devtools/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2026-04-10 11:34:33 +08:00
Code Issues Packages Projects Releases Wiki Activity
8,696 Commits 15 Branches 93 Tags
97feea9d39ccaf298082cd537e85c311bf354010
Commit Graph

99 Commits

Author SHA1 Message Date
Robert Lukierski
a94791b69a Fixes for min and abs after Benoit's comments, switched to numext. 2016-10-13 15:00:22 +01:00
Robert Lukierski
471075f7ad Fixes min() warnings. 2016-10-12 18:59:05 +01:00
Robert Lukierski
86711497c4 Adding EIGEN_DEVICE_FUNC in the Geometry module. 
Additional CUDA necessary fixes in the Core (mostly usage of
EIGEN_USING_STD_MATH).
 2016-10-12 16:35:17 +01:00
Gael Guennebaud
4057f9b1fc Enable slice-vectorization+inner-unrolling when unaligned vectorization is allowed. For instance, this permits to vectorize 5x5 matrices (including product) 2016-07-28 13:47:33 +02:00
Gael Guennebaud
66917299a9 Add debug output 2016-07-06 22:27:15 +02:00
Gael Guennebaud
367ef66af3 Re-enable some specializations for Assignment<.,Product<>> 2016-07-05 22:58:14 +02:00
Gael Guennebaud
91b3039013 Change the semantic of the last template parameter of Assignment from "Scalar" to "SFINAE" only. 
The previous "Scalar" semantic was obsolete since we allow for different scalar types in the source and destination expressions.
On can still specialize on scalar types through SFINAE and/or assignment functor.
 2016-07-04 11:02:00 +02:00
Gael Guennebaud
101ea26f5e Include the cost of stores in unrolling (also fix infinite unrolling with expression costing 0 like Constant) 2016-06-15 00:01:16 +02:00
Gael Guennebaud
66e99ab6a1 Relax mixing-type constraints for binary coefficient-wise operators: 
- Replace internal::scalar_product_traits<A,B> by Eigen::ScalarBinaryOpTraits<A,B,OP>
- Remove the "functor_is_product_like" helper (was pretty ugly)
- Currently, OP is not used, but it is available to the user for fine grained tuning
- Currently, only the following operators have been generalized: *,/,+,-,=,*=,/=,+=,-=
- TODO: generalize all other binray operators (comparisons,pow,etc.)
- TODO: handle "scalar op array" operators (currently only * is handled)
- TODO: move the handling of the "void" scalar type to ScalarBinaryOpTraits
 2016-06-06 15:11:41 +02:00
Gael Guennebaud
37197b602b Remove debuging code. 2016-05-26 11:53:10 +02:00
Gael Guennebaud
27f0434233 Introduce internal's UIntPtr and IntPtr types for pointer to integer conversions. 
This fixes "conversion from pointer to same-sized integral type" warnings by ICC.
Ideally, we would use the std::[u]intptr_t types all the time, but since they are C99/C++11 only,
let's be safe.
 2016-05-26 10:52:12 +02:00
Gael Guennebaud
e68e165a23 bug #256: enable vectorization with unaligned loads/stores. 
This concerns all architectures and all sizes.
This new behavior can be disabled by defining EIGEN_UNALIGNED_VECTORIZE=0
 2016-05-24 21:54:03 +02:00
Gael Guennebaud
64bb7576eb Clean propagation of Dest/Src alignments. 2016-05-24 17:12:12 +02:00
Christoph Hertzberg
33ca7e3c8d bug #1207: Add and fix logical-op warnings 2016-05-11 19:36:34 +02:00
Gael Guennebaud
b1bd53aa6b Fix performance regression: with AVX, unaligned stores were emitted instead of aligned ones for fixed size assignement. 2016-05-01 23:25:06 +02:00
Gael Guennebaud
06447e0a39 Improve half-packet vectorization logic to distinguish linear versus inner traversal modes. 2016-04-13 18:15:49 +02:00
Gael Guennebaud
675e0a2224 Fix static/inline keywords order. 2016-04-11 15:06:20 +02:00
Benoit Steiner
bff8cbad06 Removed executable bit from header files 2016-03-23 16:14:23 -07:00
Gael Guennebaud
a4c76f8d34 Improve inlining 2016-02-08 11:33:02 +01:00
Gael Guennebaud
d142165942 bug #667: declare several critical functions as FORECE_INLINE to make ICC happier. 
<g.gael@free.fr> HG: branch 'default' HG: changed Eigen/src/Core/ArrayBase.h HG: changed Eigen/src/Core/AssignEvaluator.h HG: changed
Eigen/src/Core/CoreEvaluators.h HG: changed Eigen/src/Core/CwiseUnaryOp.h HG: changed Eigen/src/Core/DenseBase.h HG: changed Eigen/src/Core/MatrixBase.h
 2016-01-31 16:34:10 +01:00
Gael Guennebaud
8b9dc9f0df bug #1144: fix regression in x=y+A*x (aliasing), and move evaluator_traits::AssumeAliasing to evaluator_assume_aliasing. 2016-01-09 08:30:38 +01:00
Gael Guennebaud
30b5c4cd14 Remove useless "explicit", and fix inline/static order. 2015-12-11 10:59:39 +01:00
Gael Guennebaud
8531304858 Simplify cost computations based on HugeCost being smaller that unrolling limit 2015-10-28 13:39:02 +01:00
Gael Guennebaud
77ff3386b7 Refactoring of the cost model: 
- Dynamic is now an invalid value
 - introduce a HugeCost constant to be used for runtime-cost values or arbitrarily huge cost
 - add sanity checks for cost values: must be >=0 and not too large
This change provides several benefits:
 - it fixes shortcoming is some cost computation where the Dynamic case was not properly handled.
 - it simplifies cost computation logic, and should avoid future similar shortcomings.
 - it allows to distinguish between different level of dynamic/huge/infinite cost
 - it should enable further simplifications in the computation of costs (save compilation time)
 2015-10-28 11:42:14 +01:00
Gael Guennebaud
12f50a4697 Fix assign vectorization logic with respect to fixed outer-stride 2015-10-27 11:04:19 +01:00
Gael Guennebaud
0fc8954282 Improve readibility of EIGEN_DEBUG_ASSIGN mode. 2015-10-27 10:38:49 +01:00
Gael Guennebaud
5cc7251188 Some cleaning in evaluators 2015-10-08 15:22:04 +02:00
Gael Guennebaud
aba1eda71e Help clang to inline some functions, thus fixing some regressions 2015-10-07 15:44:12 +02:00
Gael Guennebaud
41cc1f9033 Remove debuging prod() and lazyprod() function, plus some cleaning in noalias assignment 2015-10-07 15:41:22 +02:00
Gael Guennebaud
26cde4db3c Define Permutation*<>::Scalar to 'void', re-enable scalar type compatibility check in assignment while relaxing this test for void types. 2015-10-06 17:18:06 +02:00
Gael Guennebaud
fb51bab272 Some cleaning 2015-10-06 17:14:56 +02:00
Gael Guennebaud
92b9f0e102 Cleaning pass on evaluators: remove the useless and error prone evaluator<>::type indirection. 2015-09-02 21:38:40 +02:00
Gael Guennebaud
febcce34f1 Enable vectorization with half-packets 2015-08-07 20:05:31 +02:00
Gael Guennebaud
65bfa5fce7 Allow to use arbitrary packet-types during evaluation. 
This is implemented by adding a PacketType template parameter to packet and writePacket members of evaluator<>.
 2015-08-07 12:01:39 +02:00
Gael Guennebaud
ce57dbd937 Let unpacket_traits<> exposes the required alignment and make use of it everywhere 2015-08-07 10:44:01 +02:00
Gael Guennebaud
2afdef6a54 Generalize first_aligned to take the requested alignment as a template parameter, and add a first_default_aligned variante calling first_aligned with the requirement of the largest packet for the given scalar type. 2015-08-06 17:52:01 +02:00
Gael Guennebaud
1f5024332e First part of a big refactoring of alignment control to enable the handling of arbitrarily aligned buffers. It includes: 
- AlignedBit flag is deprecated. Alignment is now specified by the evaluator through the 'Alignment' enum, e.g., evaluator<Xpr>::Alignment. Its value is in Bytes.
 - Add several enums to specify alignment: Aligned8, Aligned16, Aligned32, Aligned64, Aligned128. AlignedMax corresponds to EIGEN_MAX_ALIGN_BYTES. Such enums are used to define the above Alignment value, and as the 'Options' template parameter of Map<> and Ref<>.
 - The Aligned enum is now deprecated. It is now an alias for Aligned16.
 - Currently, traits<Matrix<>>, traits<Array<>>, traits<Ref<>>, traits<Map<>>, and traits<Block<>> also expose the Alignment enum.
 2015-08-06 15:31:07 +02:00
Christoph Hertzberg
ea87561564 bug #1039: Redefining EIGEN_DEFAULT_DENSE_INDEX_TYPE may lead to errors 2015-07-13 16:08:25 +02:00
Gael Guennebaud
36643eec0c Add a call_assignment_no_alias_no_transpose shortcut 2015-06-24 17:50:43 +02:00
Gael Guennebaud
972a535288 Remove aligned-on-scalar assert and fallback to non vectorized path at runtime (first_aligned already had this runtime guard) 2015-06-14 15:04:07 +02:00
Christoph Hertzberg
e8cdbedefb bug #877, bug #572: Introduce a global Index typedef. Rename Sparse*::Index to StorageIndex, make Dense*::StorageIndex an alias to DenseIndex. Overall this commit gets rid of all Index conversion warnings. 2014-12-04 22:48:53 +01:00
Gael Guennebaud
ee06f78679 Introduce unified macros to identify compiler, OS, and architecture. They are all defined in util/Macros.h and prefixed with EIGEN_COMP_, EIGEN_OS_, and EIGEN_ARCH_ respectively. 2014-11-04 21:58:52 +01:00
Christoph Hertzberg
d3f52debc6 Make cuda_basic test compile again by adding lots of EIGEN_DEVICE_FUNC. 
Although the test passes now, there might still be some missing.
 2014-10-13 17:18:26 +02:00
Georg Drenkhahn
edaefeb978 Using Kernel::Index type instead of int to prevent possible implicit conversion from long to int. 2014-09-21 10:01:12 +02:00
Gael Guennebaud
0ca43f7e9a Remove deprecated code not used by evaluators 2014-09-18 15:15:27 +02:00
Gael Guennebaud
c83e01f2d6 Favor column major storage for inner products 2014-09-14 19:38:49 +02:00
Gael Guennebaud
26db954776 Re-enable aliasing checks when using evaluators 2014-09-14 19:06:08 +02:00
Gael Guennebaud
9b729f93a1 Resizing is done by call_assignment_noalias, so no need to perform it when dealing with aliasing. 2014-07-21 11:46:47 +02:00
Gael Guennebaud
62f332fc04 Make sure we evaluate into temporaries matching evaluator storage order requirements 2014-07-19 15:19:10 +02:00
Gael Guennebaud
b868bfb84a Make operator=(EigenBase<>) uses the new assignment mechanism and introduce a generic EigenBase to EigenBase assignment kind based on the previous evalTo mechanism. 2014-06-25 17:23:52 +02:00