| Commit message (Collapse) | Author | Age |
|
|
|
|
|
|
| |
* be aware of number of actual householder vectors
(optimization in non-full-rank case, no behavior change)
* fix applyThisOnTheRight, it was using k instead of actual_k
* QR: rename matrixQ() to householderQ() where applicable
|
| |
|
|\ |
|
| |
| |
| |
| | |
compatibility.
|
| | |
|
|\ \ |
|
| | | |
|
| | |
| | |
| | |
| | |
| | |
| | | |
* with random matrices multiplied by 1e+8 (i.e. fixed wrong absolute fuzzy compare)
* with 10,000 repetitions (i.e. the fuzzy compare is really clever)
and when it occasionnally fails, less than once in 10,000 repeats, it is only on the exact rank computation.
|
| | |
| | |
| | |
| | | |
thousands of repetitions and correctly tests matrices of all sizes. Several surprises along the way: for example, a major cause of trouble was the optimized "table of column squared norms" where the accumulation of imprecision was a serious issue; another surprise is that tests like "x!=0" before dividing by x really benefit from being replaced by fuzzy tests, as i hit real cases where i got wrong results in 1/epsilon.
|
|/ / |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
|/ |
|
| |
|
|
|
|
| |
+ rudimentary test
|
| |
|
| |
|
|
|
|
|
| |
* change block selection threshold from 1e-2 to 1e-1
* add rigorous precision test
|
|
|
|
|
|
| |
Only CMake 2.6.3 and later recognize this syntax, and at the moment we
require 2.6.2. CMake uses the right precendence, per its man page, so the
parentheses are not necessary.
|
|
|
|
| |
* expand unit tests to make sure to catch them: they nearly escaped the existing tests as these memory violations were highly dependent on the numbers of rows and cols.
|
| |
|
|
|
|
| |
* remove buggy (superfluous?) specialization in the meta-unroller
|
| |
|
| |
|
|
|
|
|
|
| |
problems with "oops forgot to install new module".
* add a file Eigen/signature_of_eigen3_matrix_library, use it to make FindEigen3.cmake more solid: able to find Eigen in either eigen3/ or eigen/ and not mix it up with Eigen2.
|
|
|
|
|
| |
* bump version to 2.91.0
* add FindEigen3.cmake
|
| |
|
| |
|
|
|
|
|
| |
* old bug: in CwiseBinaryOp: only set the LinearAccessBit if both sides have the same storage order.
* new bug: in Assign.h, only consider linear traversal if both sides have the same storage order.
|
|
|
|
|
|
| |
Actually I don't think it's buggy. But it probably triggers existing bugs, I suspect that
some xprs have LinearAccessBit and shouldn't have it.
Also this fixes the "bugs" with JacobiSVD ---> now it works again
|
|
|
|
| |
I still have JacobiSVD errors when cols>rows
|
|
|
|
| |
Rename some constants to make names match more closely what they mean.
|
| |
|
| |
|
| |
|
|
|
|
| |
subToDense ---> subTo
|
|
|
|
| |
LU: make use of that
|
| |
|
|
|
|
| |
PermutationMatrix: add resize()
|
|
|
|
| |
BandMatrix: rename toDense() ---> toDenseMatrix() for consistency
|
|
|
|
|
| |
* make multiplication order not be reversed
* release-quality documentation
|
|
|
|
| |
remain a NOP
|
| |
|
|
|
|
|
|
|
| |
* DiagonalMatrix:
- add MaxSizeAtCompileTime parameter
- DiagonalOnTheLeft ---> OnTheLeft
- fix bug in DiagonalMatrix::setIdentity()
|