Size: 1577
Comment:
|
Size: 2567
Comment:
|
Deletions are marked like this. | Additions are marked like this. |
Line 3: | Line 3: |
* Michael Abshoff * Martin Albrecht * Nick Alexander * Gregory Bard |
* Michael Abshoff -> LinBox debianization * Martin Albrecht -> GF(2), M4RI * Nick Alexander -> cylcotomic linalg * Gregory Bard -> M4RI |
Line 9: | Line 9: |
* Craig Citro * Burcin Erocal * Robert Miller (especially sparse GF(2)) * David Roe (p-adic linear algebra?) * Arne Storjohann * William Stein * Ralf-Philipp Weinmann |
* Craig Citro -> cyclotomic linalg, p-adic linalg * Burcin Erocal -> nullspace of polynomial matrices * Robert Miller (especially sparse GF(2)) -> sparse GF(2) * David Roe (p-adic linear algebra?) -> cyclotomic linear algebra * Arne Storjohann -> HNF, Sparse GF(2) Linalg * William Stein -> Cyclotomic linear algebra, HNF (LLL) * Ralf-Philipp Weinmann -> Sparse Elimination, Nullspace |
Line 36: | Line 36: |
[Arne, Ralph, Clément, Rob Miller] | |
Line 43: | Line 44: |
Line 44: | Line 46: |
* new algorithm (Storjohann) | [Arne, Clément, William] * new algorithm, based on system solving (Arne Storjohann) -> already an implementation * integrate it in IML * benchmark it against Sage * generalization: block vector system solving * implementation * benchmark * LLL trick for the existing implementation in Sage (William & Clément) == Polynomial Matrix computations == [Burcin] * Nullspace computation (Burcin) == Cyclotomic linear algebra == [Rob Bradshaw, Craig, William] == P-adic linear algebra == [David Roe, Craig, David Cohet] == LinBox debianization and 1.1.6 spkgization == [mabshoff, Clément] * get rid of gmp++ * generate and test linbox-1.1.6rc0.spkg |
Dev Days 1: Exact Linear Algebra
Michael Abshoff -> LinBox debianization
Martin Albrecht -> GF(2), M4RI
Nick Alexander -> cylcotomic linalg
Gregory Bard -> M4RI
- Rob Beezer
- Tom Boothby
Craig Citro -> cyclotomic linalg, p-adic linalg
Burcin Erocal -> nullspace of polynomial matrices
Robert Miller (especially sparse GF(2)) -> sparse GF(2)
David Roe (p-adic linear algebra?) -> cyclotomic linear algebra
Arne Storjohann -> HNF, Sparse GF(2) Linalg
William Stein -> Cyclotomic linear algebra, HNF (LLL)
Ralf-Philipp Weinmann -> Sparse Elimination, Nullspace
Dense GF(2)
- implement LQUP decomposition [Clement, Martin]
- implement LQUP routine [Clement]
- implement TRSM routine [Clement]
- implement efficient column swaps/rotations [Martin]
- SSE2 might help a lot here
- implement memory efficient mzd_addmul_strassen [Martin]
- See Clement's et al. paper on memory efficient Strassen-Winograd
- implement Arne's asymptotically fast elimination algorithm [Martin]
- implement multi-core multiplication with optimal speed-up
- OpenMP seems to be nice and easy
- 2 cores probably main target, but think about 4 cores too
- improve efficiency of M4RM
- try 7 instead of 8 Gray code tables to leave room for the actual matrix in L1
- try to fit three matrices rather than two into L2 or understand why it works so good for two
- detect L1/L2 cache sizes at runtime and choose optimal parameters for them
- implement Bill's half table idea and benchmark it
Sparse GF(2) (and other small finite fields)
- [Arne, Ralph, Clément, Rob Miller]
- Sparse Reduced Echelon form (RPW)
- Sparse Elimination:
improve LinBox gauss-domain
- eclib sparse elimination
- ....
- Sparse Elimination:
- ....
Hermite Normal Form
- [Arne, Clément, William]
new algorithm, based on system solving (Arne Storjohann) -> already an implementation
- integrate it in IML
- benchmark it against Sage
- generalization: block vector system solving
- implementation
- benchmark
LLL trick for the existing implementation in Sage (William & Clément)
Polynomial Matrix computations
- [Burcin]
- Nullspace computation (Burcin)
Cyclotomic linear algebra
- [Rob Bradshaw, Craig, William]
== P-adic linear algebra ==
- [David Roe, Craig, David Cohet]
LinBox debianization and 1.1.6 spkgization
- [mabshoff, Clément]
- get rid of gmp++
- generate and test linbox-1.1.6rc0.spkg