- 16 Dec, 2010 5 commits
-
-
Kenneth Moreland authored
-
Kenneth Moreland authored
Apparently this flag is not ubiquitous amongst gcc compilers.
-
Kenneth Moreland authored
-
Kenneth Moreland authored
I forgot to add the const to the arguments to some of them.
-
Kenneth Moreland authored
This happened when setting run lengths.
-
- 15 Dec, 2010 7 commits
-
-
Kenneth Moreland authored
-
Kenneth Moreland authored
Many parts of the code allowed implicit conversion from a 64 bit value to a 32 bit value. A lot of this was using IceTSizeType as a 32 bit integer. Using the compiler warnings, remove the conversions or make them explicit.
-
Kenneth Moreland authored
-
Kenneth Moreland authored
-
Kenneth Moreland authored
-
Kenneth Moreland authored
To simplify the indexing of image partitions in binary swap, the algorithm pretends that the image is the next largest factor of the number of processes. This ensures that the image will always be divided evenly and consistently. As a consiquence, the partition at the end of the image might get truncated. This was handled, but was not handled was the possibility that the last partition was of size 0 and the next largest is truncated. This can happen if the final partition sizes contain less pixels than number of processes. The problem was most easily demonstrated by running the OddImageSizes test on 8 processes. (4 processes would probably work too, but I did not test that.) This fix should also handle any situation where the size of the image has fewer pixels than number of processes (which is also the case in the OddImageSizes test).
-
Kenneth Moreland authored
It was always supposed to be there. I must have forgotten it.
-
- 14 Dec, 2010 2 commits
-
-
Kenneth Moreland authored
-
-
- 30 Nov, 2010 2 commits
-
-
Kenneth Moreland authored
-
Kenneth Moreland authored
When in icetCompressedComposite or icetCompressedSubComposite, the time to do the comparison/blending was not actually recorded. Fixed that.
-
- 18 Nov, 2010 5 commits
-
-
Kenneth Moreland authored
-
Kenneth Moreland authored
-
Kenneth Moreland authored
-
Kenneth Moreland authored
-
Kenneth Moreland authored
Necessary for compiling on MSVC. Also cleaned up some obvious problems with declaring variables in the middle of a block.
-
- 17 Nov, 2010 6 commits
-
-
-
Kenneth Moreland authored
-
Kenneth Moreland authored
-
Kenneth Moreland authored
-
Kenneth Moreland authored
To get around the issues, the pointers are now cast to IceTByte. It is not an elegent solution, but it solves problems on MSVC compilers.
-
Kenneth Moreland authored
This catches arithmetic with void * pointer, which is not supported by MSVC.
-
- 15 Nov, 2010 1 commit
-
-
Kenneth Moreland authored
-
- 12 Nov, 2010 1 commit
-
-
Kenneth Moreland authored
-
- 04 Nov, 2010 1 commit
-
-
Kenneth Moreland authored
Also deleted the temporary InterfaceChanges.txt file.
-
- 19 Oct, 2010 1 commit
-
-
Kenneth Moreland authored
-
- 29 Sep, 2010 1 commit
-
-
Kenneth Moreland authored
I discovered that several strategies could return an image of the wrong size when the display tile is blank. The image buffer would be set to the maximum display size and never resized to the local tile size.
-
- 22 Sep, 2010 1 commit
-
-
Kenneth Moreland authored
Also untabified reduce.c.
-
- 20 Sep, 2010 1 commit
-
-
Kenneth Moreland authored
-
- 17 Sep, 2010 1 commit
-
-
Kenneth Moreland authored
-
- 24 Aug, 2010 1 commit
-
-
Kenneth Moreland authored
-
- 09 Aug, 2010 3 commits
-
-
Kenneth Moreland authored
The copyright on all source now has the current preferred legalize. All source now also states that the license is New BSD.
-
Kenneth Moreland authored
-
Kenneth Moreland authored
The install target was installing several files that should not have been installed including one that no longer exists.
-
- 29 Jul, 2010 1 commit
-
-
Kenneth Moreland authored
When giving the error about the OpenGL layer not being initialized, hint to the user that icetGLInitialize must be called.
-