Commit message (Collapse) | Author | Age | |
---|---|---|---|
* | aberth: Compile with latest futhark | Tom Smeding | 2020-06-30 |
| | |||
* | It's Jacobi-style, not Lagrange-style | tomsmeding | 2019-04-20 |
| | |||
* | Use 'init' in max_root_norm in futhark | tomsmeding | 2019-04-20 |
| | |||
* | Chunked computations | tomsmeding | 2019-04-20 |
| | |||
* | Now it actually works on iris! | tomsmeding | 2019-04-20 |
| | |||
* | Switch to f32 for intel iris (which doesn't work) | tomsmeding | 2019-04-20 |
| | |||
* | Different image bounds layout | tomsmeding | 2019-04-20 |
| | |||
* | Larger! | tomsmeding | 2019-04-20 |
| | |||
* | Correct root bounds checking in futhark | tomsmeding | 2019-04-19 |
| | |||
* | Working Futhark! 🎉 | tomsmeding | 2019-04-19 |
| | | | | | | | | That was actually easier than expected, and (surprise!) it runs ORDERS OF MAGNITUDE faster than my C++ implementation... C++ takes ~1min on my i5-4278U CPU @ 2.60GHz using all 4 virtual cores; the Futhark takes ~11sec on 1 core. Not sure whether my own code is so terribly bad or whether Futhark is cool, but this is cool nevertheless. | ||
* | Rename lib.fut to more reasonable name | tomsmeding | 2019-04-19 |