(does not support complex and does not re-use the QR decomposition)
* Rewrite the cache friendly product to have only one instance per scalar type !
This significantly speeds up compilation time and reduces executable size.
The current drawback is that some trivial expressions might be
evaluated like conjugate or negate.
* Renamed "cache optimal" to "cache friendly"
* Added the ability to directly access matrix data of some expressions via:
- the stride()/_stride() methods
- DirectAccessBit flag (replace ReferencableBit)
* 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
(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.
* 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.
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)
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)
Updated back/forward substitution to better use Eigen's capability.
part of a matrix. Triangular also provide an optimised method for forward
and backward substitution. Further optimizations regarding assignments and
products might come later.
Updated determinant() to take into account triangular matrices.
Started the QR module with a QR decompostion algorithm.
Help needed to build a QR algorithm (eigen solver) based on it.
- 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.
are provided to handle not suported types seemlessly.
Added a generic null-ary expression with null-ary functors. They replace
Zero, Ones, Identity and Random.
- 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.
(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.
* 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.
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.
* 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
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.
- 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.
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.
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
* 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)
useless copies are made when evaluating nested expressions.
Changes:
- kill LazyBit, introduce EvalBeforeNestingBit and EvalBeforeAssigningBit
- product and random don't evaluate immediately anymore
- eval() always evaluates
- change the value of Dynamic to some large positive value,
in preparation of future simplifications
before the Product<> type is constructed. This resets template depth on each
intermediate evaluation, and gives simpler code. Introducing
ei_eval_if_expensive<Derived, n> which evaluates Derived if it's worth it
given that each of its coeffs will be accessed n times. Operator*
uses this with adequate values of n to evaluate args exactly when needed.
when to evaluate arguments and when to meta-unroll.
-use it in Product to determine when to eval args. not yet used
to determine when to unroll. for now, not used anywhere else but
that'll follow.
-fix badness of my last commit
to preserve SVN history). They are made useless by the new
ei_eval_unless_lazy.
- introduce a generic Eval member typedef so one can do e.g.
T t; U u; Product<T, U>::Eval m; m = t*u;
the cacheOptimal is only good for large enough matrices.
When taking a block in a fixed-size (hence small) matrix,
the SizeAtCompileTime is Dynamic hence that's not a good
indicator. This example shows that the good indicator is
MaxSizeAtCompileTime.
Result: +10% speed in echelon.cpp
* macro renaming: EIGEN_NDEBUG becomes EIGEN_NO_DEBUG
as this is much better (and similar to Qt) and
EIGEN_CUSTOM_ASSERT becomes EIGEN_USE_CUSTOM_ASSERT
* protect Core header by a EIGEN_CORE_H
to disable eigen's asserts without disabling one's own program's
asserts. Notice that Eigen code should now use ei_assert()
instead of assert().
* Remove findBiggestCoeff() as it's now almost redundant.
* Improve echelon.cpp: inner for loop replaced by xprs.
* remove useless "(*this)." here and there. I think they were
first introduced by automatic search&replace.
* fix compilation in Visitor.h (issue triggered by echelon.cpp)
* improve comment on swap().
* added cache efficient matrix-matrix product.
- provides a huge speed-up for large matrices.
- currently it is enabled when an explicit unrolling is not possible.
(the global funcs in MathFunctions.h and Fuzzy.h don't count as internal).
* Mainpage.dox. Add a few prospective Eigen users; change the recommended
-finline-limit from 10000 to 1000. The reason is: it could be harmful to have
a too big value here, couldn't it? (e.g. exceedingly large executables, cache
misses). Looking at gcc, a value of 900 would exactly mean "determine the inlining
of all functions as if they were marked with 'inline' keyword". So a value of
1000 seems a reasonable round number. In the benchmark that motivated this
(TestEigenSolvers) a value of 400 is enough on my system.
as well as partial redux (vertical or horizontal redux).
Includes shortcuts for: sum, minCoeff and maxCoeff.
There is no shortcut for the partial redux.
* Added a generic *visitor* mini framework. A visitor is a custom object
sequentially applied on each coefficient with knowledge of its value and
coordinates.
It is currentlly used to implement minCoeff(int*,int*) and maxCoeff(int*,int*).
findBiggestCoeff is now a shortcut for "this->cwiseAbs().maxCoeff(i,j)"
* Added coeff-wise min and max.
* fixed an issue with ei_pow(int,int) and gcc < 4.3 or ICC
- add cwiseExp(), cwiseLog()...
--> for example, doing a gamma-correction on a bitmap image stored as
an array of floats is a simple matter of:
Eigen::Map<VectorXf> m = VectorXf::map(bitmap,size);
m = m.cwisePow(gamma);
- apidoc improvements, reorganization of the \name's
- remove obsolete examples
- remove EIGEN_ALWAYS_INLINE on lazyProduct(), it seems useless.
internal classes: AaBb -> ei_aa_bb
IntAtRunTimeIfDynamic -> ei_int_if_dynamic
unify UNROLLING_LIMIT (there was no reason to have operator= use
a higher limit)
etc...
Finally the importing macro is named EIGEN_BASIC_PUBLIC_INTERFACE
because it does not only import the ei_traits, it also makes the base class
a friend, etc.
template parameter "Scalar" is removed. This is achieved by introducting a
template <typename Derived> struct Scalar to achieve a forward-declaration of
the Scalar typedefs.
If the number of coefficients does not match the matrix size, then an assertion is raised.
No support for xpr on the right side for the moment.
* Added support for assertion checking. This allows to test that an assertion is indeed raised
when it should be.
* Fixed a mistake in the CwiseUnary example.
- compatible with current STL's functors as well as with the extention proposal (TR1)
* thanks to the above, Cast and ScalarMultiple have been removed
* benchmark_suite is more flexible (compiler and matrix size)
* functor templates are not template template parameter anymore
(this allows to make templated functors !)
* Main page: extented compiler discussion
* A small hack to support gcc 3.4 and 4.0 (see the main page)
* Fix a cast type issue in Cast
* Various doxygen updates (mainly Cwise stuff and added doxygen groups
in MatrixBase to split the huge memeber list, still not perfect though)
* Updated Gael's email address
and long double.
-define scalar-multiple operators only for the current Scalar type;
thanks to Gael for expaining how to make the compiler understand
when automatic casting is needed.
-take ScalarMultiple take only 1 template param, again.
We lose some flexibility especially when dealing with complex numbers,
but we gain a lot of extensibility to new scalar types.
Rework the matrix storage to ensure optimal sizeof in all cases, while
keeping the decoupling of matrix sizes versus storage sizes.
Also fixing (recently introduced) bugs caused by unwanted
reallocations of the buffers.
- finally get the Eval stuff right. get back to having Eval as
a subclass of Matrix with limited functionality, and then,
add a typedef MatrixType to get the actual matrix type.
- add swap(), findBiggestCoeff()
- bugfix by Ramon in Transpose
- new demo: doc/echelon.cpp
- reduction of sizeof(Block) for vector types
- variants of block() and fixedBlock() for vector blocks
- convenience methods start() and end() for vectors
- convenience method corner() for matrices
dimension. The advantage is that evaluating a dynamic-sized block in a fixed-size
matrix no longer causes a dynamic memory allocation. Other new thing:
IntAtRunTimeIfDynamic allows storing an integer at zero cost if it is known at
compile time.
column-major order, even if storage is row-major. Benchmark showed that adapting
the traversal order to the storage order brought no benefit.
Also do some cleanup after Gael's big patch.
1) Eigen2 co-installable with Eigen1 without conflict, without affecting programs including either.
2) #include<Eigen/Core> without the .h without conflict with the Core/ directory
3) Uniformize coding style of the CMakeLists.
This is an optimization for complex matrices, allowing to do only a real multiplication
when a complex multiplication is not needed, e.g. in normalized().
with minimal code duplication. There now are only two (2)
const_cast remaining in the whole source code.
- eigen2 now fully allows copying a row-vector into a column-vector.
added a unit-test for that.
- split unit tests, improve docs, various improvements.
- make vectors use a separate loop unroller, so that copying a
row-vector into a col-vector is now possible
- add much more documentation
- misc improvements
