- May 09, 2022
-
-
Andreas Klöckner authored
-
- May 01, 2022
-
-
- Apr 19, 2022
-
-
Andreas Klöckner authored
-
-
Andreas Klöckner authored
-
- Apr 17, 2022
-
-
Matthias Diener authored
* Warn when a kernel has many arguments * count and check arg sizes * check for all context devices * refactor the size checking code and adjust pocl's limits * cleanup comments * Reword GPU warning Co-authored-by: Andreas Klöckner <inform@tiker.net> * Only warn when approaching the limit if param size is estimate Co-authored-by: Andreas Klöckner <inform@tiker.net>
-
- Apr 04, 2022
-
-
- Apr 03, 2022
-
-
- Feb 18, 2022
-
-
Matthias Diener authored
* remove some Python 2 code * also remove unused function
-
- Feb 17, 2022
-
-
- Feb 15, 2022
-
-
- Feb 14, 2022
-
-
- Feb 10, 2022
-
-
Andreas Klöckner authored
-
-
- Jan 17, 2022
-
-
Andreas Klöckner authored
-
-
- Jan 16, 2022
-
-
- Dec 30, 2021
-
-
Andreas Klöckner authored
-
Alexandru Fikl authored
-
- Dec 13, 2021
-
-
Matthias Diener authored
* Cache: reduce sleep time * Update cache.py * print warnings fix * restructure * add comment about our choice for the timeout
-
- Dec 12, 2021
-
-
- Dec 07, 2021
-
-
Andreas Klöckner authored
-
Andreas Klöckner authored
-
- Dec 05, 2021
-
-
nchristensen authored
* Make if_positive keep strides of then_ if known. The previous version of `if_positive` always returns a C-ordered array even if `then_` is not C-ordered. This breaks artificial viscosity in Mirgecom when an array context uses a Fortran style data layout. This change makes the output have the same strides as `then_` if the shape of `then_` is known, otherwise it defaults to use the same layout as `criterion`. * Increment version * Resolve Flake8 failures * Use more idiomatic != in shape comparison Co-authored-by: Andreas Klöckner <inform@tiker.net>
-
- Nov 24, 2021
-
-
- Oct 20, 2021
-
-
- Oct 18, 2021
-
-
Andreas Klöckner authored
-
-
- Oct 07, 2021
-
-
Andreas Klöckner authored
-
-