Releases: eyalroz/cuda-kat
Release 0.2
Features added:
- grid-info: More functions and aliases; some function renames.
- Non-builtins:
rotate_left()
,rotate_right()
,log2()
. - #88: Simple integer ranges, and special ranges for warp, block and grid-stride iteration, instead of calling at_grid_stride, at_block_stride etc.
Features removed:
strf
-based stringstreams: They're currently broken. They may make a comeback in a future release.
Other noteworhty changes:
Bugfix release 0.1.1
This point release fixes bug #72 , which prevented for use of the reduction, scan and scan-and-reduce primitives, at the warp and block level, with floating-point types. Note that these templated primitives now have one less template parameter (albeit one with a default assigned value).
Initial release
This is the first proper release of the Kernel Author's Toolkit. The repository has existed for while now - almost a full year - so there has been quite a bit of "percolation" for the code to "settle" in a releasable state.
What marks the time for this release, then, is the completion of extensive work on unit test coverage (details here), which has significantly improved the robustness of the code - to the point it is ready for wider public use and scrutiny.
Users of this initial release are especially encouraged to provide feedback - positive and negative - on the choice of functionality, on style, on performance, compatibility with other libraries and so forth.