Commit Graph

347 Commits

Author SHA1 Message Date
Gael Guennebaud
804a239d30 patch from Moritz Lenz to allow solving transposed problem with superlu 2009-04-10 19:54:43 +00:00
Gael Guennebaud
22edf77470 add a 4x4 inverse case which is not handled by the current
ei_compute_inverse_in_size4_case (reported by mikola on IRC)
2009-04-09 21:55:29 +00:00
Gael Guennebaud
5b1d0cebc5 sparse module: new API proposal for triangular solves and experimental
solver support with a sparse matrix as the rhs.
2009-04-05 16:30:10 +00:00
Benoit Jacob
2f45eeb0c6 More Cholesky fixes.
* Cholesky decs are NOT rank revealing so remove all the rank/isPositiveDefinite etc stuff.
* fix bug in LLT: s/return/continue/
* introduce machine_epsilon constants, they are actually needed for Higman's formula determining
  the cutoff in Cholesky. Btw fix the page reference to his book (chat with Keir).
* solve methods always return true, since this isn't a rank revealing dec. Actually... they already did always return true!! Now it's explicit.
* updated dox and unit-test
2009-04-01 00:21:16 +00:00
Benoit Jacob
bf596d0b3a add adjointInPlace() and add documentation warnings on adjoint() and transpose() about aliasing effects. 2009-03-31 13:55:40 +00:00
Benoit Jacob
a1ba995f05 Many improvements in LLT and LDLT:
* in LDLT, support the negative semidefinite case
* fix bad floating-point comparisons, improves greatly the accuracy of methods like
  isPositiveDefinite() and rank()
* simplifications
* identify (but not resolve) bug: claim that only triangular part is used, is inaccurate
* expanded unit-tests
2009-03-30 21:45:45 +00:00
Gael Guennebaud
49fc1e3e84 add vectorization of sqrt for float 2009-03-27 14:41:46 +00:00
Gael Guennebaud
17860e578c add SSE2 versions of sin, cos, log, exp using code from Julien
Pommier. They are for float only, and they return exactly the same
result as the standard versions in about 90% of the cases. Otherwise the max error
is below 1e-7. However, for very large values (>1e3) the accuracy of sin and cos
slighlty decrease. They are about 3 or 4 times faster than 4 calls to their respective
standard versions. So, is it ok to enable them by default in their respective functors ?
2009-03-25 12:26:13 +00:00
Gael Guennebaud
f4cf5e9b26 split and extend eigen-solver tests 2009-03-23 14:38:59 +00:00
Gael Guennebaud
fbf415c547 add vectorization of unary operator-() (the AltiVec version is probably
broken)
2009-03-20 10:03:24 +00:00
Gael Guennebaud
b8f46090ff add optimized cross3 function (code from Rohit Garg) 2009-03-11 14:20:36 +00:00
Gael Guennebaud
3f80c68be5 add the vectorization of abs 2009-03-09 18:40:09 +00:00
Gael Guennebaud
3ac42fed94 big rework of the Transform class:
* add Projective and AffineCompact modes as an optional third template
  argument
* extend Transform::operator* to support more use cases
2009-03-08 11:35:30 +00:00
Gael Guennebaud
fa9f7708d4 add efficient matrix product specializations for Homogeneous 2009-03-05 16:40:56 +00:00
Gael Guennebaud
0be89a4796 big addons:
* add Homogeneous expression for vector and set of vectors (aka matrix)
  => the next step will be to overload operator*
* add homogeneous normalization (again for vector and set of vectors)
* add a Replicate expression (with uni-directional replication
  facilities)
=> for all of them I'll add examples once we agree on the API
* fix gcc-4.4 warnings
* rename reverse.cpp array_reverse.cpp
2009-03-05 10:25:22 +00:00
Gael Guennebaud
40774c625e add a proof of concept autodiff jacobian helper class based on adolc
with unit test and FindAdolc cmake module
2009-02-27 16:19:13 +00:00
Gael Guennebaud
170128770a compilation fix + test orho methods for complex 2009-02-26 10:09:23 +00:00
Gael Guennebaud
de014efdaf * split CacheFriendlyProduct into multiple smaller files
* add an efficient selfadjoint * vector implementation (= blas symv)
  perf are inbetween MKL and GOTO
  => the interface is still missing (have to be rethougth)
2009-02-21 20:20:38 +00:00
Gael Guennebaud
ee04c5c874 add tests showing bug in unitOrthogonal such that we don't forget it! 2009-02-17 09:57:32 +00:00
Gael Guennebaud
e6f1104b57 * fix Quaternion::setFromTwoVectors (thanks to "benv" from the forum)
* extend PartialRedux::cross() to any matrix sizes with automatic
  vectorization when possible
* unit tests: add "geo_" prefix to all unit tests related to the
  geometry module and start splitting the big "geometry.cpp" tests to
  multiple smaller ones (also include new tests)
2009-02-17 09:53:05 +00:00
Gael Guennebaud
c5245a3388 update vectorization_logic unit test wrt previous sum/redux change 2009-02-13 08:45:19 +00:00
Gael Guennebaud
20a8bb96eb fix m = m*m with m sparse (gug found by Frederik Heinz) 2009-02-12 15:57:13 +00:00
Gael Guennebaud
51c991af45 * exit Sum.h, exit Prod.h, welcome vectorization of redux() !
* add vectorization for minCoeff and maxCoeff
2009-02-12 15:18:59 +00:00
Gael Guennebaud
cbbc6d940b * add ei_predux_mul internal function
* apply Ricard Marxer's prod() patch with fixes for the vectorized path
2009-02-10 18:06:05 +00:00
Gael Guennebaud
a9688f0b71 - add diagonal * sparse product as an expression
- split sparse_basic unit test
- various fixes in sparse module
2009-02-09 09:59:30 +00:00
Gael Guennebaud
3009d79a1f * allow Matrix to be resized to 0 (solve a lot of troubles with
some containers)
* new workaround for std::vector which is supposed to work for any
  classes having EIGEN_MAKE_ALIGNED_OPERATOR_NEW as discussed on ML
2009-02-07 11:16:15 +00:00
Gael Guennebaud
24808dc090 force the use of debug version of QtCore unless it is not available 2009-02-06 21:50:18 +00:00
Gael Guennebaud
1c24f5bbc5 eventually MSVC does not like my /O2 flags (incompatibility with other option set by default) 2009-02-06 15:29:40 +00:00
Gael Guennebaud
f5d96df800 Add vectorization of Reverse (was more tricky than I thought) and
simplify the index based functions
2009-02-06 12:40:38 +00:00
Gael Guennebaud
4dc4ab3abb Reverse::coeff*(int) functions are for vector only 2009-02-06 09:13:04 +00:00
Gael Guennebaud
6fbca94803 apply Ricard patch for Reverse with minor modifications 2009-02-06 09:01:50 +00:00
Gael Guennebaud
dc97079905 add optimization flag for MSVC and heavy tests
remove unsupported namespace
2009-02-05 22:13:50 +00:00
Gael Guennebaud
910b387438 Add sub/super-diagonal expression (read/write) as a trivial extension of
DiagonalCoeffs. The current API is simply:
  m.diagonal<1>() => 1st super diagonal
  m.diagonal<-2>() => the 2nd sub diagonal
I'll add a code snippet once we agree on this API.
2009-02-05 18:37:21 +00:00
Gael Guennebaud
7c374394ee remove explicit fortran dependency in FindCholmod 2009-02-05 10:34:41 +00:00
Gael Guennebaud
da45184635 add custom FindBLAS FindLAPACK working for c++ compiler
fix issues in Cholmod/Taucs supports
2009-02-05 09:36:52 +00:00
Benoit Jacob
5d69c5102b oops, #ifdef instead of #if ---> bug 2009-02-04 16:57:28 +00:00
Benoit Jacob
f81479d392 forgot to update this unit test... 2009-02-04 16:55:38 +00:00
Benoit Jacob
93a089adc8 disable alignment altogether outside of the platforms which potentially have SSE or AltiVec
This should remove most portability issues to other platforms where data alignment issues (including
overloading operator new and new[]) can be tricky, and where data alignment is not needed in the first place.
2009-02-04 16:53:03 +00:00
Gael Guennebaud
95db32fcdc setup the unsupported directory structure.
The unsupported module documentation is automatically generated in:
  build/doc/unsupported/
with bidirectional cross references.
I leave a class Foo in AdolcForward module to illustrate the
cross-reference behavior. I will remove it in the next commit.
2009-02-04 15:37:00 +00:00
Keir Mierle
b9a82be727 Add full pivoting to LDLT decomposition. 2009-02-03 17:50:35 +00:00
Keir Mierle
b4777379d4 Add Matrix::resizeLike(other) convenience function and test. 2009-02-03 01:43:59 +00:00
Gael Guennebaud
36c8a64923 add MatrixBase::stableNorm() avoiding over/under-flow
using it in QR reduced the error of Keir test from 1e-12 to 1e-24 but
that's much more expensive !
2009-01-28 22:11:56 +00:00
Gael Guennebaud
1b194193ef Big change in DiagonalMatrix and Geometry/Scaling:
* previous DiagonalMatrix expression is now DiagonalMatrixWrapper
* DiagonalMatrix class is now for storage
* add the DiagonalMatrixBase class to factorize code of the
  two previous classes
* remove Scaling class (it is now a global function)
* add UniformScaling helper class
  (don't use it directly, use the Scaling function)
* add the Scaling global function to simplify the creation
  of scaling objects
There is still a lot to do, in particular about DiagonalProduct for which
the goal is to get rid of the "if()" in the coeff() function. At least
it is not worse than before ! Also need to uptade the tutorial and add more doc.
2009-01-28 16:26:06 +00:00
Gael Guennebaud
0f15a8d829 QR: add isInjective(), isSurjective(),
mark isFullRank() deprecated,
    add solve() (mix of Keir's patch and LU::solve())
=> there is big problem with complex which are not working
2009-01-28 09:45:53 +00:00
Gael Guennebaud
8ce4503494 add support for read/write sub sets of inner vectors (sparse module) 2009-01-27 22:48:17 +00:00
Benoit Jacob
d384671793 now these tests succeed with 10,000 repeats 2009-01-27 20:47:12 +00:00
Gael Guennebaud
4ac8cabf8a fix my previous commit with EIGEN_EMPTY macro bug 2009-01-27 19:08:20 +00:00
Gael Guennebaud
4b09865b8f check GSL version in cmake files 2009-01-27 16:04:16 +00:00
Gael Guennebaud
d7f60257dd small fix related to GSL cmake stuff 2009-01-26 20:00:41 +00:00
Benoit Jacob
874ff5a0b4 fix msvc warnings (useful ones again) reported by gael on CDash 2009-01-26 17:56:04 +00:00