915587d03d* add CommaInitializer::finished to allow the use of (Matrix3() << v0, v1, v2).finished() as an argument of a function. Other possibilities for the name could be "end" or "matrix" ?? * various update in Quaternion, in particular I added a lot of FIXME about the API options, these have to be discussed and fixed.
Gael Guennebaud
2008-06-03 15:50:09 +0000
196f38f5dbimproved Quaternion class: - Euler angles and angle axis conversions, - stable spherical interpolation - documentation - update the respective unit test
Gael Guennebaud
2008-06-03 13:43:29 +0000
bcb32839c2fix building of examples
Gael Guennebaud
2008-06-03 09:43:59 +0000
a9cf229e15add a geometry unit test and fix a couple of typo in Quaternion.h
Gael Guennebaud
2008-06-03 07:32:12 +0000
8de4d92b70- get the doc of the enums in MatrixBase right - get the doc of the flags in Constants right - finally give up with SEPARATE_MEMBER_PAGES: it triggers too big Doxygen bugs, and produces too many small pages. So we have one huge page for MatrixBase at currently 300kb and going up, so the solution especially for users with low bandwidth will be to provide an archive of the html documentation.
Benoit Jacob
2008-06-03 02:06:18 +0000
366971bea4* start of the Geometry module with a cross product and quaternion expressions (haven't tried them yet) * applied the meta selector rule to MatrixBase::swap()
Gael Guennebaud
2008-06-02 22:58:36 +0000
75de41a00bbig changes in Doxygen configuration; work around bug with doxygen parsing of initialized enum values showing the last word the initializer instead of the actual enum value's name; add some more docs.
Benoit Jacob
2008-06-02 20:08:37 +0000
ac88feebb7work around Doxygen bug triggered by r814874, which caused many classes to disappear from the docs.
Benoit Jacob
2008-06-02 19:29:23 +0000
6209bbe286doc improvements: fix linking in Mainpage.dox, improved Doxyfile.in
Benoit Jacob
2008-06-02 18:27:37 +0000
f2ebbee274added unit tests for sizeof and dynamic memory allocation
Gael Guennebaud
2008-06-02 14:54:52 +0000
54ae2ac7e8Added the computation of eigen vectors in the symmetric eigen solver. However the eigen vectors are not correct yet, but I really cannot find the problem.
Gael Guennebaud
2008-06-02 12:52:08 +0000
3b0523041asince m*m.adjoint() is positive, so are its eigenvalues, so no need for cwiseAbs()
Benoit Jacob
2008-06-02 04:45:02 +0000
0444e3601a- add MatrixBase::eigenvalues() convenience method - add MatrixBase::matrixNorm(); in the non-selfadjoint case, we reduce to the selfadjoint case by using the "C*-identity" a.k.a. norm of x = sqrt(norm of x * x.adjoint())
Benoit Jacob
2008-06-02 04:42:45 +0000
92b7e2d6a1fix a couple of issues making the eigensolver test compile and run without aborting on an assert. Had to fix a stupid bug in Block -- very strange we hadn't hit it before.
Benoit Jacob
2008-06-02 02:06:33 +0000
001b01a290Rewrite from scratch of the eigen solver for symmetric matrices which now supports selfadjoint matrix. The implementation follows Golub's famous book.
Gael Guennebaud
2008-06-02 00:30:26 +0000
06752b2b77* added a Tridiagonalization class for selfadjoint matrices * added MatrixBase::real() * added the ability to extract a selfadjoint matrix from the lower or upper part of a matrix, e.g.: m.extract<Upper|SelfAdjoint>() will ignore the strict lower part and return a selfadjoint. This is compatible with ZeroDiag and UnitDiag.
Gael Guennebaud
2008-06-01 17:20:18 +0000
dc5fd8dfffmeagre outcome for so much time spent! * fix inverse() bug discovered by Gael's test * fix warnings introduced by the new Diagonal stuff * update Doxyfile to v1.5.6
Benoit Jacob
2008-06-01 03:36:49 +0000
d5cbb1d002added a unit test for Inverse discovered that m1 == m1.inverse().inverse() fails though m1.inverse()*m1 == I
Gael Guennebaud
2008-05-31 23:44:11 +0000
64169389edadded an *optional* Eigen2 dynamic library. it allows the possiblity to save some compilation time by linking to it *and* defining the token EIGEN_EXTERN_INSTANCIATIONS
Gael Guennebaud
2008-05-31 23:21:49 +0000
fcf4457b78added optimized matrix times diagonal matrix product via Diagonal flag shortcut.
Gael Guennebaud
2008-05-31 21:35:11 +0000
310f7aa096moved purely "array" related stuff to a new module Array. This include: - cwise Pow,Sin,Cos,Exp... - cwise Greater and other comparison operators - .any(), .all() and partial reduction - random
Gael Guennebaud
2008-05-31 18:11:48 +0000
a2f71f9d7eupdated EigenSolver to use .coeff / .coeffRef
Gael Guennebaud
2008-05-31 16:31:10 +0000
c9fb248c36simply a bit the basic product moving dynamic loops to the corresponding special case of the unrollers. the latter ones are therefore re-named *product_impl.
Gael Guennebaud
2008-05-31 15:06:26 +0000
f5e599e489* replace compile-time-if by meta-selector in Assign.h as it speed up compilation. * fix minor typo introduced in the previous commit
Gael Guennebaud
2008-05-31 14:42:07 +0000
e2ac5d244eAdded ArrayBit to get the ability to manipulate a Matrix like a simple scalar. In particular this flag changes the behavior of operator* to a coeff wise product.
Gael Guennebaud
2008-05-29 22:33:07 +0000
b501e08d81now the unit-tests (hence all of Eigen) don't depend on Qt at all anymore.
Benoit Jacob
2008-05-29 03:37:16 +0000
486fdb26a1many small fixes and documentation improvements, this should be alpha5.
Benoit Jacob
2008-05-29 03:12:30 +0000
c1559d3079* updated the assignement operator macro so that overloads in MatrixBase work * removed product_selector and cleaned Product.h a bit * cleaned Assign.h a bit
Gael Guennebaud
2008-05-28 22:56:19 +0000
8711e26c8a* change Flagged to take into account NestByValue only * bugfix in Assign and cache friendly product (weird that worked before) * improved argument evaluation in Product
Gael Guennebaud
2008-05-28 22:11:47 +0000
73084dc754* added _*coeffRef members in NestedByValue * added ConjugateReturnType and AdjointReturnType that are type-defined to Derived& and Transpose<Derived> if the scalar type is not complex: this avoids abusive copies in the cache friendly Product
Gael Guennebaud
2008-05-28 09:09:18 +0000
f54760c889hehe, the complicated nesting scheme in Flagged in the previous commit was a sign that we were doing something wrong. In fact, having NestByValue as a special case of Flagged was wrong, and the previous commit, while not buggy, was inefficient because then when the resulting NestByValue xpr was nested -- hence copied -- the original xpr which was already nested by value was copied again; hence instead of 1 copy we got 3 copies. The solution was to ressuscitate the old Temporary.h (renamed NestByValue.h) as it was the right approach.
Benoit Jacob
2008-05-28 05:14:16 +0000
aebecae510* find the proper way of nesting the expression in Flagged: finally that's more subtle than just using ei_nested, because when flagging with NestByValueBit we want to store the expression by value already, regardless of whether it already had the NestByValueBit set. * rename temporary() ----> nestByValue() * move the old Product.h to disabled/, replace by what was ProductWIP.h * tweak -O and -g flags for tests and examples * reorder the tests -- basic things go first * simplifications, e.g. in many methoeds return derived() and count on implicit casting to the actual return type. * strip some not-really-useful stuff from the heaviest tests
Benoit Jacob
2008-05-28 04:38:16 +0000
559233c73e* fix the QR module to use extract/part instead of the previous triangular stuff * added qr and eigensolver tests * fix a compilation warning in Matrix copy constructor
Gael Guennebaud
2008-05-27 09:16:27 +0000
5aa00f6870part 2 of big change: rename Triangular.h -> Extract.h (svn required to commit that separately)
Benoit Jacob
2008-05-27 05:50:36 +0000
953efdbfe7- introduce Part and Extract classes, splitting and extending the former Triangular class - full meta-unrolling in Part - move inverseProduct() to MatrixBase - compilation fix in ProductWIP: introduce a meta-selector to only do direct access on types that support it. - phase out the old Product, remove the WIP_DIRTY stuff. - misc renaming and fixes
Benoit Jacob
2008-05-27 05:47:30 +0000
8f1fc80a77some documentation fixes (Cwise* and Cholesky)
Gael Guennebaud
2008-05-22 16:31:00 +0000
94e1629a1b* improved product performance: - fallback to normal product for small dynamic matrices - overloaded "c += (a * b).lazy()" to avoid the expensive and useless temporary and setZero() in such very common cases. * fix a couple of issues with the flags
Gael Guennebaud
2008-05-22 14:51:25 +0000
106a0c1befrestored the product test
Gael Guennebaud
2008-05-22 12:35:09 +0000
9ab6e186ebremove Like1DArrayBit in Transpose
Gael Guennebaud
2008-05-22 12:25:11 +0000
522e24f2d7update of the testing framework: replaced the QTestLib framework my custom macros and a (optional) custom script to run the tests from ctest.
Gael Guennebaud
2008-05-22 12:18:55 +0000
c6789a279cFix compilation issues with MSVC and NVCC. Added a few typedef of complex return types in MatrixBase (Needed by MSVC)
Gael Guennebaud
2008-05-15 09:40:11 +0000
5da60897abIntroduce generic Flagged xpr, remove already Lazy.h and Temporary.h Rename DefaultLostFlagMask --> HerediraryBits
Benoit Jacob
2008-05-14 08:20:15 +0000
fd2e9e5c3c* Clean a bit the eigenvalue solver: if the matrix is known to be selfadjoint at compile time, then it returns real eigenvalues. * Fix a couple of bugs with the new product.
Gael Guennebaud
2008-05-13 07:40:25 +0000
3eccfd1a78-fix certain #includes -fix CMakeLists, public headers weren't getting installed
Benoit Jacob
2008-05-12 21:15:17 +0000
4317fad869* Added several cast to int of the enums (needed for some compilers) * Fix a mistake in CwiseNullary. * Added a CoreDeclarions header that declares only the forward declarations and related basic stuffs.
Gael Guennebaud
2008-05-12 18:09:30 +0000
678f18fce4put inline keywords everywhere appropriate. So we don't need anymore to pass -finline-limit=1000 to gcc to get good performance. By the way some cleanup.
Benoit Jacob
2008-05-12 17:34:46 +0000
f0eb3d2d3bupdated product test to carefully test all scalar types and fix an issue in the triangular test
Gael Guennebaud
2008-05-12 10:26:10 +0000
45cda6704a* Draft of a eigenvalues solver (does not support complex and does not re-use the QR decomposition)
Gael Guennebaud
2008-05-12 10:23:09 +0000
dca416cacemove arch-specific code to arch/SSE and arch/AltiVec subdirs. rename the noarch PacketMath.h to DummyPacketMath.h
Benoit Jacob
2008-05-12 08:30:42 +0000
3562b01105* Give Konstantinos a copyright line * Fix compilation of Inverse.h with vectorisation * Introduce EIGEN_GNUC_AT_LEAST(x,y) macro doing future-proof (e.g. gcc v5.0) check * Only use ProductWIP if vectorisation is enabled * rename EIGEN_ALWAYS_INLINE -> EIGEN_INLINE with fall-back to inline keyword * some cleanup/indentation
Benoit Jacob
2008-05-12 08:12:40 +0000
4f6d7abc87only include SSE3 headers if compiling with SSE3 support
Benoit Jacob
2008-05-08 09:15:16 +0000
4754fa4868removed "sort brief" in doxygen documentation
Gael Guennebaud
2008-05-08 08:13:38 +0000
bf5326c3ca* Added ReferencableBit flag to known if coeffRef is available. (needed by the new product implementation) * Make the packet* members template to support aligned and unaligned access. This makes Block vectorizable. Combined with ReferencableBit, we should be able to determine at runtime (in some specific cases) if an aligned vectorization is possible or not. * Improved the new product implementation to robustly handle all cases, it now passes all the tests. * Renamed the packet version ei_predux to ei_preduxp to avoid name collision.
Gael Guennebaud
2008-05-08 08:12:52 +0000
64c49de7ba* split PacketMath.h to SSE and Altivec specific files * improved the flexibility of the new product implementation, now all sizes seems to be properly handled.
Gael Guennebaud
2008-05-05 17:19:47 +0000
46fa4c713f* Started support for unaligned vectorization. * Introduce a new highly optimized matrix-matrix product for large matrices. The code is still highly experimental and it is activated only if you define EIGEN_WIP_PRODUCT at compile time. Currently the third dimension of the product must be a factor of the packet size (x4 for floats) and the right handed side matrix must be column major. Moreover, currently c = a*b; actually computes c += a*b !! Therefore, the code is provided for experimentation purpose only ! These limitations will be fixed soon or later to become the default product implementation.
Gael Guennebaud
2008-05-05 10:23:29 +0000
8c6007f80e* Patch by Konstantinos Margaritis: AltiVec vectorization. * Fix several warnings, temporarily disable determinant test.
Benoit Jacob
2008-05-03 12:21:23 +0000
0545df2149slighly improved the cache friendly product to use mul-add only
Gael Guennebaud
2008-05-03 10:01:30 +0000
a6655dd91aadded packet mul-add function (ei_pmad) and updated Product to use it. this change nothing for current SSE architecture but might be helpful for altivec/cell and up comming AMD processors.
Gael Guennebaud
2008-05-03 00:45:08 +0000
102e029dadRemoved ei_pload1, use posix_memalign to allocate aligned memory, and make Product ok when only one side is vectorizable (and the product is still vectorized)
Gael Guennebaud
2008-05-02 13:30:12 +0000
e19f9bc523added a test for triangular matrices
Gael Guennebaud
2008-05-02 11:35:59 +0000
890a8de962Make products always eval into expressions. Improves performance in benchmark. Still not as fasts as explicit eval(), strangely.
Benoit Jacob
2008-05-02 08:53:23 +0000
ef5b20bc50fix flag and cost computations for nested expressions
Gael Guennebaud
2008-05-01 18:58:30 +0000
5588def0cfnullary xpr are now vectorized
Gael Guennebaud
2008-05-01 14:28:53 +0000
02f1615d2aEnable vectorization of product with dynamic matrices, extended cache optimal product to work in any row/column major situations, and a few bugfixes (forgot to add the Cholesky header, vectorization of CwiseBinary)
Gael Guennebaud
2008-05-01 13:53:05 +0000
6486991ac3some cleaning in Cholesky and removed evil ei_sqrt of complex
Gael Guennebaud
2008-04-27 18:57:28 +0000
64bacf1c3f* added ei_sqrt for complex * updated Cholesky to support complex * correct result_type for abs and abs2 functors
Gael Guennebaud
2008-04-27 14:05:40 +0000
4ffffa670eadded Cholesky module
Gael Guennebaud
2008-04-27 10:57:32 +0000
1ec2d21ca5Fixed a couple of issues introduced in previous commits. Added a test for Triangular.
Gael Guennebaud
2008-04-26 20:28:27 +0000
b4c974d059Added triangular assignement, e.g.: m.upper() = a+b; only updates the upper triangular part of m. Note that: m = (a+b).upper(); updates all coefficients of m (but half of the additions will be skiped)
Gael Guennebaud
2008-04-26 19:20:26 +0000
4c92150676Added Triangular expression to extract upper or lower (strictly or not) part of a matrix. Triangular also provide an optimised method for forward and backward substitution. Further optimizations regarding assignments and products might come later.
Gael Guennebaud
2008-04-26 18:26:05 +0000
62bf0bbd59fix a bug in determinant of 4x4 matrices and a small type issue in Inverse
Gael Guennebaud
2008-04-26 08:56:52 +0000
173e582e3cadded a tough test to check the determinant that currently fails
Gael Guennebaud
2008-04-25 23:13:20 +0000
6f2c72fb53Various fixes in: - vector to vector assign - PartialRedux - Vectorization criteria of Product - returned type of normalized - SSE integer mul
Gael Guennebaud
2008-04-25 23:10:37 +0000
a451835bceMake the explicit vectorization much more flexible: - support dynamic sizes - support arbitrary matrix size when the matrix can be seen as a 1D array (except for fixed size matrices where the size in Bytes must be a factor of 16, this is to allow compact storage of a vector of matrices) Note that the explict vectorization is still experimental and far to be completely tested.
Gael Guennebaud
2008-04-25 15:46:18 +0000
30d47b5250forgot to add a file in the previous commit
Gael Guennebaud
2008-04-24 20:25:55 +0000
9385793f71Fix a couple of issue with the vectorization. In particular, default ei_p* functions are provided to handle not suported types seemlessly.
Gael Guennebaud
2008-04-24 18:35:39 +0000
6ae037dfb5give up on OpenMP... for now
Benoit Jacob
2008-04-18 07:57:46 +0000
acfd6f3bda- add _packetCoeff() to Inverse, allowing vectorization. - let Inverse take template parameter MatrixType instead of ExpressionType, in order to reduce executable code size when taking inverses of xpr's. - introduce ei_corrected_matrix_flags : the flags template parameter to the Matrix class is only a suggestion. This is also useful in ei_eval.
Benoit Jacob
2008-04-16 07:18:27 +0000
43e2bc14fe+5% optimization in 4x4 inverse: -only evaluate block expressions for which that is beneficial -don't check for invertibility unless requested
Benoit Jacob
2008-04-15 20:39:27 +0000
6747b45ae7for 4x4 matrices implement the special algorithm that Markos proposed, falling back to the general algorithm in the bad case.
Benoit Jacob
2008-04-15 20:15:36 +0000
2a86f052a5- optimized determinant calculations for small matrices (size <= 4) (only 30 muls for size 4) - rework the matrix inversion: now using cofactor technique for size<=3, so the ugly unrolling is only used for size 4 anymore, and even there I'm looking to get rid of it.
Benoit Jacob
2008-04-14 17:07:12 +0000
9789c04467when evaluating an xpr, the result can now be vectorizable even if the xpr itself wasn't vectorizable.
Benoit Jacob
2008-04-14 08:55:12 +0000
ea3ccb1e8c* Start of the LU module, with matrix inversion already there and fully optimized. * Even if LargeBit is set, only parallelize for large enough objects (controlled by EIGEN_PARALLELIZATION_TRESHOLD).
Benoit Jacob
2008-04-14 08:20:24 +0000
ab4046970b* Add fixed-size template versions of corner(), start(), end(). * Use them to write an unrolled path in echelon.cpp, as an experiment before I do this LU module. * For floating-point types, make ei_random() use an amplitude of 1.
Benoit Jacob
2008-04-12 17:37:27 +0000
dcebc46cdc- cleaner use of OpenMP (no code duplication anymore) using a macro and _Pragma. - use OpenMP also in cacheOptimalProduct and in the vectorized paths as well - kill the vector assignment unroller. implement in operator= the logic for assigning a row-vector in a col-vector. - CMakeLists support for building tests/examples with -fopenmp and/or -msse2 - updates in bench/, especially replace identity() by ones() which prevents underflows from perturbing bench results.
Benoit Jacob
2008-04-11 14:28:42 +0000
7bee90a62aMerge Gael's experimental OpenMP parallelization support into Assign.h.
Benoit Jacob
2008-04-11 08:18:47 +0000
187b1543ceadded a vectorized version of Product::_cacheOptimalProduct, added the possibility to disable the vectorization using EIGEN_DONT_VECTORIZE (some architectures has SSE support by default)
Gael Guennebaud
2008-04-10 12:34:22 +0000
613c49b475* add typedefs for matrices/vectors with LargeBit * add -pedantic to CXXFLAGS * cleanup intricated expressions with && and || which gave warnings because of "missing" parentheses * fix compile error in NumTraits, apparently discovered by -pedantic
Benoit Jacob
2008-04-10 10:33:50 +0000
ca448d2537split those files in util/ some more renaming
Benoit Jacob
2008-04-10 09:41:13 +0000
9d8876ce82* rename XprCopy -> Nested * rename OperatorEquals -> Assign * move Util.h and FwDecl.h to a util/ subdir
Benoit Jacob
2008-04-10 09:01:28 +0000
212da8ffe0fix priority operator bugs in the computation of the VectorizableBit flag, now benchmark.cpp is properly vectorized
Gael Guennebaud
2008-04-09 18:24:13 +0000
8f957564eca better bugfix in ei_matrix_operator_equals_packet_unroller
Gael Guennebaud
2008-04-09 18:04:26 +0000
d95d952e92bugfix in ei_matrix_operator_equals_packet_unroller
Gael Guennebaud
2008-04-09 17:44:59 +0000
1985fb0551Added initial experimental support for explicit vectorization. Currently only the following platform/operations are supported: - SSE2 compatible architecture - compiler compatible with intel's SSE2 intrinsics - float, double and int data types - fixed size matrices with a storage major dimension multiple of 4 (or 2 for double) - scalar-matrix product, component wise: +,-,*,min,max - matrix-matrix product only if the left matrix is vectorizable and column major or the right matrix is vectorizable and row major, e.g.: a.transpose() * b is not vectorized with the default column major storage. To use it you must define EIGEN_VECTORIZE and EIGEN_INTEL_PLATFORM.
Gael Guennebaud
2008-04-09 12:31:55 +0000
4920f2011efinish making use of CoeffReadCost and the new XprCopy everywhere seems appropriate to me.
Benoit Jacob
2008-04-08 14:15:01 +0000
371d302efb- merge ei_xpr_copy and ei_eval_if_needed_before_nesting - make use of CoeffReadCost to determine when to unroll the loops, for now only in Product.h and in OperatorEquals.h performance remains the same: generally still not as good as before the big changes.
Benoit Jacob
2008-04-06 18:01:03 +0000
30ec34de36fix compilation (finish removal of EIGEN_UNROLLED_LOOPS)
Benoit Jacob
2008-04-05 14:20:30 +0000
61e58cf602fixes as discussed with Gael on IRC. Mainly, in Fuzzy.h, and Dot.h, use ei_xpr_copy to evaluate args when needed. Had to introduce an ugly trick with ei_unref as when the XprCopy type is a reference one can't directly access member typedefs such as Scalar.
Benoit Jacob
2008-04-05 14:15:02 +0000
b4a156671f* make use of the EvalBeforeNestingBit and EvalBeforeAssigningBit in ei_xpr_copy and operator=, respectively. * added Matrix::lazyAssign() when EvalBeforeAssigningBit must be skipped (mainly internal use only) * all expressions are now stored by const reference * added Temporary xpr: .temporary() must be called on any temporary expression not directly returned by a function (mainly internal use only) * moved all functors in the Functors.h header * added some preliminaries stuff for the explicit vectorization
Gael Guennebaud
2008-04-05 11:10:54 +0000
048910caae* added cwise comparisons * added "all" and "any" special redux operators * added support bool matrices * added support for cost model of STL functors via ei_functor_traits (By default ei_functor_traits query the functor member Cost)
Gael Guennebaud
2008-04-03 18:13:27 +0000
Gael Guennebaud