|
1 | | -### v3.3.20170222 |
| 1 | +### v3.5.20170718 |
| 2 | +- Feature parity with ArrayFire 3.5. |
| 3 | + - `canny`: Canny Edge detector |
| 4 | + - `Array.scalar`: Return the first element of the array |
| 5 | + - `dot`: Now support option to return scalar |
| 6 | + - `print_mem_info`: Prints memory being used / locked by arrayfire memory manager. |
| 7 | + - `Array.allocated`: Returs the amount of memory allocated for the given buffer. |
| 8 | + - `set_fft_plan_cache_size`: Sets the size of the fft plan cache. |
| 9 | + |
| 10 | +- Bug Fixes: |
| 11 | + - `sort_by_key` had key and value flipped in documentation. |
| 12 | + |
| 13 | +- Improvements and bugfixes from upstream include: |
| 14 | + - CUDA backend uses nvrtc instead of nvvm |
| 15 | + - Performance improvements to arrayfire.reorder |
| 16 | + - Faster unified backend |
| 17 | + - You can find more information at arrayfire's [release notes](https://github.com/arrayfire/arrayfire/blob/v3.5.0/docs/pages/release_notes.md) |
| 18 | + |
| 19 | +### v3.4.20170222 |
2 | 20 | - Bugfix: Fixes typo in `approx1`. |
3 | 21 | - Bugfix: Fixes typo in `hamming_matcher` and `nearest_neighbour`. |
4 | 22 | - Bugfix: Added necessary copy and lock mechanisms in interop.py. |
5 | 23 | - Example / Benchmark: New conjugate gradient benchmark. |
6 | 24 | - Feature: Added support to create arrayfire arrays from numba. |
7 | 25 | - Behavior change: af.print() only prints full arrays for smaller sizes. |
8 | 26 |
|
9 | | -### v3.3.20161126 |
| 27 | +### v3.4.20161126 |
10 | 28 | - Fixing memory leak in array creation. |
11 | 29 | - Supporting 16 bit integer types in interop. |
12 | 30 |
|
|
0 commit comments