Rasmus Munk Larsen 3ed67cb0bb Fix a bug in the implementation of Carmack's fast sqrt algorithm in Eigen (enabled by EIGEN_FAST_MATH), which causes the vectorized parts of the computation to return -0.0 instead of NaN for negative arguments. ...

Benchmark speed in Giga-sqrts/s
Intel(R) Xeon(R) CPU E5-1650 v3 @ 3.50GHz
-----------------------------------------
                    SSE        AVX
Fast=1              2.529G     4.380G
Fast=0              1.944G     1.898G
Fast=1 fixed        2.214G     3.739G

This table illustrates the worst case in terms speed impact: It was measured by repeatedly computing the sqrt of an n=4096 float vector that fits in L1 cache. For large vectors the operation becomes memory bound and the differences between the different versions almost negligible.

2016-10-04 14:22:56 -07:00

..

arch

Fix a bug in the implementation of Carmack's fast sqrt algorithm in Eigen (enabled by EIGEN_FAST_MATH), which causes the vectorized parts of the computation to return -0.0 instead of NaN for negative arguments.

2016-10-04 14:22:56 -07:00

functors

bug #1195: move NumTraits::Div<>::Cost to internal::scalar_div_cost (with some specializations in arch/SSE and arch/AVX)

2016-09-08 08:36:27 +02:00

products

Fix alignement of statically allocated temporaries in symv, and trmv.

2016-09-21 17:34:24 +02:00

util

Updated the list of warnings to reflect the new message ids introduced in cuda 8.0

2016-09-28 17:42:59 -07:00

Array.h

Doc: split customizing-eigen page into sub-pages and re-structure a bit the different topics

2016-08-30 11:10:08 +02:00

ArrayBase.h

bug #828: clarify documentation of SparseMatrixBase's unary methods.

2016-09-16 10:40:50 +02:00

ArrayWrapper.h

bug #178: remove additional const on nested expression, and remove several const_cast.

2016-01-28 21:43:20 +01:00

Assign_MKL.h

Fix support of Intel's VML

2016-07-06 14:07:32 +02:00

Assign.h

Remove deprecated code not used by evaluators

2014-09-18 15:15:27 +02:00

AssignEvaluator.h

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

BandMatrix.h

Fix numerous doxygen shortcomings, and workaround some clang -Wdocumentation warnings

2016-01-01 21:45:06 +01:00

Block.h

Fix shortcoming in fixed-value deduction of startRow/startCol

2016-02-29 10:31:27 +01:00

BooleanRedux.h

Simplify cost computations based on HugeCost being smaller that unrolling limit

2015-10-28 13:39:02 +01:00

CommaInitializer.h

bug #1272: Disable assertion when total number of columns is zero.

2016-08-12 15:15:34 +02:00

ConditionEstimator.h

Fix with expession template scalar types.

2016-07-26 11:33:28 +02:00

CoreEvaluators.h

bug #1317: fix performance regression with some Block expressions and clang by helping it to remove dead code.

2016-10-01 15:37:00 +02:00

CoreIterators.h

Cleaning pass on evaluators: remove the useless and error prone evaluator<>::type indirection.

2015-09-02 21:38:40 +02:00

CwiseBinaryOp.h

Relax mixing-type constraints for binary coefficient-wise operators:

2016-06-06 15:11:41 +02:00

CwiseNullaryOp.h

fix documentation of LinSpaced

2016-09-22 14:50:07 +02:00

CwiseTernaryOp.h

Moved static assertions into the class constructor to make the code more portable

2016-06-05 12:57:48 -07:00

CwiseUnaryOp.h

Fixed compilation errors triggered by duplicate inline declaration

2016-01-31 10:48:49 -08:00

CwiseUnaryView.h

bug #178: remove additional const on nested expression, and remove several const_cast.

2016-01-28 21:43:20 +01:00

DenseBase.h

bug #828: clarify documentation of SparseMatrixBase's methods returning a sub-matrix.

2016-09-16 11:23:19 +02:00

DenseCoeffsBase.h

Add static assertion to y(), z(), w() accessors

2016-01-20 09:18:44 +01:00

DenseStorage.h

bug #725: make move ctor/assignment noexcept.

2016-06-03 14:28:25 +02:00

Diagonal.h

bug #178: remove additional const on nested expression, and remove several const_cast.

2016-01-28 21:43:20 +01:00

DiagonalMatrix.h

Change the semantic of the last template parameter of Assignment from "Scalar" to "SFINAE" only.

2016-07-04 11:02:00 +02:00

DiagonalProduct.h

Remove deprecated code not used by evaluators

2014-09-18 15:15:27 +02:00

Dot.h

Relax mixing-type constraints for binary coefficient-wise operators:

2016-06-06 15:11:41 +02:00

EigenBase.h

Relax mixing-type constraints for binary coefficient-wise operators:

2016-06-06 15:11:41 +02:00

ForceAlignedAccess.h

Make cuda_basic test compile again by adding lots of EIGEN_DEVICE_FUNC.

2014-10-13 17:18:26 +02:00

Fuzzy.h

bug #701: workaround (min) and (max) blocking ADL by introducing numext::mini and numext::maxi internal functions and a EIGEN_NOT_A_MACRO macro.

2014-10-20 15:55:32 +02:00

GeneralProduct.h

Fix alignement of statically allocated temporaries in gemv.

2016-09-17 12:52:27 +02:00

GenericPacketMath.h

bug #1249: enable use of __builtin_prefetch for GCC, clang, and ICC only.

2016-07-25 15:17:45 +02:00

GlobalFunctions.h

Add missing Eigen::rsqrt global function

2016-07-20 11:59:49 +02:00

Inverse.h

Added several missing EIGEN_DEVICE_FUNC qualifiers

2016-09-14 14:06:21 -07:00

IO.h

Add NumTraits::digits10() function based on numeric_limits::digits10 and make use of it for printing matrices.

2016-07-25 11:13:01 +02:00

Map.h

Fix numerous doxygen shortcomings, and workaround some clang -Wdocumentation warnings

2016-01-01 21:45:06 +01:00

MapBase.h

Doc: split customizing-eigen page into sub-pages and re-structure a bit the different topics

2016-08-30 11:10:08 +02:00

MathFunctions.h

Added a specialization of Eigen::numext::real and Eigen::numext::imag for std::complex<T> to be used when compiling a cuda kernel. This is unfortunately necessary to be able to process complex numbers from a CUDA kernel on MacOS.

2016-09-22 15:52:23 -07:00

MathFunctionsImpl.h

Add a note regarding gcc bug #72867

2016-09-22 11:18:52 +02:00

Matrix.h

Doc: split customizing-eigen page into sub-pages and re-structure a bit the different topics

2016-08-30 11:10:08 +02:00

MatrixBase.h

Removed EIGEN_DEVICE_FUNC qualifers for the lu(), fullPivLu(), partialPivLu(), and inverse() functions since they aren't ready to run on GPU

2016-09-19 14:13:20 -07:00

NestByValue.h

Fix numerous doxygen shortcomings, and workaround some clang -Wdocumentation warnings

2016-01-01 21:45:06 +01:00

NoAlias.h

Relax mixing-type constraints for binary coefficient-wise operators:

2016-06-06 15:11:41 +02:00

NumTraits.h

bug #1195: move NumTraits::Div<>::Cost to internal::scalar_div_cost (with some specializations in arch/SSE and arch/AVX)

2016-09-08 08:36:27 +02:00

PermutationMatrix.h

Fix numerous doxygen shortcomings, and workaround some clang -Wdocumentation warnings

2016-01-01 21:45:06 +01:00

PlainObjectBase.h

Doc: split customizing-eigen page into sub-pages and re-structure a bit the different topics

2016-08-30 11:10:08 +02:00

Product.h

Cleanup useless helper: internal::product_result_scalar

2016-06-15 00:04:10 +02:00

ProductEvaluators.h

Use explicit type casting to generate packets of zeros.

2016-10-04 08:23:38 -07:00

Random.h

bug #1286: automatically detect the available prototypes of functors passed to CwiseNullaryExpr such that functors have only to implement the operators that matters among:

2016-08-31 15:45:25 +02:00

Redux.h

Big 279: enable mixing types for comparisons, min, and max.

2016-06-10 15:05:43 +02:00

Ref.h

Workaround MSVC instantiation faillure of has_*ary_operator at the level of triats<Ref>::match so that the has_*ary_operator are really properly instantiated throughout the compilation unit.

2016-09-06 15:47:04 +02:00

Replicate.h

Fix numerous doxygen shortcomings, and workaround some clang -Wdocumentation warnings

2016-01-01 21:45:06 +01:00

ReturnByValue.h

Fix numerous doxygen shortcomings, and workaround some clang -Wdocumentation warnings

2016-01-01 21:45:06 +01:00

Reverse.h

Fix numerous doxygen shortcomings, and workaround some clang -Wdocumentation warnings

2016-01-01 21:45:06 +01:00

Select.h

bug #697: make sure empty classes are at the end in case of multiple inheritence

2014-12-02 14:40:19 +01:00

SelfAdjointView.h

Implement scalar multiples and division by a scalar as a binary-expression with a constant expression.

2016-06-14 11:26:57 +02:00

SelfCwiseBinaryOp.h

Relax mixing-type constraints for binary coefficient-wise operators:

2016-06-06 15:11:41 +02:00

Solve.h

Change the semantic of the last template parameter of Assignment from "Scalar" to "SFINAE" only.

2016-07-04 11:02:00 +02:00

SolverBase.h

Add LU::transpose().solve() and LU::adjoint().solve() API.

2015-12-01 14:38:47 +01:00

SolveTriangular.h

Do not copy for trivial 1x1 case. This also avoids a "maybe-uninitialized" warning in some situations.

2016-05-11 23:50:13 +02:00

StableNorm.h

bug #1203: by-pass large stack-allocation in stableNorm if EIGEN_STACK_ALLOCATION_LIMIT is too small

2016-04-15 22:34:11 +02:00

Stride.h

Fix numerous doxygen shortcomings, and workaround some clang -Wdocumentation warnings

2016-01-01 21:45:06 +01:00

Swap.h

Allow to use arbitrary packet-types during evaluation.

2015-08-07 12:01:39 +02:00

Transpose.h

bug #178: remove additional const on nested expression, and remove several const_cast.

2016-01-28 21:43:20 +01:00

Transpositions.h

bug #178: remove additional const on nested expression, and remove several const_cast.

2016-01-28 21:43:20 +01:00

TriangularMatrix.h

Change the semantic of the last template parameter of Assignment from "Scalar" to "SFINAE" only.

2016-07-04 11:02:00 +02:00

VectorBlock.h

Fix numerous doxygen shortcomings, and workaround some clang -Wdocumentation warnings

2016-01-01 21:45:06 +01:00

VectorwiseOp.h

Add missing non-const reverse method in VectorwiseOp.

2016-07-16 15:19:28 +02:00

Visitor.h

Method is called visit, not visitor

2016-01-24 15:50:59 +01:00