| Commit message (Collapse) | Author | Age | Files | Lines | |
|---|---|---|---|---|---|
| * | Neon uint8x8_t corners and minor fixes | enricotenuti | 2024-10-15 | 2 | -51/+44 |
| | | |||||
| * | Removed un-common function from common.h | Sebastiano Tronto | 2024-10-10 | 1 | -1/+0 |
| | | |||||
| * | Remove double return | Sebastiano Tronto | 2024-10-07 | 1 | -1/+0 |
| | | |||||
| * | Revert table-based transformations | Sebastiano Tronto | 2024-10-04 | 4 | -115/+40 |
| | | | | | Long live the giant switch! | ||||
| * | Table-based transformations (will revert immediately) | Sebastiano Tronto | 2024-10-04 | 4 | -40/+115 |
| | | | | | | | | | This commit changes transformations to being based on a lookup table intead of a big switch. This is something I had planned for a long time and the code is nicer with this change. But unfortunately the performance is noticeably worse, so I'll revert all these changes in the next commit. I wanted to push this anyway so we can keep track of this. | ||||
| * | Improved gendata_h48k2 performance | Sebastiano Tronto | 2024-09-07 | 1 | -2/+2 |
| | | | | | | | | | | | Performance for the "real coordinate" case (h0) is ok, but still around 3x slower than the gendata_h48k4 method. Performance for the intermediate tables is not good, but for now I'll keep it like this. This also shows that for h11 we should use the gendata_h48k4 method for h11. For intermediate table, h8, h9 and h10 might be too slow to compute in an acceptable time. We could restrict to h7 (largest with base = 9). | ||||
| * | More renaming | Sebastiano Tronto | 2024-09-05 | 3 | -42/+42 |
| | | |||||
| * | Rename constants from _underscore to CAPS | Sebastiano Tronto | 2024-09-05 | 4 | -116/+116 |
| | | |||||
| * | Fix error with neon | Sebastiano Tronto | 2024-08-18 | 1 | -0/+3 |
| | | |||||
| * | Removed duplication | Sebastiano Tronto | 2024-08-18 | 4 | -70/+35 |
| | | |||||
| * | Reorganized folder structure | Sebastiano Tronto | 2024-08-18 | 5 | -0/+999 |
