summaryrefslogtreecommitdiff
path: root/numpy/core/blasdot
Commit message (Collapse)AuthorAgeFilesLines
* ENH: Move vdot to multiarray.Charles Harris2014-09-042-811/+0
| | | | | | | | | | Remove vdot from _dotblas and implement it in multiarray. Remove the files in core/blasdot as they are no longer needed. Fix tests and build to reflect the changes. The vdot function is now a bit faster in the common case as ravel is used instead of flatten. There is also no need to conjugate the files for clongdouble.
* ENH: Move dotblas_innerproduct down into multiarray.Charles Harris2014-09-041-289/+0
| | | | | | | Move the dotblas_innerproduct function from the _dotblas.c file to a new cblasfuncs.c file in multiarray and rename it cblas_innerproduct. Modify it so that it can called directly from PyArray_InnerProduct and do so. Fix up numeric.py and the tests to reflect these changes.
* ENH: Move dotblas_matrixproduct down into multiarray.Charles Harris2014-09-041-571/+14
| | | | | | | 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.
* MAINT, STY: Remove use of alterdot, restoredot in _dotblas.c.Charles Harris2014-09-041-151/+108
| | | | These are no longer needed. Also do C style cleanups.
* ENH: When cblas is available use it in descr->f->dot.Charles Harris2014-09-041-133/+2
| | | | | | | | | | Importing _dotblas currently executes _dotblas.alterdot, which replaces the default descr->f->dot function with a cblas based version for float, double, complex float, and complex double data types. This PR changes the default descr->f->dot to use cblas whenever it is available. After this change, the alterdot and restoredot functions serve no purpose, so are changed to do nothing and deprecated. Note that those functions were already doing nothing when _dotblas was not available.
* ENH: dotblas: re-route all cblas_?dot calls to chunked versionsLars Buitinck2013-11-211-76/+71
| | | | Also factored out and simplified NumPy->BLAS stride conversion.
* BUG/ENH: blasdot: make *dot calls 64-bit safeLars Buitinck2013-11-211-37/+83
| | | | | | | | | | | | | | BLAS uses int for all its size types, so the size is going to be silently truncated on platforms where sizeof npy_intp > sizeof int, i.e. on modern 64-bit platforms. In this case, we do a dot product in reasonably-sized chunks and add the result ourselves. On platforms where sizeof(int) == sizeof(npy_intp), i.e. typical 32-bit platforms, it's up to the compiler to optimize the loop away. Also tried to make the code more readable, replaced int by npy_intp in a few places and removed useless register declarations (modern compilers ignore those).
* MAINT: dotblas: factor out all gemm and gemv callsLars Buitinck2013-10-261-167/+83
|
* BLD: fix Bento build.Ralf Gommers2013-09-191-1/+1
| | | | Issue was introduced with PR-3524, the ufunc override stuff.
* ENH: Add ufunc override functionality to ufuncs and dots.Blake Griffith2013-09-051-2/+26
|
* MAINT: Remove outdated version checks.Charles Harris2013-07-111-6/+0
| | | | | | | | | | Because Numpy 1.8.0 will no longer supports Python versions < 2.6 we no longer need to check for that and can also remove the code that is specific to those earlier versions. To make this a bit safer, the toplevel setup.py file now contains a check of the Python version number and raises an error when run by an unsupported version.
* Removed trailing printf statementNicolas Scheffer2012-11-121-1/+0
|
* ENH: Remove the need for temporary copies in numpy.dotNicolas Scheffer2012-11-121-3/+15
| | | | | | | in numpy core, modified the blas function call to handle C and F order arrays without the need for copy. This brings a significant speed up for matrix multiplication Added a full test for matrix multiplication in test_blasdot
* Use PyMODINIT_FUNC and update docs accordingly.cgohlke2012-09-021-1/+1
| | | | See https://github.com/scipy/scipy/pull/279
* WRN: Fix some compiler warnings.Charles Harris2012-04-061-7/+11
|
* UPD: Make _dotblas.c use non-deprecated macros/functions.Charles Harris2012-04-061-249/+249
|
* UPD: Add NPY_NO_DEPRECATED_API to some more files.Charles Harris2012-04-061-2/+8
| | | | _dotblas needs more work to remove direct access to ndarray internals.
* UPD: Various fixes, Remove #define NPY_NO_PREFIX from files in core.Charles Harris2012-02-051-1/+1
|
* STY: Replace more found macros in old_defines.h with new.Charles Harris2012-02-041-78/+78
|
* ENH: core: Allow user to pass in output array for dot()Luis Pedro Coelho2011-02-121-12/+44
| | | | | | | | | This avoids the memory allocation. It is strict in checking that the types are correct, but since it is intended as an optimisation, it should only be used when the user knows what they are doing. The out parameter is added both to the BLAS and non-BLAS versions of dot(). Tests are included.
* 3K: fix core/dotblas module initialization.Pauli Virtanen2010-08-041-3/+30
|
* BUG: ensure Py_TYPE is available in _dotblas.c on Python 2.4 and 2.5Pauli Virtanen2010-02-201-0/+6
|
* 3K: Some remaining Py_TYPE issuesPauli Virtanen2010-02-201-1/+1
|
* Fix _dotblas ob_type pointer for py3k.Charles Harris2010-02-131-4/+4
|
* Hard tab removal: _dotblas.cCharles Harris2009-10-201-663/+663
|
* Ensure that documentation for dot, vdot, inner, alterdot, restoredot is the ↵Pauli Virtanen2009-03-211-23/+32
| | | | same, independent of whether these functions come from _dotblas or multiarray/numeric.py
* Fixed #950: fix bug in dotblas 0D input detection.Pauli Virtanen2009-03-101-0/+5
| | | | | One of the branches in dotblas_matrixproduct tried to dereference the arr->data of a 0D array. This should plug it.
* Fix signed/unsigned warnings by doing explicit cast (as done by the compiler ↵David Cournapeau2009-03-091-10/+10
| | | | implicitely anyway).
* Do not use old noprefix header in dotblas (#945).David Cournapeau2009-03-021-44/+44
|
* Use NPY_UNUSED in numpy.core sources.David Cournapeau2008-10-081-5/+5
|
* Fix buggy sentinel for blasdot docstring.David Cournapeau2008-08-031-1/+1
|
* Remove blank line.Charles Harris2008-07-161-1/+0
|
* Fix ticket #849. Thanks to Michael Abbott.Charles Harris2008-07-161-4/+9
| | | | | | The added check for NULL descr isn't really needed here because the typenums used don't allow such returns. But knowing that requires knowledge of PyArray_DescrFromType internals and not checking makes the code fragile.
* Coding style cleanups.Charles Harris2008-07-161-69/+125
|
* * Make matrices return 1-dimensional array on item selection for 1.2\n * ↵Travis Oliphant2008-05-071-0/+5
| | | | Remove (now un-needed) tolist method from matrices\n * For ticket #551, copy data if start of memory is not aligned on itemsize location in optimized blas wrapper.
* Fix Ticket #588: problem with negative striding and fast blas implementation ↵Travis Oliphant2007-10-091-7/+28
| | | | of dot
* Cleanup whitespace.Charles Harris2007-04-112-174/+174
|
* Fix warnings found by Intel compiler due to unused variables that were set. ↵Travis Oliphant2006-12-021-2/+2
| | | | Make ones work for compound types.
* Fix reference count problem with dtypes in vdotTravis Oliphant2006-11-171-3/+3
|
* Add output arguments to a few more functions for consistencyTravis Oliphant2006-08-101-1/+1
|
* Convert so that axis arguments are consistent for methods and functions. ↵Travis Oliphant2006-08-041-1/+1
| | | | Place functions for which this changes the Numeric default into oldnumeric.
* Fix-up usage of NPY_ALLOW_THREADS. Default is now WITH_THREAD but an ↵Travis Oliphant2006-07-181-6/+17
| | | | environment variable NUMPY_NOSMP can be set which will make the default 0
* Make sure generated C-API uses new names.Travis Oliphant2006-07-181-1/+0
|
* Allow threads at several places in the code.Travis Oliphant2006-07-181-2/+11
|
* Fix ticket #177Travis Oliphant2006-07-121-1/+0
|
* Apply npy_ and NPY_ prefixes to all C-API names that don't already have ↵Travis Oliphant2006-07-081-1/+1
| | | | PyArray_ prefixes.
* Fix defect #115Travis Oliphant2006-05-141-2/+2
|
* Fix ticket #109Travis Oliphant2006-05-111-1/+1
|
* Fix ticket #106Travis Oliphant2006-05-101-4/+6
|
* Fix-up optimized dot when strides is not a multiple of the element size.Travis Oliphant2006-05-091-5/+35
|