ad-dual/bench, branch master
[no description]
https://git.tomsmeding.com/ad-dual/atom?h=master
2025-02-24T21:11:24Z
Dual arrays is >100x faster than 'ad' on large fneural
2025-02-24T21:11:24Z
Tom Smeding
tom@tomsmeding.com
2025-02-24T21:10:47Z
urn:sha1:3631b758acfb2585809fdb0755e1a8e7afe3b9b7
ad:Numeric.AD.Double / ad-dual:Numeric.ADDual.Array.Internal
Prelude> 1.129e-3 / 41.89e-6 -- neural-100
26.951539746956314
Prelude> 34.67e-3 / 156.9e-6 -- neural-180
220.9687699171447
Prelude> 79.03e-3 / 178.6e-6 -- neural-500
442.4972004479283
Prelude> 365.3e-3 / 665.5e-6 -- neural-2000
548.9105935386928
Optimise by backpropagating scalar computation in C
2025-02-23T20:44:23Z
Tom Smeding
t.j.smeding@uu.nl
2025-02-23T20:44:23Z
urn:sha1:5f7a81acc7f75415d62dac86c5b50c848ab15341
Working tests and benchmarks against 'ad'
2025-02-21T12:35:26Z
Tom Smeding
t.j.smeding@uu.nl
2025-02-21T12:35:26Z
urn:sha1:a17bd53598ee5266fc3a1c45f8f4bb4798dc495e
Try debugging crash
2025-02-20T21:41:53Z
Tom Smeding
t.j.smeding@uu.nl
2025-02-20T21:41:53Z
urn:sha1:20754e0ae9226f658590f46105399aee5c0dcee2
Attempt at a benchmark (crashes)
2025-02-20T09:11:57Z
Tom Smeding
tom@tomsmeding.com
2025-02-20T09:11:57Z
urn:sha1:fe3132304b6c25e5bebc9fb327e3ea5d6018be7a