summaryrefslogtreecommitdiff
path: root/numpy/linalg
Commit message (Collapse)AuthorAgeFilesLines
...
* ENH: Add keepdims to linalg.normEric Moore2014-10-172-15/+72
|
* ENH: Move dotblas_matrixproduct down into multiarray.Charles Harris2014-09-041-1/+1
| | | | | | | Move the dotblas_matrixproduct function from the _dotblas.c file to a new cblasfuncs.c file in multiarray and rename it cblas_matrixproduct. Modify it so that it can called directly from PyArray_MatrixProduct2 and do so. Fix up numeric.py and core/__init__.py to reflect these changes.
* TST: pass sys.argv to run_module_suite by defaultJulian Taylor2014-09-041-3/+5
| | | | | allows passing flags like --pdb to test files also add call to files where its missing
* MAINT: linalg: use sizeof instead of magic constantLars Buitinck2014-08-271-3/+2
| | | | Allows changing the error message at some point.
* DOC: fix return shape of tensorinv docFrederic2014-07-171-1/+1
|
* TST: add more regression testing related to ↵alex2014-05-261-0/+3
| | | | https://github.com/numpy/numpy/issues/4733
* BUG: svd ufunc typoalex2014-05-222-2/+10
|
* BUG: Delay npyiter size check when size may changeSebastian Berg2014-03-121-0/+6
| | | | | | | | | | | | | When a multi index is tracked and RemoveAxis can be called, the size of the iterator may still change. This was causing failures for example for the SVD, because the gufunc machinery requires a temporarily larger iterator for output allocation. Thanks to Jaime (jaime.frio@gmail.com) for noting that this is plausible since the size check can be delayed pretty ok up until GetIterNext (or similar functions). Closes gh-4442
* DOC: Fixed documentation on lstsq function on when it return an empty ↵Cimarron Mittelsteadt2014-03-061-1/+1
| | | | residuals array
* BUG: Fix wrong variable in test_build.py exception.Charles Harris2014-02-151-5/+5
| | | | | | | The variable used was not defined, causing an error when the exception was raised. Closes #3146.
* MAINT: fix build warnings for umath_linalgRalf Gommers2014-01-191-25/+4
| | | | | The removal of CFLOAT for eig/eigvals is related to a bug, see c355550dd80. Can be put back once that issue is fixed.
* BUG: fix build with single-threaded pythonUlrich Seidl2014-01-161-2/+8
| | | | | | | Some symbols are not defined leading to: ImportError Undefined symbol: PyGILState_Release Closes gh-3255
* MAINT: linalg: prune unused routines from lapack_litePauli Virtanen2014-01-122-595/+6
|
* Merge pull request #3895 from larsmans/norm-dotCharles Harris2014-01-091-1/+6
|\ | | | | ENH: use dot in linalg.norm
| * ENH: optimize complex vector normLars Buitinck2014-01-031-2/+6
| | | | | | | | Also used ravel(order='K') to prevent copies.
| * ENH: use dot (BLAS!) in linalg.normLars Buitinck2013-10-111-1/+2
| |
* | Merge pull request #4149 from argriffing/doc-matrix-rankCharles Harris2014-01-021-0/+1
|\ \ | | | | | | DOC: matrix_rank
| * | DOC: more matrix_rank docsalex2013-12-261-0/+1
| | |
* | | BUG: fix complex norm of higher orderJulian Taylor2014-01-022-1/+13
| | | | | | | | | | | | | | | | | | asfarray truncates the complex part, so it must be avoided for complex types. Closes gh-4156.
* | | BUG: linalg: norm fails on longdouble, signed intLars Buitinck2013-12-262-4/+24
|/ / | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This fixes the following bug with longdouble: >>> x = np.arange(10, dtype=np.longdouble) >>> np.linalg.norm(x, ord=3) Traceback (most recent call last): File "<ipython-input-5-7ee53a8ac142>", line 1, in <module> np.linalg.norm(x, ord=3) File "/tmp/v/lib/python2.7/site-packages/numpy/linalg/linalg.py", line 2090, in norm return add.reduce(absx**ord, axis=axis)**(1.0/ord) UnboundLocalError: local variable 'absx' referenced before assignment As well as the handling of minimal values for signed integers: >>> x = np.array([-2**31], dtype=np.int32) >>> np.linalg.norm(x, ord=3) /tmp/v/lib/python2.7/site-packages/numpy/linalg/linalg.py:2090: RuntimeWarning: invalid value encountered in double_scalars return add.reduce(absx**ord, axis=axis)**(1.0/ord) nan
* | Merge pull request #3974 from juliantaylor/fenv-improveCharles Harris2013-11-061-3/+3
|\ \ | | | | | | ENH: avoid expensive clears in fenv functions
| * | ENH: avoid expensive clears in fenv functionsJulian Taylor2013-11-061-3/+3
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Clearing is 50-100 times more expensive than checking on x86, so check if there is anything needs to be cleared first. This speeds up scalar operations by 10%-20%. Based on Arink Verma code in #3739. Implement the functions as new C-API functions npy_get_floatstatus and npy_clear_floatstatus in npy_math.
* | | MAINT: Refactor eigh and eigvalsh and associated tests.Charles Harris2013-10-262-47/+56
| | | | | | | | | | | | | | | | | | Do not convert the UPLO argument to a bytestring, it is not necessary. Distribute parts of the eigh tests into the appropriate TestEigvalsh or TestEigh test class.
* | | BUG: reject invalid UPLO with ValueError in eigh/eigvalshJulian Taylor2013-10-252-5/+35
| | | | | | | | | | | | to prevent unintentional use of wrong function. Restores 1.7 behavior.
* | | BUG: fix broken UPLO of eigh in python3Julian Taylor2013-10-252-4/+26
| | | | | | | | | | | | | | | | | | UPLO was cast to bytes and compared to a string which is always false in python3. closes gh-3977
* | | TST: Initialize some arrays to to avoid valgrind warningsSebastian Berg2013-10-241-1/+1
|/ /
* | MAINT: double underscores in Lapack-lite C moduleLars Buitinck2013-10-191-68/+70
| | | | | | | | | | | | Strictly speaking, any non-standard name with "__" in it causes undefined behavior in C. It's also completely unnecessary to use any in this module.
* | STY: linalg: fix C comment stylePauli Virtanen2013-10-121-6/+10
| |
* | BUG: linalg: don't tell BLAS to use zero stridesPauli Virtanen2013-10-122-6/+52
|/ | | | At least OSX Accelerate fails for this case.
* TST: linalg: better rtol choicePauli Virtanen2013-10-081-1/+5
|
* TST: linalg: add slightly bigger test case + don't catch KeyboardInterruptPauli Virtanen2013-10-081-1/+9
|
* MAINT: linalg: strip out unused gufuncs_linalg codePauli Virtanen2013-10-073-3109/+4
| | | | | Some of the functions removed were untested, and apparently did not function correctly (cholesky_up, inner1d, maybe more).
* BUG: linalg: fix eigvalsh return type (always real-valued)Pauli Virtanen2013-10-071-1/+1
|
* BUG: linalg: use correct BLAS incx convention in xCOPYPauli Virtanen2013-10-071-6/+22
|
* TST: linalg: add more comprehensive test cases for linalg funcsPauli Virtanen2013-10-071-223/+332
| | | | | | Also remove TestCase subclassing, so that generator tests work. Also fix bugs in the existing generator tests that were not actually run previously.
* BUG: Allos linalg.solve to handle zero equations.Sebastian Berg2013-10-012-4/+23
| | | | | This was previously correct, but the special case was missing since the gufunc code cannot handle it.
* BUG: fix small issues found with cppcheckJulian Taylor2013-09-291-1/+1
| | | | | | uninitialized use in _strided_to_strided_one_to_n_with_finish wrong error NULL checks in PyArray_MapIterNew and einsum_sub_op_from_str double free and memleak in umath_linalg.c.src init_@lapack_func@
* MAINT: Make the qr raw mode test independent of the LAPACK library.Charles Harris2013-09-121-38/+22
| | | | | | | | The qr factorization is not unique and the values returned by the raw mode may differ between LAPACK versions. Consequently, the results cannot be checked against known values. Closes #3703.
* MAINT: Refactor test_xerbla to hide printed error messages.Charles Harris2013-09-111-36/+25
| | | | | | | | | | | Combine the xerbla override tests with the previous _is_xerbla_safe decorator. The xerbla checks are then made in a forked process where the outputs are closed so that if xerbla prints a message it does not appear. It would be nice to figure out a Windows version of this test. Closes #3704. Closes #3687.
* STY: Giant comma spacing fixup.Charles Harris2013-08-186-120/+120
| | | | | | | Run the 2to3 ws_comma fixer on *.py files. Some lines are now too long and will need to be broken at some point. OTOH, some lines were already too long and need to be broken at some point. Now seems as good a time as any to do this with open PRs at a minimum.
* STY: Giant whitespace cleanup.Charles Harris2013-08-188-302/+299
| | | | Now is as good a time as any with open PR's at a low.
* Merge pull request #2721 from abalkin/masterCharles Harris2013-08-151-1/+1
|\ | | | | Mistake in linalg.eig() documentation
| * Fixed a mistake in linalg.eig() documentation.Alexander Belopolsky2012-11-091-1/+1
| |
* | ENH: inv/solve work with empty inner and others empty outer arraySebastian Berg2013-08-042-16/+82
| | | | | | | | | | | | | | This makes the inverse of a 0x0 array simply be 0x0 again. It also modifies the no-empty array check in favor of a no-empty *inner* array, since the gufuncs seem to handle the other case fine.
* | MAINT: linalg: some more clean up in norm: 'nd' was defined but not usedWarren Weckesser2013-07-201-5/+4
| |
* | BUG: linalg: norm was computing and then ignoring a productWarren Weckesser2013-07-191-2/+1
| |
* | MAINT: Use np.errstate context manager.Charles Harris2013-07-111-5/+6
| | | | | | | | | | | | | | | | | | | | | | | | | | Now that Python < 2.6 is no longer supported we can use the errstate context manager in places where constructs like ``` old = seterr(invalid='ignore') try: blah finally: seterr(**old) ``` were used.
* | DOC: another minor tweak of linalg.lstsq doc stringFazlul Shahriar2013-06-131-1/+2
| |
* | DOC: minor tweak of linalg.lstsq doc stringFazlul Shahriar2013-06-131-2/+1
| |
* | DOC: fix typo in linalg.lstsq doc stringFazlul Shahriar2013-06-131-2/+2
| |
View Lorry Controller Status