Benoit Jacob
309d540d4a
add parentheses; hopefully this solves Koldos MSVC compilation issue...
2009-08-12 10:14:15 -04:00
Benoit Jacob
22d65d47d0
finally, the good approach was two-sided Jacobi. Indeed, it allows
...
to guarantee the precision of the output, which is very valuable.
Here, we guarantee that the diagonal matrix returned by the SVD is
actually diagonal, to machine precision.
Performance isn't bad at all at 50% of the current householder SVD
performance for a 200x200 matrix (no vectorization) and we have
lots of room for improvement.
2009-08-12 02:35:07 -04:00
Benoit Jacob
ce033ebdfe
add EIGEN_DEBUG_VAR
2009-08-11 16:12:34 -04:00
Gael Guennebaud
afbd73b5cd
overload operartor* with a ProductBase such that "scalar * (mat * mat)" is optimized
...
as one could naturally expect
2009-08-11 15:15:06 +02:00
Gael Guennebaud
ea884e6f48
remove #include Bidiagonalization, and add missing ";"
2009-08-11 15:08:03 +02:00
Gael Guennebaud
35b4077a5d
merge
2009-08-09 23:11:25 +02:00
Gael Guennebaud
ef55e7f4ce
make custom asm directive volatile
2009-08-09 23:09:46 +02:00
Benoit Jacob
216ee335ac
LinearVectorization: If the destination isn't aligned,
...
we have to do runtime checks and we don't unroll, so it's only good for large enough sizes
2009-08-09 22:19:12 +02:00
Benoit Jacob
1f1705868b
now you can #define EIGEN_DEBUG_ASSIGN, and all the values in ei_assign_traits are printed
2009-08-09 21:35:13 +02:00
Benoit Jacob
527557672a
disable the assembly for fast unaligned stores. indeed, there is a strange bug that is triggered by this code:
...
#include<Eigen/Core>
int main()
{
Eigen::Matrix4f m;
m <<1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16;
m.col(0).swap(m.col(1));
std::cout << m << std::endl;
}
when the fast unaligned stores are used, the column is copied instead of being swapped.
2009-08-09 20:49:55 +02:00
Benoit Jacob
3ed83fa681
* add Jacobi transformations
...
* add Jacobi (Hestenes) SVD decomposition for square matrices
* add function for trivial Householder
2009-08-09 16:58:13 +02:00
Gael Guennebaud
d1dc088ef0
* implement a second level of micro blocking (faster for small sizes)
...
* workaround GCC bad implementation of _mm_set1_p*
2009-08-07 11:09:34 +02:00
Gael Guennebaud
3ac01b1400
compilation fix in EigenSolver,
...
bugfix in PartialLU
2009-08-06 16:41:54 +02:00
Gael Guennebaud
e82e30862a
typo
2009-08-06 15:04:42 +02:00
Gael Guennebaud
03febf00a0
fix VS compilation issue in MapBase::operator+= and -=
2009-08-06 15:03:37 +02:00
Gael Guennebaud
1d4fea48b5
fix a couple of compilations issues
2009-08-06 14:10:02 +02:00
Gael Guennebaud
56d00779db
more product refactoring
2009-08-06 12:20:02 +02:00
Hauke Heibel
8163757cf0
fix vs.net compilation issue
2009-08-06 11:27:25 +02:00
Gael Guennebaud
84a7659bef
implement the missing outer product,
...
and attempt to workaround a gcc internal error
2009-08-05 17:39:11 +02:00
Gael Guennebaud
88147e0a91
big refactoring in Product.h:
...
- all specialized products now inherits ProductBase
- the default product evaluated by Assign is still here,
but it is currently enabled for small fixed sizes only
- => this significantly speed up compilation for large matrices
- I left the OuterProduct specialization empty as an exercise...
2009-08-05 15:23:35 +02:00
Gael Guennebaud
7d607048a9
implement a ProductBase class and, as a proof of concept, update TriangularProduct
...
and SelfAdjointMatrixProduct to take advantage of it => fewer LOC
2009-08-04 16:54:17 +02:00
Benoit Jacob
523cdedf58
make the dot product linear in the second variable, not the first variable
2009-08-03 17:20:45 +02:00
Gael Guennebaud
912da9fade
merge with special_matrix branch
2009-08-03 16:17:32 +02:00
Gael Guennebaud
a8f943127c
merge
2009-08-03 16:11:30 +02:00
Benoit Jacob
d10c710b15
add new Householder module
2009-08-03 16:06:57 +02:00
Gael Guennebaud
3cf5bb31f6
* Bye bye MultiplierBase, extend a bit AnyMatrixBase to allow =, +=, and -=
...
* This probably makes ReturnByValue needless
2009-08-03 16:05:15 +02:00
Benoit Jacob
66ee2044ce
small fixes
2009-08-03 16:05:07 +02:00
Benoit Jacob
3cde9c0e35
apply Gael's idea for auto transpose in mixed fixed/dynamic case
2009-08-03 16:04:15 +02:00
Gael Guennebaud
0103de8512
bugfix in trsm
2009-08-02 15:32:43 +02:00
Gael Guennebaud
48fc64458c
add blocked LLT, and bugfix in trsm asserts
2009-08-01 23:42:51 +02:00
Gael Guennebaud
18429156a1
add selfadjointView from a trinagularView
2009-07-31 17:35:55 +02:00
Gael Guennebaud
a156f5a869
faster trsm kernel and fix a couple of issues
2009-07-31 13:18:19 +02:00
Gael Guennebaud
21f686846b
s/std::atan2/ei_atan2
2009-07-31 10:08:23 +02:00
Manuel Yguel
ae5e26a363
add missing ei_atan2 without painfull warnings
2009-07-31 09:21:31 +02:00
Gael Guennebaud
ff20a2ba94
add explicit "on the right" triangular solving,
...
=> no temporary when the rhs/unknows is row major
2009-07-30 16:03:06 +02:00
Gael Guennebaud
62d9b9b7b5
fix typo
2009-07-29 09:26:20 +02:00
Gael Guennebaud
864171df5c
fix a couple of issues related to recent products
2009-07-28 18:11:30 +02:00
Gael Guennebaud
1ba35248e9
synch with main branch
2009-07-28 17:37:22 +02:00
Gael Guennebaud
54804eb626
synch with main branch
2009-07-28 17:35:07 +02:00
Gael Guennebaud
264fe82c65
add a debug mechanism to compute the number of intermediate evaluations (only for dynamic size)
2009-07-28 17:13:13 +02:00
Gael Guennebaud
7ed7ec64b5
improve the expression analyzer to bypass Transpose expression
2009-07-28 14:02:12 +02:00
Gael Guennebaud
94cc30180e
compilation fixes
2009-07-27 13:50:23 +02:00
Gael Guennebaud
0590c18555
various compilation and bug fixes in selfadjoint stuff
2009-07-27 13:17:39 +02:00
Gael Guennebaud
b5e4064289
cleaning and fix a perf issue
2009-07-27 12:13:53 +02:00
Gael Guennebaud
f95b77be62
trmm is now fully working and available via TriangularView::operator*
2009-07-27 11:42:54 +02:00
Gael Guennebaud
6aba84719d
trmm is now working in all storage order configurations
2009-07-27 10:27:01 +02:00
Gael Guennebaud
1d4d9a37fd
some cleaning
2009-07-26 13:53:24 +02:00
Gael Guennebaud
f3fde74695
finalize trsm: works in all situations, and it is now used by solve() and solveInPlace()
2009-07-26 13:01:37 +02:00
Gael Guennebaud
282e18da49
ok, now trsm works very well for upper triangular matrices
...
TODO: link it with the meta triangular_solve_selector and handle
the case where the rhs is row major by copying it to a col-major
temporary + handle right solving: X = B * M^-1
2009-07-26 00:49:17 +02:00
Gael Guennebaud
f4112dcff3
The new trsm is working very very well (read very fast) for
...
lower triangular matrix and row or col major lhs.
TODO: handle upper triangular and row major rhs cases
2009-07-25 21:41:01 +02:00
