Improvement on grevlex gb computations over QQ

Open mohabsafey opened this issue 8 months ago • 2 comments

uses hash tables from Christian
reorganisation of CRT and rational reconstruction calls
as a consequence: less memory consumption
as a consequence: several bug fixes
as a consequence: faster on several examples

Apr 09 '25 07:04 mohabsafey

Wait a little bit before merging it ; I would like to do some more intensive tests and add specific tests for this PR

Apr 09 '25 07:04 mohabsafey

These new commits fix some bugs arising on large computations (mainly because of "bad" primes issues). Some more intensive tests are running ; thus it is better to wait a little bit before merging this PR.

Apr 27 '25 19:04 mohabsafey

I think this PR is ready to be merged. Note that the added tests take 2 secs which is longer than what we are used to, but I could not find a test which is smaller. Performance improvements (both in terms of memory usage and CPU time) are confirmed.

May 07 '25 14:05 mohabsafey