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
a4f6642518
fix issue #36 (missing return *this in Rotation2D
2009-08-11 15:11:47 +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
8e08680119
don't depend on uninitialized value
2009-08-09 17:18:42 +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
fe813911f2
make LU::solve() not to crash when rank=0
2009-08-09 00:06:53 +02:00
Benoit Jacob
5f8d58f36a
fix bug in sorting of singular values
2009-08-09 00:05:38 +02:00
Gael Guennebaud
f5e1c896c7
replace custom rank one update in LU by an expression
2009-08-08 00:01:43 +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
543a785756
Fix compilation in sparse module
2009-08-06 17:28:49 +02:00
Gael Guennebaud
2707a6b87c
fix determinant in PartialLU
2009-08-06 17:28:31 +02:00
Gael Guennebaud
1d1e4884da
oops, one more bug fix in homogeneous
2009-08-06 16:56:10 +02:00
Gael Guennebaud
9822493aaf
fixes in determinant and homogeneous
2009-08-06 16:54:55 +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
2e46e9f2b4
shame on me
2009-08-06 14:57:38 +02:00
Gael Guennebaud
d34c5ef509
fix my bad fix of Hauke's fix ;)
2009-08-06 14:54:25 +02:00
Gael Guennebaud
1d4fea48b5
fix a couple of compilations issues
2009-08-06 14:10:02 +02:00
Hauke Heibel
c2861dd41a
fixed inversion for AffineCompact matrices
2009-08-06 12:25:18 +02:00
Gael Guennebaud
56d00779db
more product refactoring
2009-08-06 12:20:02 +02:00
Hauke Heibel
6b2ab13ac5
fix vs.net compilation issue
2009-08-06 11:40:25 +02:00
Hauke Heibel
8163757cf0
fix vs.net compilation issue
2009-08-06 11:27:25 +02:00
Gael Guennebaud
fa55cf5ce7
fix compilation and segfault issues
2009-08-06 11:19:36 +02:00
Gael Guennebaud
b9b17ea5a5
add the missing Affine Transform * set of column vectors products...
2009-08-06 11:02:03 +02:00
Benoit Jacob
0744638b6f
remove remnant of MultiplierBase
2009-08-06 10:35:13 +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
Benoit Jacob
014c581a5b
fix assertions, improve docs.
...
we never assert on conditions that depend on the result of a computation!!
also the assertion that rank>0 amounts to matrix!=0 which we have to leave under the responsibility of the user.
2009-08-05 10:15:28 +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
Gael Guennebaud
f3a6bc48c4
fix a couple of compilation issue due to the removal of MultiplierBase
2009-08-04 13:16:40 +02:00
Gael Guennebaud
2089a263f8
merge
2009-08-04 11:31:25 +02:00
Gael Guennebaud
4bec101470
implement two levels of blocking in PartialLU => high speedup
2009-08-04 11:28:02 +02:00
Benoit Jacob
4436a4d68c
use explicit Block/VectorBlock xprs to make sure that compile-time known sizes are used
2009-08-04 00:27:58 +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
ce1dc1ab16
implements a blocked version of PartialLU
2009-08-03 12:11:18 +02:00
Gael Guennebaud
0103de8512
bugfix in trsm
2009-08-02 15:32:43 +02:00
