Commit Graph

432 Commits

Author SHA1 Message Date
Gael Guennebaud
a7a05382d1 Add a LU decomposition action in BTL and various cleaning in BTL. For instance
all per plot settings have been moved to a single file, go_mean now takes an
optional second argument "tiny" to generate plots for tiny matrices, and
output of comparison information wrt to previous benchs (if any).
2008-08-04 23:12:48 +00:00
Benoit Jacob
c2f8ecf466 * LU decomposition, supporting all rectangular matrices, with full
pivoting for better numerical stability. For now the only application is
determinant.
* New determinant unit-test.
* Disable most of Swap.h for now as it makes LU fail (mysterious).
Anyway Swap needs a big overhaul as proposed on IRC.
* Remnants of old class Inverse removed.
* Some warnings fixed.
2008-08-04 04:45:59 +00:00
Gael Guennebaud
f81dfcf00b fix two perf issues in product.
fix positive definite test in Cholesky.
remove #include <cstring> in CoreDeclaration.
2008-08-03 20:23:06 +00:00
Benoit Jacob
49ae3fca89 fix compile errors with gcc 4.3: unresolved func call to
ei_cache_friendly_product, and undeclared memcpy
2008-08-03 15:44:06 +00:00
Gael Guennebaud
6d11a07e5e Added a ei_palign function align a packet from two others.
This allows much faster code dealing with unligned as
in the updated matrix-vector product functions.
2008-08-03 15:15:46 +00:00
Gael Guennebaud
55aeb1f83a Optimizations:
* faster matrix-matrix and matrix-vector products (especially for not aligned cases)
 * faster tridiagonalization (make it using our matrix-vector impl.)
Others:
 * fix Flags of Map
 * split the test_product to two smaller ones
2008-08-01 23:44:59 +00:00
Gael Guennebaud
b32b186c14 removed the packet specializations of some functors
(GCC generates better code without those "optimizations")
2008-07-31 21:03:11 +00:00
Gael Guennebaud
842c4f8bfa Several compilation fixes for MSVC and NVCC, basically:
- added explicit enum to int conversion where needed
- if a function is not defined as declared and the return type is "tricky"
  then the type must be typedefined somewhere. A "tricky return type" can be:
  * a template class with a default parameter which depends on another template parameter
  * a nested template class, or type of a nested template class
2008-07-29 16:33:07 +00:00
Gael Guennebaud
e0215ee510 BTL: - added tridiagonalization and hessenberg decomposition bench
- added GOTO library
2008-07-28 20:48:21 +00:00
Gael Guennebaud
44d95e0540 fix some internal asserts in CacheFrinedlyProduct 2008-07-27 22:14:08 +00:00
Gael Guennebaud
02a7efa910 forgot to include this file in previous commit 2008-07-27 14:24:32 +00:00
Gael Guennebaud
93115619c2 * updated benchmark files according to recent renamings
* various improvements in BTL including trisolver and cholesky bench
2008-07-27 11:39:47 +00:00
Gael Guennebaud
e9e5261664 Fix a couple issues introduced in the previous commit:
* removed DirectAccessBit from Part
* use a template specialization in inverseProduct() to transform a Part xpr to a Flagged xpr
2008-07-26 23:05:44 +00:00
Gael Guennebaud
e77ccf2928 * Rewrite the triangular solver so that we can take advantage of our efficient matrix-vector products:
=> up to 6 times faster !
* Added DirectAccessBit to Part
* Added an exemple of a cwise operator
* Renamed perpendicular() => someOrthogonal() (geometry module)
* Fix a weired bug in ei_constant_functor: the default copy constructor did not copy
  the imaginary part when the single member of the class is a complex...
2008-07-26 20:40:29 +00:00
Gael Guennebaud
2940617e6f bugfix in some internal asserts of CacheFriendlyProduct 2008-07-26 12:26:27 +00:00
Benoit Jacob
f997a3e902 update the inverse test a little
make use of static asserts in Map
fix 2 warnings in CacheFriendlyProduct: unused var 'Vectorized'
2008-07-26 12:08:28 +00:00
Gael Guennebaud
b466c266a0 * Fix some complex alignment issues in the cache friendly matrix-vector products.
* Minor update of the cores of the Cholesky algorithms to make them more friendly
  wrt to matrix-vector products => speedup x5 !
2008-07-23 17:30:00 +00:00
Gael Guennebaud
172000aaeb Add .perpendicular() function in Geometry module (adapted from Eigen1)
Documentation:
 * add an overview for each module.
 * add an example for .all() and Cwise::operator<
2008-07-22 10:54:42 +00:00
Gael Guennebaud
516db2c3b9 Fix compilation issues with icc and g++ < 4.1. Those include:
- conflicts with operator * overloads
 - discard the use of ei_pdiv for interger
   (g++ handles operators on __m128* types, this is why it worked)
 - weird behavior of icc in fixed size Block() constructor complaining
   the initializer of m_blockRows and m_blockCols were missing while
   we are in fixed size (maybe this hide deeper problem since this is a
   recent one, but icc gives only little feedback)
2008-07-21 12:40:56 +00:00
Gael Guennebaud
c10f069b6b * Merge Extract and Part to the Part expression.
Renamed "MatrixBase::extract() const" to "MatrixBase::part() const"
* Renamed static functions identity, zero, ones, random with an upper case
  first letter: Identity, Zero, Ones and Random.
2008-07-21 00:34:46 +00:00
Gael Guennebaud
ce425d92f1 Various documentation improvements, in particualr in Cholesky and Geometry module.
Added doxygen groups for Matrix typedefs and the Geometry module
2008-07-20 15:18:54 +00:00
Gael Guennebaud
269f683902 Add cholesky's members to MatrixBase
Various documentation improvements including new snippets (AngleAxis and Cholesky)
2008-07-19 22:59:05 +00:00
Gael Guennebaud
6e2c53e056 Added an automatically generated list of selected examples in the documentation.
Added the custom gemetry_module tag, and use it.
2008-07-19 20:36:41 +00:00
Gael Guennebaud
05ad083467 Added MatrixBase::Unit*() static function to easily create unit/basis vectors.
Removed EulerAngles, addes typdefs for Quaternion and AngleAxis,
and added automatic conversions from Quaternion/AngleAxis to Matrix3 such that:
 Matrix3f m = AngleAxisf(0.2,Vector3f::UnitX) * AngleAxisf(0.2,Vector3f::UnitY);
just works.
2008-07-19 13:03:23 +00:00
Gael Guennebaud
7245c63067 Complete rewrite of partial reduction according to mailing list discussions. 2008-07-19 11:36:32 +00:00
Benoit Jacob
8b4945a5a2 add some static asserts, use them, fix gcc 4.3 warning in Product.h. 2008-07-19 00:25:41 +00:00
Gael Guennebaud
22a816ade8 * Fix a couple of issues related to the recent cache friendly products
* Improve the efficiency of matrix*vector in unaligned cases
* Trivial fixes in the destructors of MatrixStorage
* Removed the matrixNorm in test/product.cpp (twice faster and
  that assumed the matrix product was ok while checking that !!)
2008-07-19 00:09:01 +00:00
Benoit Jacob
62ec1dd616 * big rework of Inverse.h:
- remove all invertibility checking, will be redundant with LU
  - general case: adapt to matrix storage order for better perf
  - size 4 case: handle corner cases without falling back to gen case.
  - rationalize with selectors instead of compile time if
  - add C-style computeInverse()
* update inverse test.
* in snippets, default cout precision to 3 decimal places
* add some cmake module from kdelibs to support btl with cmake 2.4
2008-07-15 23:56:17 +00:00
Gael Guennebaud
b970a9c8aa trivial fix in EulerAngles constructor 2008-07-15 22:42:55 +00:00
Gael Guennebaud
c8cbc1665e enhancements of the plot generator:
- removed the ugly X11 and PNG gnuplots terminals
- use enhanced postscript terminal
- use imagemagick to generate the png files (with compression)
- disable the fortran impl by default since it is as meaningless as a "C impl"
- update line settings
2008-07-13 11:46:36 +00:00
Gael Guennebaud
99a625243f Optimization: added super efficient rowmajor * vector product (and vector * colmajor).
It basically performs 4 dot products at once reducing loads of the vector and improving
instructions scheduling. With 3 cache friendly algorithms, we now handle all product
configurations with outstanding perf for large matrices.
2008-07-13 01:22:54 +00:00
Benoit Jacob
51e6ee39f0 SVN_SILENT trivial fix 2008-07-12 23:42:19 +00:00
Gael Guennebaud
bd0183f850 fix a cmake issue in FindTvmet and FindMKL 2008-07-12 23:34:42 +00:00
Benoit Jacob
e979e6485f another occurence of that little cmake fix 2008-07-12 23:27:41 +00:00
Gael Guennebaud
861d18d553 * Optimization: added a specialization of Block for xpr with DirectAccessBit
* some simplifications and fixes in cache friendly products
2008-07-12 22:59:34 +00:00
Benoit Jacob
1bbaea9885 little cmake fix 2008-07-12 22:13:03 +00:00
Gael Guennebaud
10c4e36b39 disable MKL check and fortran for cmake <2.6 2008-07-12 21:54:02 +00:00
Gael Guennebaud
ed6e07b2f6 various improvements of the plot generator in BTL 2008-07-12 21:41:32 +00:00
Gael Guennebaud
8233de8b69 various minor updates in the benchmark suite like non inlining
of some functions as well as the experimental C code used to design
efficient eigen's matrix vector products.
2008-07-12 12:14:08 +00:00
Gael Guennebaud
b7bd1b3446 Add a *very efficient* evaluation path for both col-major matrix * vector
and vector * row-major products. Currently, it is enabled only is the matrix
has DirectAccessBit flag and the product is "large enough".
Added the respective unit tests in test/product/cpp.
2008-07-12 12:12:02 +00:00
Gael Guennebaud
6f71ef8277 resurrected tvmet, added mt4, intel's MKL and handcoded vectorized backends
in the benchmark suite
2008-07-10 18:28:50 +00:00
Benoit Jacob
2b53fd4d53 some performance fixes in Assign.h reported by Gael. Some doc update in
Cwise.
2008-07-10 16:15:55 +00:00
Gael Guennebaud
7b4c6b8862 in BTL: a specific bench/action can be selected at runtime, e.g.:
BTL_CONFIG="-a ata" ctest -V -R eigen
  run the all benchmarks having "ata" in their name for all
  libraries matching the regexp "eigen"
2008-07-09 22:35:11 +00:00
Gael Guennebaud
c9b046d5d5 * added optimized paths for matrix-vector and vector-matrix products
(using either a cache friendly strategy or re-using dot-product
  vectorized implementation)
* add LinearAccessBit to Transpose
2008-07-09 22:30:18 +00:00
Benoit Jacob
25904802bc raah, results were corrupted by overflow. Now slice vectorization is
about a +25% speedup which is still nice as i expected zero or even
negative benefit.
2008-07-09 16:46:26 +00:00
Benoit Jacob
8f21a5e862 add benchmark for slice vectorization... expected it to be little or
zero benefit... turns out to be 20x speedup. Something is wrong.
2008-07-09 16:43:11 +00:00
Gael Guennebaud
28539e7597 imported a reworked version of BTL (Benchmark for Templated Libraries).
the modifications to initial code follow:
* changed build system from plain makefiles to cmake
* added eigen2 (4 versions: vec/novec and fixed/dynamic), GMM++, MTL4 interfaces
* added "transposed matrix * vector" product action
* updated blitz interface to use condensed products instead of hand coded loops
* removed some deprecated interfaces
* changed default storage order to column major for all libraries
* new generic bench timer strategy which is supposed to be more accurate
* various code clean-up
2008-07-09 14:04:48 +00:00
Gael Guennebaud
5f55ab524c * added a lazyAssign overload skipping .lazy() such that c = (<xpr>).lazy() such that
lazyAssign overloads of <xpr> are automatically called (this also reduces assign instansiations)
2008-07-09 13:54:21 +00:00
Gael Guennebaud
783eb6da9b I forgot that the previous commit needed minor changes outside the bench folder 2008-07-08 17:25:58 +00:00
Gael Guennebaud
77a622f2bb add Cholesky and eigensolver benchmark 2008-07-08 17:20:17 +00:00