_encoder\gpu_vc08.sln for Visual Studio 2008 Remember to choose the correct configuration (Release for GPU code, EmuRelease for CPU emulation code)! - use src_gui\dotnet_gui.sln to compile the .Net GUI (currently only for Visual Studio 2008) Linux: -------------------------------------------------------------- $ cd src_encoder $ make optional: (execute as root) $ make install (copies the executable to /usr/local/bin/) For cleaning up: $ make clean You may use the options emu=1 and dbg=1 to enable debug and/or emulation; e.g.: $ make emu=1 dbg=1 $ make install emu=1 dbg=1 (you must use the same options in both commands, also for make clean) The executable and temporary files are stored in subdirectories depending on the supplied options. **** **** ***** **** **** **** ***** **** **** **** ***** **** Linux troubleshooting: 1 ) When we tested it, compilation only worked with gcc 4.3 and not with gcc 4.4; but this might be fixed by NVIDIA in the meantime, so check for the latest Toolkit and SDK versions. If you can't compile, install gcc and g++ 4.3. Two options: a) uninstall gcc and g++ 4.4 before installing version 4.3 b) after installing version 4.3, ensure that the commands gcc and g++ correspond to version 4.3. Should work about like this (might be different on your system, no guarantee!) (become root) $ cd /usr/bin $ rm gcc g++ $ ln -s gcc-4.3 gcc $ ln -s g++-4.3 g++ $ chmod a+x gcc g++ 2 ) If you get runtime errors like "... cudaGetDeviceProperties not implemented", you probably have an older CUDA version. You need to recompile cuj2k: Uncomment the line "COMMONFLAGS += -DNO_DEVICE_PROP" in the makefile, then run $ make clean $ make 3 ) If the executable from the binary package does not work, and you get other error messages not mentioned here, just try compiling it by yourself, because then probably some library versions do not match. **** **** ***** **** **** **** ***** **** **** **** ***** ****
About
API updates and Supports 32bpc
Topics
Resources
Stars
Watchers
Forks
Releases
No releases published
Packages 0
No packages published