Commit Graph

13 Commits

Author SHA1 Message Date
Martin Diehl 10d8a63cb6 not used
and current code is not good (not using Horner scheme)
2022-05-08 17:18:15 +02:00
Martin Diehl d713026f7e fast evaluation for (most common) case of constant 2022-05-08 16:49:30 +02:00
Martin Diehl b2052cb3c7 less strict tolerances
values can be in the order of 1e5, so 1e-6 precision is not too bad
2022-05-08 16:41:19 +02:00
Martin Diehl 8f9fbb30e5 use fused multiply-add where possible
only possible for Intel compiler
2022-05-08 15:48:13 +02:00
Martin Diehl 240426402c using (faster) Horner evaluation
https://rosettacode.org/wiki/Horner%27s_rule_for_polynomial_evaluation
2022-05-08 13:32:14 +02:00
Martin Diehl 72c29f744c better tests 2022-05-08 12:49:29 +02:00
Martin Diehl ec184cb8fe need to relax absolute tolerance
absolute values for quartic become quite high
2022-05-08 08:30:58 +02:00
Martin Diehl 9605a5c2de polynomial expansion up to order 4 2022-05-07 23:04:06 +02:00
Martin Diehl 71e4fa222c generic code for variable order 2022-05-07 22:53:31 +02:00
Martin Diehl 70b9943920 allow to specify only quadratic term in Dict 2022-05-07 22:47:13 +02:00
Martin Diehl 7cf3bcb9c9 easier to read 2022-05-07 22:01:47 +02:00
Martin Diehl 8fead8e306 relaxed tolerances
for optimized code, the last digit might differ
2022-02-02 13:48:04 +01:00
Martin Diehl 39aa243695 Polynomial Class 2022-01-31 14:05:15 +00:00