GitHub-Proxy/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2025-04-23 18:19:34 +08:00
Code Issues Packages Projects Releases Wiki Activity

Remove "using namespace Eigen" from blas/common.h.

Browse Source
This commit is contained in:
Antonio Sánchez 2024-02-22 22:51:42 +00:00 committed by Rasmus Munk Larsen
parent 6ed4d80cc8
commit 8a73c6490f
19 changed files with 503 additions and 335 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
blas/BandTriangularSolver.h
View File

@ -10,6 +10,7 @@
#ifndef EIGEN_BAND_TRIANGULARSOLVER_H #ifndef EIGEN_BAND_TRIANGULARSOLVER_H
#define EIGEN_BAND_TRIANGULARSOLVER_H #define EIGEN_BAND_TRIANGULARSOLVER_H
namespace Eigen {
namespace internal { namespace internal {
/* \internal /* \internal
@ -77,6 +78,7 @@ struct band_solve_triangular_selector<Index, Mode, LhsScalar, ConjLhs, RhsScalar
} }
}; };
} // end namespace internal } // namespace internal
} // namespace Eigen
#endif // EIGEN_BAND_TRIANGULARSOLVER_H #endif // EIGEN_BAND_TRIANGULARSOLVER_H

4
blas/GeneralRank1Update.h
View File

@ -10,6 +10,7 @@
#ifndef EIGEN_GENERAL_RANK1UPDATE_H #ifndef EIGEN_GENERAL_RANK1UPDATE_H
#define EIGEN_GENERAL_RANK1UPDATE_H #define EIGEN_GENERAL_RANK1UPDATE_H
namespace Eigen {
namespace internal { namespace internal {
/* Optimized matrix += alpha * uv' */ /* Optimized matrix += alpha * uv' */
@ -35,6 +36,7 @@ struct general_rank1_update<Scalar, Index, RowMajor, ConjLhs, ConjRhs> {
} }
}; };
} // end namespace internal } // namespace internal
} // namespace Eigen
#endif // EIGEN_GENERAL_RANK1UPDATE_H #endif // EIGEN_GENERAL_RANK1UPDATE_H

4
blas/PackedSelfadjointProduct.h
View File

@ -10,6 +10,7 @@
#ifndef EIGEN_SELFADJOINT_PACKED_PRODUCT_H #ifndef EIGEN_SELFADJOINT_PACKED_PRODUCT_H
#define EIGEN_SELFADJOINT_PACKED_PRODUCT_H #define EIGEN_SELFADJOINT_PACKED_PRODUCT_H
namespace Eigen {
namespace internal { namespace internal {
/* Optimized matrix += alpha * uv' /* Optimized matrix += alpha * uv'
@ -45,6 +46,7 @@ struct selfadjoint_packed_rank1_update<Scalar, Index, RowMajor, UpLo, ConjLhs, C
} }
}; };
} // end namespace internal } // namespace internal
} // namespace Eigen
#endif // EIGEN_SELFADJOINT_PACKED_PRODUCT_H #endif // EIGEN_SELFADJOINT_PACKED_PRODUCT_H

4
blas/PackedTriangularMatrixVector.h
View File

@ -10,6 +10,7 @@
#ifndef EIGEN_PACKED_TRIANGULAR_MATRIX_VECTOR_H #ifndef EIGEN_PACKED_TRIANGULAR_MATRIX_VECTOR_H
#define EIGEN_PACKED_TRIANGULAR_MATRIX_VECTOR_H #define EIGEN_PACKED_TRIANGULAR_MATRIX_VECTOR_H
namespace Eigen {
namespace internal { namespace internal {
template <typename Index, int Mode, typename LhsScalar, bool ConjLhs, typename RhsScalar, bool ConjRhs, template <typename Index, int Mode, typename LhsScalar, bool ConjLhs, typename RhsScalar, bool ConjRhs,
@ -75,6 +76,7 @@ struct packed_triangular_matrix_vector_product<Index, Mode, LhsScalar, ConjLhs,
}; };
}; };
} // end namespace internal } // namespace internal
} // namespace Eigen
#endif // EIGEN_PACKED_TRIANGULAR_MATRIX_VECTOR_H #endif // EIGEN_PACKED_TRIANGULAR_MATRIX_VECTOR_H

4
blas/PackedTriangularSolverVector.h
View File

@ -10,6 +10,7 @@
#ifndef EIGEN_PACKED_TRIANGULAR_SOLVER_VECTOR_H #ifndef EIGEN_PACKED_TRIANGULAR_SOLVER_VECTOR_H
#define EIGEN_PACKED_TRIANGULAR_SOLVER_VECTOR_H #define EIGEN_PACKED_TRIANGULAR_SOLVER_VECTOR_H
namespace Eigen {
namespace internal { namespace internal {
template <typename LhsScalar, typename RhsScalar, typename Index, int Side, int Mode, bool Conjugate, int StorageOrder> template <typename LhsScalar, typename RhsScalar, typename Index, int Side, int Mode, bool Conjugate, int StorageOrder>
@ -69,6 +70,7 @@ struct packed_triangular_solve_vector<LhsScalar, RhsScalar, Index, OnTheRight, M
} }
}; };
} // end namespace internal } // namespace internal
} // namespace Eigen
#endif // EIGEN_PACKED_TRIANGULAR_SOLVER_VECTOR_H #endif // EIGEN_PACKED_TRIANGULAR_SOLVER_VECTOR_H

4
blas/Rank2Update.h
View File

@ -10,6 +10,7 @@
#ifndef EIGEN_RANK2UPDATE_H #ifndef EIGEN_RANK2UPDATE_H
#define EIGEN_RANK2UPDATE_H #define EIGEN_RANK2UPDATE_H
namespace Eigen {
namespace internal { namespace internal {
/* Optimized selfadjoint matrix += alpha * uv' + conj(alpha)*vu' /* Optimized selfadjoint matrix += alpha * uv' + conj(alpha)*vu'
@ -49,6 +50,7 @@ struct packed_rank2_update_selector {
} }
}; };
} // end namespace internal } // namespace internal
} // namespace Eigen
#endif // EIGEN_RANK2UPDATE_H #endif // EIGEN_RANK2UPDATE_H

67
blas/common.h
View File

@ -29,6 +29,13 @@
#include "blas.h" #include "blas.h"
#include "BandTriangularSolver.h"
#include "GeneralRank1Update.h"
#include "PackedSelfadjointProduct.h"
#include "PackedTriangularMatrixVector.h"
#include "PackedTriangularSolverVector.h"
#include "Rank2Update.h"
#define NOTR 0 #define NOTR 0
#define TR 1 #define TR 1
#define ADJ 2 #define ADJ 2
@ -59,58 +66,58 @@ inline bool check_side(const char* side) { return SIDE(*side) != 0xff; }
inline bool check_uplo(const char* uplo) { return UPLO(*uplo) != 0xff; } inline bool check_uplo(const char* uplo) { return UPLO(*uplo) != 0xff; }
namespace Eigen {
#include "BandTriangularSolver.h"
#include "GeneralRank1Update.h"
#include "PackedSelfadjointProduct.h"
#include "PackedTriangularMatrixVector.h"
#include "PackedTriangularSolverVector.h"
#include "Rank2Update.h"
} // namespace Eigen
using namespace Eigen;
typedef SCALAR Scalar; typedef SCALAR Scalar;
typedef NumTraits<Scalar>::Real RealScalar; typedef Eigen::NumTraits<Scalar>::Real RealScalar;
typedef std::complex<RealScalar> Complex; typedef std::complex<RealScalar> Complex;
enum { IsComplex = Eigen::NumTraits<SCALAR>::IsComplex, Conj = IsComplex }; enum { IsComplex = Eigen::NumTraits<SCALAR>::IsComplex, Conj = IsComplex };
typedef Matrix<Scalar, Dynamic, Dynamic, ColMajor> PlainMatrixType; typedef Eigen::Matrix<Scalar, Eigen::Dynamic, Eigen::Dynamic, Eigen::ColMajor> PlainMatrixType;
typedef Map<Matrix<Scalar, Dynamic, Dynamic, ColMajor>, 0, OuterStride<> > MatrixType; typedef Eigen::Map<Eigen::Matrix<Scalar, Eigen::Dynamic, Eigen::Dynamic, Eigen::ColMajor>, 0, Eigen::OuterStride<> >
typedef Map<const Matrix<Scalar, Dynamic, Dynamic, ColMajor>, 0, OuterStride<> > ConstMatrixType; MatrixType;
typedef Map<Matrix<Scalar, Dynamic, 1>, 0, InnerStride<Dynamic> > StridedVectorType; typedef Eigen::Map<const Eigen::Matrix<Scalar, Eigen::Dynamic, Eigen::Dynamic, Eigen::ColMajor>, 0,
typedef Map<Matrix<Scalar, Dynamic, 1> > CompactVectorType; Eigen::OuterStride<> >
ConstMatrixType;
typedef Eigen::Map<Eigen::Matrix<Scalar, Eigen::Dynamic, 1>, 0, Eigen::InnerStride<Eigen::Dynamic> > StridedVectorType;
typedef Eigen::Map<Eigen::Matrix<Scalar, Eigen::Dynamic, 1> > CompactVectorType;
template <typename T> template <typename T>
Map<Matrix<T, Dynamic, Dynamic, ColMajor>, 0, OuterStride<> > matrix(T* data, int rows, int cols, int stride) { Eigen::Map<Eigen::Matrix<T, Eigen::Dynamic, Eigen::Dynamic, Eigen::ColMajor>, 0, Eigen::OuterStride<> > matrix(
return Map<Matrix<T, Dynamic, Dynamic, ColMajor>, 0, OuterStride<> >(data, rows, cols, OuterStride<>(stride)); T* data, int rows, int cols, int stride) {
return Eigen::Map<Eigen::Matrix<T, Eigen::Dynamic, Eigen::Dynamic, Eigen::ColMajor>, 0, Eigen::OuterStride<> >(
data, rows, cols, Eigen::OuterStride<>(stride));
} }
template <typename T> template <typename T>
Map<const Matrix<T, Dynamic, Dynamic, ColMajor>, 0, OuterStride<> > matrix(const T* data, int rows, int cols, Eigen::Map<const Eigen::Matrix<T, Eigen::Dynamic, Eigen::Dynamic, Eigen::ColMajor>, 0, Eigen::OuterStride<> > matrix(
int stride) { const T* data, int rows, int cols, int stride) {
return Map<const Matrix<T, Dynamic, Dynamic, ColMajor>, 0, OuterStride<> >(data, rows, cols, OuterStride<>(stride)); return Eigen::Map<const Eigen::Matrix<T, Eigen::Dynamic, Eigen::Dynamic, Eigen::ColMajor>, 0, Eigen::OuterStride<> >(
data, rows, cols, Eigen::OuterStride<>(stride));
} }
template <typename T> template <typename T>
Map<Matrix<T, Dynamic, 1>, 0, InnerStride<Dynamic> > make_vector(T* data, int size, int incr) { Eigen::Map<Eigen::Matrix<T, Eigen::Dynamic, 1>, 0, Eigen::InnerStride<Eigen::Dynamic> > make_vector(T* data, int size,
return Map<Matrix<T, Dynamic, 1>, 0, InnerStride<Dynamic> >(data, size, InnerStride<Dynamic>(incr)); int incr) {
return Eigen::Map<Eigen::Matrix<T, Eigen::Dynamic, 1>, 0, Eigen::InnerStride<Eigen::Dynamic> >(
data, size, Eigen::InnerStride<Eigen::Dynamic>(incr));
} }
template <typename T> template <typename T>
Map<const Matrix<T, Dynamic, 1>, 0, InnerStride<Dynamic> > make_vector(const T* data, int size, int incr) { Eigen::Map<const Eigen::Matrix<T, Eigen::Dynamic, 1>, 0, Eigen::InnerStride<Eigen::Dynamic> > make_vector(const T* data,
return Map<const Matrix<T, Dynamic, 1>, 0, InnerStride<Dynamic> >(data, size, InnerStride<Dynamic>(incr)); int size,
int incr) {
return Eigen::Map<const Eigen::Matrix<T, Eigen::Dynamic, 1>, 0, Eigen::InnerStride<Eigen::Dynamic> >(
data, size, Eigen::InnerStride<Eigen::Dynamic>(incr));
} }
template <typename T> template <typename T>
Map<Matrix<T, Dynamic, 1> > make_vector(T* data, int size) { Eigen::Map<Eigen::Matrix<T, Eigen::Dynamic, 1> > make_vector(T* data, int size) {
return Map<Matrix<T, Dynamic, 1> >(data, size); return Eigen::Map<Eigen::Matrix<T, Eigen::Dynamic, 1> >(data, size);
} }
template <typename T> template <typename T>
Map<const Matrix<T, Dynamic, 1> > make_vector(const T* data, int size) { Eigen::Map<const Eigen::Matrix<T, Eigen::Dynamic, 1> > make_vector(const T* data, int size) {
return Map<const Matrix<T, Dynamic, 1> >(data, size); return Eigen::Map<const Eigen::Matrix<T, Eigen::Dynamic, 1> >(data, size);
} }
template <typename T> template <typename T>

2
blas/double.cpp
View File

@ -19,7 +19,7 @@
#include "level2_real_impl.h" #include "level2_real_impl.h"
#include "level3_impl.h" #include "level3_impl.h"
double EIGEN_BLAS_FUNC_NAME(sdot)(int* n, float* x, int* incx, float* y, int* incy) { extern "C" double EIGEN_BLAS_FUNC_NAME(sdot)(int* n, float* x, int* incx, float* y, int* incy) {
if (*n <= 0) return 0; if (*n <= 0) return 0;
if (*incx == 1 && *incy == 1) if (*incx == 1 && *incy == 1)

16
blas/level1_cplx_impl.h
View File

@ -11,7 +11,7 @@
struct scalar_norm1_op { struct scalar_norm1_op {
typedef RealScalar result_type; typedef RealScalar result_type;
inline RealScalar operator()(const Scalar &a) const { return numext::norm1(a); } inline RealScalar operator()(const Scalar &a) const { return Eigen::numext::norm1(a); }
}; };
namespace Eigen { namespace Eigen {
namespace internal { namespace internal {
@ -40,7 +40,7 @@ extern "C" int EIGEN_CAT(i, EIGEN_BLAS_FUNC_NAME(amax))(int *n, RealScalar *px,
if (*n <= 0) return 0; if (*n <= 0) return 0;
Scalar *x = reinterpret_cast<Scalar *>(px); Scalar *x = reinterpret_cast<Scalar *>(px);
DenseIndex ret; Eigen::DenseIndex ret;
if (*incx == 1) if (*incx == 1)
make_vector(x, *n).unaryExpr<scalar_norm1_op>().maxCoeff(&ret); make_vector(x, *n).unaryExpr<scalar_norm1_op>().maxCoeff(&ret);
else else
@ -52,7 +52,7 @@ extern "C" int EIGEN_CAT(i, EIGEN_BLAS_FUNC_NAME(amin))(int *n, RealScalar *px,
if (*n <= 0) return 0; if (*n <= 0) return 0;
Scalar *x = reinterpret_cast<Scalar *>(px); Scalar *x = reinterpret_cast<Scalar *>(px);
DenseIndex ret; Eigen::DenseIndex ret;
if (*incx == 1) if (*incx == 1)
make_vector(x, *n).unaryExpr<scalar_norm1_op>().minCoeff(&ret); make_vector(x, *n).unaryExpr<scalar_norm1_op>().minCoeff(&ret);
else else
@ -132,16 +132,16 @@ EIGEN_BLAS_FUNC(EIGEN_CAT(REAL_SCALAR_SUFFIX, rot))
StridedVectorType vx(make_vector(x, *n, std::abs(*incx))); StridedVectorType vx(make_vector(x, *n, std::abs(*incx)));
StridedVectorType vy(make_vector(y, *n, std::abs(*incy))); StridedVectorType vy(make_vector(y, *n, std::abs(*incy)));
Reverse<StridedVectorType> rvx(vx); Eigen::Reverse<StridedVectorType> rvx(vx);
Reverse<StridedVectorType> rvy(vy); Eigen::Reverse<StridedVectorType> rvy(vy);
// TODO implement mixed real-scalar rotations // TODO implement mixed real-scalar rotations
if (*incx < 0 && *incy > 0) if (*incx < 0 && *incy > 0)
internal::apply_rotation_in_the_plane(rvx, vy, JacobiRotation<Scalar>(c, s)); Eigen::internal::apply_rotation_in_the_plane(rvx, vy, Eigen::JacobiRotation<Scalar>(c, s));
else if (*incx > 0 && *incy < 0) else if (*incx > 0 && *incy < 0)
internal::apply_rotation_in_the_plane(vx, rvy, JacobiRotation<Scalar>(c, s)); Eigen::internal::apply_rotation_in_the_plane(vx, rvy, Eigen::JacobiRotation<Scalar>(c, s));
else else
internal::apply_rotation_in_the_plane(vx, vy, JacobiRotation<Scalar>(c, s)); Eigen::internal::apply_rotation_in_the_plane(vx, vy, Eigen::JacobiRotation<Scalar>(c, s));
} }
EIGEN_BLAS_FUNC(EIGEN_CAT(REAL_SCALAR_SUFFIX, scal))(int *n, RealScalar *palpha, RealScalar *px, int *incx) { EIGEN_BLAS_FUNC(EIGEN_CAT(REAL_SCALAR_SUFFIX, scal))(int *n, RealScalar *palpha, RealScalar *px, int *incx) {

4
blas/level1_impl.h
View File

@ -88,10 +88,10 @@ EIGEN_BLAS_FUNC(rotg)(RealScalar *pa, RealScalar *pb, RealScalar *pc, RealScalar
a = b; a = b;
} else { } else {
scale = abs(a) + abs(b); scale = abs(a) + abs(b);
norm = scale * sqrt((numext::abs2(a / scale)) + (numext::abs2(b / scale))); norm = scale * sqrt((Eigen::numext::abs2(a / scale)) + (Eigen::numext::abs2(b / scale)));
alpha = a / abs(a); alpha = a / abs(a);
*c = abs(a) / norm; *c = abs(a) / norm;
*s = alpha * numext::conj(b) / norm; *s = alpha * Eigen::numext::conj(b) / norm;
a = alpha * norm; a = alpha * norm;
} }
#endif #endif

14
blas/level1_real_impl.h
View File

@ -28,7 +28,7 @@ extern "C" int EIGEN_CAT(i, EIGEN_BLAS_FUNC_NAME(amax))(int *n, Scalar *px, int
if (*n <= 0) return 0; if (*n <= 0) return 0;
Scalar *x = reinterpret_cast<Scalar *>(px); Scalar *x = reinterpret_cast<Scalar *>(px);
DenseIndex ret; Eigen::DenseIndex ret;
if (*incx == 1) if (*incx == 1)
make_vector(x, *n).cwiseAbs().maxCoeff(&ret); make_vector(x, *n).cwiseAbs().maxCoeff(&ret);
else else
@ -40,7 +40,7 @@ extern "C" int EIGEN_CAT(i, EIGEN_BLAS_FUNC_NAME(amin))(int *n, Scalar *px, int
if (*n <= 0) return 0; if (*n <= 0) return 0;
Scalar *x = reinterpret_cast<Scalar *>(px); Scalar *x = reinterpret_cast<Scalar *>(px);
DenseIndex ret; Eigen::DenseIndex ret;
if (*incx == 1) if (*incx == 1)
make_vector(x, *n).cwiseAbs().minCoeff(&ret); make_vector(x, *n).cwiseAbs().minCoeff(&ret);
else else
@ -97,15 +97,15 @@ EIGEN_BLAS_FUNC(rot)(int *n, Scalar *px, int *incx, Scalar *py, int *incy, Scala
StridedVectorType vx(make_vector(x, *n, std::abs(*incx))); StridedVectorType vx(make_vector(x, *n, std::abs(*incx)));
StridedVectorType vy(make_vector(y, *n, std::abs(*incy))); StridedVectorType vy(make_vector(y, *n, std::abs(*incy)));
Reverse<StridedVectorType> rvx(vx); Eigen::Reverse<StridedVectorType> rvx(vx);
Reverse<StridedVectorType> rvy(vy); Eigen::Reverse<StridedVectorType> rvy(vy);
if (*incx < 0 && *incy > 0) if (*incx < 0 && *incy > 0)
internal::apply_rotation_in_the_plane(rvx, vy, JacobiRotation<Scalar>(c, s)); Eigen::internal::apply_rotation_in_the_plane(rvx, vy, Eigen::JacobiRotation<Scalar>(c, s));
else if (*incx > 0 && *incy < 0) else if (*incx > 0 && *incy < 0)
internal::apply_rotation_in_the_plane(vx, rvy, JacobiRotation<Scalar>(c, s)); Eigen::internal::apply_rotation_in_the_plane(vx, rvy, Eigen::JacobiRotation<Scalar>(c, s));
else else
internal::apply_rotation_in_the_plane(vx, vy, JacobiRotation<Scalar>(c, s)); Eigen::internal::apply_rotation_in_the_plane(vx, vy, Eigen::JacobiRotation<Scalar>(c, s));
} }
/* /*

28
blas/level2_cplx_impl.h
View File

@ -22,9 +22,11 @@ EIGEN_BLAS_FUNC(hemv)
typedef void (*functype)(int, const Scalar *, int, const Scalar *, Scalar *, Scalar); typedef void (*functype)(int, const Scalar *, int, const Scalar *, Scalar *, Scalar);
static const functype func[2] = { static const functype func[2] = {
// array index: UP // array index: UP
(internal::selfadjoint_matrix_vector_product<Scalar, int, ColMajor, Upper, false, false>::run), (Eigen::internal::selfadjoint_matrix_vector_product<Scalar, int, Eigen::ColMajor, Eigen::Upper, false,
false>::run),
// array index: LO // array index: LO
(internal::selfadjoint_matrix_vector_product<Scalar, int, ColMajor, Lower, false, false>::run), (Eigen::internal::selfadjoint_matrix_vector_product<Scalar, int, Eigen::ColMajor, Eigen::Lower, false,
false>::run),
}; };
const Scalar *a = reinterpret_cast<const Scalar *>(pa); const Scalar *a = reinterpret_cast<const Scalar *>(pa);
@ -107,9 +109,9 @@ EIGEN_BLAS_FUNC(hpr)(char *uplo, int *n, RealScalar *palpha, RealScalar *px, int
typedef void (*functype)(int, Scalar *, const Scalar *, RealScalar); typedef void (*functype)(int, Scalar *, const Scalar *, RealScalar);
static const functype func[2] = { static const functype func[2] = {
// array index: UP // array index: UP
(internal::selfadjoint_packed_rank1_update<Scalar, int, ColMajor, Upper, false, Conj>::run), (Eigen::internal::selfadjoint_packed_rank1_update<Scalar, int, Eigen::ColMajor, Eigen::Upper, false, Conj>::run),
// array index: LO // array index: LO
(internal::selfadjoint_packed_rank1_update<Scalar, int, ColMajor, Lower, false, Conj>::run), (Eigen::internal::selfadjoint_packed_rank1_update<Scalar, int, Eigen::ColMajor, Eigen::Lower, false, Conj>::run),
}; };
Scalar *x = reinterpret_cast<Scalar *>(px); Scalar *x = reinterpret_cast<Scalar *>(px);
@ -149,9 +151,9 @@ EIGEN_BLAS_FUNC(hpr2)
typedef void (*functype)(int, Scalar *, const Scalar *, const Scalar *, Scalar); typedef void (*functype)(int, Scalar *, const Scalar *, const Scalar *, Scalar);
static const functype func[2] = { static const functype func[2] = {
// array index: UP // array index: UP
(internal::packed_rank2_update_selector<Scalar, int, Upper>::run), (Eigen::internal::packed_rank2_update_selector<Scalar, int, Eigen::Upper>::run),
// array index: LO // array index: LO
(internal::packed_rank2_update_selector<Scalar, int, Lower>::run), (Eigen::internal::packed_rank2_update_selector<Scalar, int, Eigen::Lower>::run),
}; };
Scalar *x = reinterpret_cast<Scalar *>(px); Scalar *x = reinterpret_cast<Scalar *>(px);
@ -195,9 +197,9 @@ EIGEN_BLAS_FUNC(her)(char *uplo, int *n, RealScalar *palpha, RealScalar *px, int
typedef void (*functype)(int, Scalar *, int, const Scalar *, const Scalar *, const Scalar &); typedef void (*functype)(int, Scalar *, int, const Scalar *, const Scalar *, const Scalar &);
static const functype func[2] = { static const functype func[2] = {
// array index: UP // array index: UP
(selfadjoint_rank1_update<Scalar, int, ColMajor, Upper, false, Conj>::run), (Eigen::selfadjoint_rank1_update<Scalar, int, Eigen::ColMajor, Eigen::Upper, false, Conj>::run),
// array index: LO // array index: LO
(selfadjoint_rank1_update<Scalar, int, ColMajor, Lower, false, Conj>::run), (Eigen::selfadjoint_rank1_update<Scalar, int, Eigen::ColMajor, Eigen::Lower, false, Conj>::run),
}; };
Scalar *x = reinterpret_cast<Scalar *>(px); Scalar *x = reinterpret_cast<Scalar *>(px);
@ -242,9 +244,9 @@ EIGEN_BLAS_FUNC(her2)
typedef void (*functype)(int, Scalar *, int, const Scalar *, const Scalar *, Scalar); typedef void (*functype)(int, Scalar *, int, const Scalar *, const Scalar *, Scalar);
static const functype func[2] = { static const functype func[2] = {
// array index: UP // array index: UP
(internal::rank2_update_selector<Scalar, int, Upper>::run), (Eigen::internal::rank2_update_selector<Scalar, int, Eigen::Upper>::run),
// array index: LO // array index: LO
(internal::rank2_update_selector<Scalar, int, Lower>::run), (Eigen::internal::rank2_update_selector<Scalar, int, Eigen::Lower>::run),
}; };
Scalar *x = reinterpret_cast<Scalar *>(px); Scalar *x = reinterpret_cast<Scalar *>(px);
@ -313,7 +315,8 @@ EIGEN_BLAS_FUNC(geru)
Scalar *x_cpy = get_compact_vector(x, *m, *incx); Scalar *x_cpy = get_compact_vector(x, *m, *incx);
Scalar *y_cpy = get_compact_vector(y, *n, *incy); Scalar *y_cpy = get_compact_vector(y, *n, *incy);
internal::general_rank1_update<Scalar, int, ColMajor, false, false>::run(*m, *n, a, *lda, x_cpy, y_cpy, alpha); Eigen::internal::general_rank1_update<Scalar, int, Eigen::ColMajor, false, false>::run(*m, *n, a, *lda, x_cpy, y_cpy,
alpha);
if (x_cpy != x) delete[] x_cpy; if (x_cpy != x) delete[] x_cpy;
if (y_cpy != y) delete[] y_cpy; if (y_cpy != y) delete[] y_cpy;
@ -351,7 +354,8 @@ EIGEN_BLAS_FUNC(gerc)
Scalar *x_cpy = get_compact_vector(x, *m, *incx); Scalar *x_cpy = get_compact_vector(x, *m, *incx);
Scalar *y_cpy = get_compact_vector(y, *n, *incy); Scalar *y_cpy = get_compact_vector(y, *n, *incy);
internal::general_rank1_update<Scalar, int, ColMajor, false, Conj>::run(*m, *n, a, *lda, x_cpy, y_cpy, alpha); Eigen::internal::general_rank1_update<Scalar, int, Eigen::ColMajor, false, Conj>::run(*m, *n, a, *lda, x_cpy, y_cpy,
alpha);
if (x_cpy != x) delete[] x_cpy; if (x_cpy != x) delete[] x_cpy;
if (y_cpy != y) delete[] y_cpy; if (y_cpy != y) delete[] y_cpy;

226
blas/level2_impl.h
View File

@ -13,12 +13,13 @@ template <typename Index, typename Scalar, int StorageOrder, bool ConjugateLhs,
struct general_matrix_vector_product_wrapper { struct general_matrix_vector_product_wrapper {
static void run(Index rows, Index cols, const Scalar *lhs, Index lhsStride, const Scalar *rhs, Index rhsIncr, static void run(Index rows, Index cols, const Scalar *lhs, Index lhsStride, const Scalar *rhs, Index rhsIncr,
Scalar *res, Index resIncr, Scalar alpha) { Scalar *res, Index resIncr, Scalar alpha) {
typedef internal::const_blas_data_mapper<Scalar, Index, StorageOrder> LhsMapper; typedef Eigen::internal::const_blas_data_mapper<Scalar, Index, StorageOrder> LhsMapper;
typedef internal::const_blas_data_mapper<Scalar, Index, RowMajor> RhsMapper; typedef Eigen::internal::const_blas_data_mapper<Scalar, Index, Eigen::RowMajor> RhsMapper;
internal::general_matrix_vector_product<Index, Scalar, LhsMapper, StorageOrder, ConjugateLhs, Scalar, RhsMapper, Eigen::internal::general_matrix_vector_product<Index, Scalar, LhsMapper, StorageOrder, ConjugateLhs, Scalar,
ConjugateRhs>::run(rows, cols, LhsMapper(lhs, lhsStride), RhsMapper, ConjugateRhs>::run(rows, cols, LhsMapper(lhs, lhsStride),
RhsMapper(rhs, rhsIncr), res, resIncr, alpha); RhsMapper(rhs, rhsIncr), res, resIncr,
alpha);
} }
}; };
@ -26,12 +27,13 @@ EIGEN_BLAS_FUNC(gemv)
(const char *opa, const int *m, const int *n, const RealScalar *palpha, const RealScalar *pa, const int *lda, (const char *opa, const int *m, const int *n, const RealScalar *palpha, const RealScalar *pa, const int *lda,
const RealScalar *pb, const int *incb, const RealScalar *pbeta, RealScalar *pc, const int *incc) { const RealScalar *pb, const int *incb, const RealScalar *pbeta, RealScalar *pc, const int *incc) {
typedef void (*functype)(int, int, const Scalar *, int, const Scalar *, int, Scalar *, int, Scalar); typedef void (*functype)(int, int, const Scalar *, int, const Scalar *, int, Scalar *, int, Scalar);
static const functype func[4] = {// array index: NOTR static const functype func[4] = {
(general_matrix_vector_product_wrapper<int, Scalar, ColMajor, false, false>::run), // array index: NOTR
// array index: TR (general_matrix_vector_product_wrapper<int, Scalar, Eigen::ColMajor, false, false>::run),
(general_matrix_vector_product_wrapper<int, Scalar, RowMajor, false, false>::run), // array index: TR
// array index: ADJ (general_matrix_vector_product_wrapper<int, Scalar, Eigen::RowMajor, false, false>::run),
(general_matrix_vector_product_wrapper<int, Scalar, RowMajor, Conj, false>::run), 0}; // array index: ADJ
(general_matrix_vector_product_wrapper<int, Scalar, Eigen::RowMajor, Conj, false>::run), 0};
const Scalar *a = reinterpret_cast<const Scalar *>(pa); const Scalar *a = reinterpret_cast<const Scalar *>(pa);
const Scalar *b = reinterpret_cast<const Scalar *>(pb); const Scalar *b = reinterpret_cast<const Scalar *>(pb);
@ -84,31 +86,43 @@ EIGEN_BLAS_FUNC(trsv)
(const char *uplo, const char *opa, const char *diag, const int *n, const RealScalar *pa, const int *lda, (const char *uplo, const char *opa, const char *diag, const int *n, const RealScalar *pa, const int *lda,
RealScalar *pb, const int *incb) { RealScalar *pb, const int *incb) {
typedef void (*functype)(int, const Scalar *, int, Scalar *); typedef void (*functype)(int, const Scalar *, int, Scalar *);
using Eigen::ColMajor;
using Eigen::Lower;
using Eigen::OnTheLeft;
using Eigen::RowMajor;
using Eigen::UnitDiag;
using Eigen::Upper;
static const functype func[16] = { static const functype func[16] = {
// array index: NOTR | (UP << 2) | (NUNIT << 3) // array index: NOTR | (UP << 2) | (NUNIT << 3)
(internal::triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Upper | 0, false, ColMajor>::run), (Eigen::internal::triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Upper | 0, false, ColMajor>::run),
// array index: TR | (UP << 2) | (NUNIT << 3) // array index: TR | (UP << 2) | (NUNIT << 3)
(internal::triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Lower | 0, false, RowMajor>::run), (Eigen::internal::triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Lower | 0, false, RowMajor>::run),
// array index: ADJ | (UP << 2) | (NUNIT << 3) // array index: ADJ | (UP << 2) | (NUNIT << 3)
(internal::triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Lower | 0, Conj, RowMajor>::run), 0, (Eigen::internal::triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Lower | 0, Conj, RowMajor>::run), 0,
// array index: NOTR | (LO << 2) | (NUNIT << 3) // array index: NOTR | (LO << 2) | (NUNIT << 3)
(internal::triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Lower | 0, false, ColMajor>::run), (Eigen::internal::triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Lower | 0, false, ColMajor>::run),
// array index: TR | (LO << 2) | (NUNIT << 3) // array index: TR | (LO << 2) | (NUNIT << 3)
(internal::triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Upper | 0, false, RowMajor>::run), (Eigen::internal::triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Upper | 0, false, RowMajor>::run),
// array index: ADJ | (LO << 2) | (NUNIT << 3) // array index: ADJ | (LO << 2) | (NUNIT << 3)
(internal::triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Upper | 0, Conj, RowMajor>::run), 0, (Eigen::internal::triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Upper | 0, Conj, RowMajor>::run), 0,
// array index: NOTR | (UP << 2) | (UNIT << 3) // array index: NOTR | (UP << 2) | (UNIT << 3)
(internal::triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Upper | UnitDiag, false, ColMajor>::run), (Eigen::internal::triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Upper | UnitDiag, false,
ColMajor>::run),
// array index: TR | (UP << 2) | (UNIT << 3) // array index: TR | (UP << 2) | (UNIT << 3)
(internal::triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Lower | UnitDiag, false, RowMajor>::run), (Eigen::internal::triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Lower | UnitDiag, false,
RowMajor>::run),
// array index: ADJ | (UP << 2) | (UNIT << 3) // array index: ADJ | (UP << 2) | (UNIT << 3)
(internal::triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Lower | UnitDiag, Conj, RowMajor>::run), 0, (Eigen::internal::triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Lower | UnitDiag, Conj, RowMajor>::run),
0,
// array index: NOTR | (LO << 2) | (UNIT << 3) // array index: NOTR | (LO << 2) | (UNIT << 3)
(internal::triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Lower | UnitDiag, false, ColMajor>::run), (Eigen::internal::triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Lower | UnitDiag, false,
ColMajor>::run),
// array index: TR | (LO << 2) | (UNIT << 3) // array index: TR | (LO << 2) | (UNIT << 3)
(internal::triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Upper | UnitDiag, false, RowMajor>::run), (Eigen::internal::triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Upper | UnitDiag, false,
RowMajor>::run),
// array index: ADJ | (LO << 2) | (UNIT << 3) // array index: ADJ | (LO << 2) | (UNIT << 3)
(internal::triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Upper | UnitDiag, Conj, RowMajor>::run), 0}; (Eigen::internal::triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Upper | UnitDiag, Conj, RowMajor>::run),
0};
const Scalar *a = reinterpret_cast<const Scalar *>(pa); const Scalar *a = reinterpret_cast<const Scalar *>(pa);
Scalar *b = reinterpret_cast<Scalar *>(pb); Scalar *b = reinterpret_cast<Scalar *>(pb);
@ -140,32 +154,46 @@ EIGEN_BLAS_FUNC(trmv)
(const char *uplo, const char *opa, const char *diag, const int *n, const RealScalar *pa, const int *lda, (const char *uplo, const char *opa, const char *diag, const int *n, const RealScalar *pa, const int *lda,
RealScalar *pb, const int *incb) { RealScalar *pb, const int *incb) {
typedef void (*functype)(int, int, const Scalar *, int, const Scalar *, int, Scalar *, int, const Scalar &); typedef void (*functype)(int, int, const Scalar *, int, const Scalar *, int, Scalar *, int, const Scalar &);
using Eigen::ColMajor;
using Eigen::Lower;
using Eigen::OnTheLeft;
using Eigen::RowMajor;
using Eigen::UnitDiag;
using Eigen::Upper;
static const functype func[16] = { static const functype func[16] = {
// array index: NOTR | (UP << 2) | (NUNIT << 3) // array index: NOTR | (UP << 2) | (NUNIT << 3)
(internal::triangular_matrix_vector_product<int, Upper | 0, Scalar, false, Scalar, false, ColMajor>::run), (Eigen::internal::triangular_matrix_vector_product<int, Upper | 0, Scalar, false, Scalar, false, ColMajor>::run),
// array index: TR | (UP << 2) | (NUNIT << 3) // array index: TR | (UP << 2) | (NUNIT << 3)
(internal::triangular_matrix_vector_product<int, Lower | 0, Scalar, false, Scalar, false, RowMajor>::run), (Eigen::internal::triangular_matrix_vector_product<int, Lower | 0, Scalar, false, Scalar, false, RowMajor>::run),
// array index: ADJ | (UP << 2) | (NUNIT << 3) // array index: ADJ | (UP << 2) | (NUNIT << 3)
(internal::triangular_matrix_vector_product<int, Lower | 0, Scalar, Conj, Scalar, false, RowMajor>::run), 0, (Eigen::internal::triangular_matrix_vector_product<int, Lower | 0, Scalar, Conj, Scalar, false, RowMajor>::run),
0,
// array index: NOTR | (LO << 2) | (NUNIT << 3) // array index: NOTR | (LO << 2) | (NUNIT << 3)
(internal::triangular_matrix_vector_product<int, Lower | 0, Scalar, false, Scalar, false, ColMajor>::run), (Eigen::internal::triangular_matrix_vector_product<int, Lower | 0, Scalar, false, Scalar, false, ColMajor>::run),
// array index: TR | (LO << 2) | (NUNIT << 3) // array index: TR | (LO << 2) | (NUNIT << 3)
(internal::triangular_matrix_vector_product<int, Upper | 0, Scalar, false, Scalar, false, RowMajor>::run), (Eigen::internal::triangular_matrix_vector_product<int, Upper | 0, Scalar, false, Scalar, false, RowMajor>::run),
// array index: ADJ | (LO << 2) | (NUNIT << 3) // array index: ADJ | (LO << 2) | (NUNIT << 3)
(internal::triangular_matrix_vector_product<int, Upper | 0, Scalar, Conj, Scalar, false, RowMajor>::run), 0, (Eigen::internal::triangular_matrix_vector_product<int, Upper | 0, Scalar, Conj, Scalar, false, RowMajor>::run),
0,
// array index: NOTR | (UP << 2) | (UNIT << 3) // array index: NOTR | (UP << 2) | (UNIT << 3)
(internal::triangular_matrix_vector_product<int, Upper | UnitDiag, Scalar, false, Scalar, false, ColMajor>::run), (Eigen::internal::triangular_matrix_vector_product<int, Upper | UnitDiag, Scalar, false, Scalar, false,
ColMajor>::run),
// array index: TR | (UP << 2) | (UNIT << 3) // array index: TR | (UP << 2) | (UNIT << 3)
(internal::triangular_matrix_vector_product<int, Lower | UnitDiag, Scalar, false, Scalar, false, RowMajor>::run), (Eigen::internal::triangular_matrix_vector_product<int, Lower | UnitDiag, Scalar, false, Scalar, false,
RowMajor>::run),
// array index: ADJ | (UP << 2) | (UNIT << 3) // array index: ADJ | (UP << 2) | (UNIT << 3)
(internal::triangular_matrix_vector_product<int, Lower | UnitDiag, Scalar, Conj, Scalar, false, RowMajor>::run), (Eigen::internal::triangular_matrix_vector_product<int, Lower | UnitDiag, Scalar, Conj, Scalar, false,
RowMajor>::run),
0, 0,
// array index: NOTR | (LO << 2) | (UNIT << 3) // array index: NOTR | (LO << 2) | (UNIT << 3)
(internal::triangular_matrix_vector_product<int, Lower | UnitDiag, Scalar, false, Scalar, false, ColMajor>::run), (Eigen::internal::triangular_matrix_vector_product<int, Lower | UnitDiag, Scalar, false, Scalar, false,
ColMajor>::run),
// array index: TR | (LO << 2) | (UNIT << 3) // array index: TR | (LO << 2) | (UNIT << 3)
(internal::triangular_matrix_vector_product<int, Upper | UnitDiag, Scalar, false, Scalar, false, RowMajor>::run), (Eigen::internal::triangular_matrix_vector_product<int, Upper | UnitDiag, Scalar, false, Scalar, false,
RowMajor>::run),
// array index: ADJ | (LO << 2) | (UNIT << 3) // array index: ADJ | (LO << 2) | (UNIT << 3)
(internal::triangular_matrix_vector_product<int, Upper | UnitDiag, Scalar, Conj, Scalar, false, RowMajor>::run), (Eigen::internal::triangular_matrix_vector_product<int, Upper | UnitDiag, Scalar, Conj, Scalar, false,
RowMajor>::run),
0}; 0};
const Scalar *a = reinterpret_cast<const Scalar *>(pa); const Scalar *a = reinterpret_cast<const Scalar *>(pa);
@ -189,7 +217,7 @@ EIGEN_BLAS_FUNC(trmv)
if (*n == 0) return; if (*n == 0) return;
Scalar *actual_b = get_compact_vector(b, *n, *incb); Scalar *actual_b = get_compact_vector(b, *n, *incb);
Matrix<Scalar, Dynamic, 1> res(*n); Eigen::Matrix<Scalar, Eigen::Dynamic, 1> res(*n);
res.setZero(); res.setZero();
int code = OP(*opa) | (UPLO(*uplo) << 2) | (DIAG(*diag) << 3); int code = OP(*opa) | (UPLO(*uplo) << 2) | (DIAG(*diag) << 3);
@ -345,34 +373,40 @@ EIGEN_BLAS_FUNC(tbmv)(char *uplo, char *opa, char *diag, int *n, int *k, RealSca
EIGEN_BLAS_FUNC(tbsv) EIGEN_BLAS_FUNC(tbsv)
(char *uplo, char *op, char *diag, int *n, int *k, RealScalar *pa, int *lda, RealScalar *px, int *incx) { (char *uplo, char *op, char *diag, int *n, int *k, RealScalar *pa, int *lda, RealScalar *px, int *incx) {
typedef void (*functype)(int, int, const Scalar *, int, Scalar *); typedef void (*functype)(int, int, const Scalar *, int, Scalar *);
using Eigen::ColMajor;
using Eigen::Lower;
using Eigen::OnTheLeft;
using Eigen::RowMajor;
using Eigen::UnitDiag;
using Eigen::Upper;
static const functype func[16] = { static const functype func[16] = {
// array index: NOTR | (UP << 2) | (NUNIT << 3) // array index: NOTR | (UP << 2) | (NUNIT << 3)
(internal::band_solve_triangular_selector<int, Upper | 0, Scalar, false, Scalar, ColMajor>::run), (Eigen::internal::band_solve_triangular_selector<int, Upper | 0, Scalar, false, Scalar, ColMajor>::run),
// array index: TR | (UP << 2) | (NUNIT << 3) // array index: TR | (UP << 2) | (NUNIT << 3)
(internal::band_solve_triangular_selector<int, Lower | 0, Scalar, false, Scalar, RowMajor>::run), (Eigen::internal::band_solve_triangular_selector<int, Lower | 0, Scalar, false, Scalar, RowMajor>::run),
// array index: ADJ | (UP << 2) | (NUNIT << 3) // array index: ADJ | (UP << 2) | (NUNIT << 3)
(internal::band_solve_triangular_selector<int, Lower | 0, Scalar, Conj, Scalar, RowMajor>::run), (Eigen::internal::band_solve_triangular_selector<int, Lower | 0, Scalar, Conj, Scalar, RowMajor>::run),
0, 0,
// array index: NOTR | (LO << 2) | (NUNIT << 3) // array index: NOTR | (LO << 2) | (NUNIT << 3)
(internal::band_solve_triangular_selector<int, Lower | 0, Scalar, false, Scalar, ColMajor>::run), (Eigen::internal::band_solve_triangular_selector<int, Lower | 0, Scalar, false, Scalar, ColMajor>::run),
// array index: TR | (LO << 2) | (NUNIT << 3) // array index: TR | (LO << 2) | (NUNIT << 3)
(internal::band_solve_triangular_selector<int, Upper | 0, Scalar, false, Scalar, RowMajor>::run), (Eigen::internal::band_solve_triangular_selector<int, Upper | 0, Scalar, false, Scalar, RowMajor>::run),
// array index: ADJ | (LO << 2) | (NUNIT << 3) // array index: ADJ | (LO << 2) | (NUNIT << 3)
(internal::band_solve_triangular_selector<int, Upper | 0, Scalar, Conj, Scalar, RowMajor>::run), (Eigen::internal::band_solve_triangular_selector<int, Upper | 0, Scalar, Conj, Scalar, RowMajor>::run),
0, 0,
// array index: NOTR | (UP << 2) | (UNIT << 3) // array index: NOTR | (UP << 2) | (UNIT << 3)
(internal::band_solve_triangular_selector<int, Upper | UnitDiag, Scalar, false, Scalar, ColMajor>::run), (Eigen::internal::band_solve_triangular_selector<int, Upper | UnitDiag, Scalar, false, Scalar, ColMajor>::run),
// array index: TR | (UP << 2) | (UNIT << 3) // array index: TR | (UP << 2) | (UNIT << 3)
(internal::band_solve_triangular_selector<int, Lower | UnitDiag, Scalar, false, Scalar, RowMajor>::run), (Eigen::internal::band_solve_triangular_selector<int, Lower | UnitDiag, Scalar, false, Scalar, RowMajor>::run),
// array index: ADJ | (UP << 2) | (UNIT << 3) // array index: ADJ | (UP << 2) | (UNIT << 3)
(internal::band_solve_triangular_selector<int, Lower | UnitDiag, Scalar, Conj, Scalar, RowMajor>::run), (Eigen::internal::band_solve_triangular_selector<int, Lower | UnitDiag, Scalar, Conj, Scalar, RowMajor>::run),
0, 0,
// array index: NOTR | (LO << 2) | (UNIT << 3) // array index: NOTR | (LO << 2) | (UNIT << 3)
(internal::band_solve_triangular_selector<int, Lower | UnitDiag, Scalar, false, Scalar, ColMajor>::run), (Eigen::internal::band_solve_triangular_selector<int, Lower | UnitDiag, Scalar, false, Scalar, ColMajor>::run),
// array index: TR | (LO << 2) | (UNIT << 3) // array index: TR | (LO << 2) | (UNIT << 3)
(internal::band_solve_triangular_selector<int, Upper | UnitDiag, Scalar, false, Scalar, RowMajor>::run), (Eigen::internal::band_solve_triangular_selector<int, Upper | UnitDiag, Scalar, false, Scalar, RowMajor>::run),
// array index: ADJ | (LO << 2) | (UNIT << 3) // array index: ADJ | (LO << 2) | (UNIT << 3)
(internal::band_solve_triangular_selector<int, Upper | UnitDiag, Scalar, Conj, Scalar, RowMajor>::run), (Eigen::internal::band_solve_triangular_selector<int, Upper | UnitDiag, Scalar, Conj, Scalar, RowMajor>::run),
0, 0,
}; };
@ -420,40 +454,52 @@ EIGEN_BLAS_FUNC(tbsv)
*/ */
EIGEN_BLAS_FUNC(tpmv)(char *uplo, char *opa, char *diag, int *n, RealScalar *pap, RealScalar *px, int *incx) { EIGEN_BLAS_FUNC(tpmv)(char *uplo, char *opa, char *diag, int *n, RealScalar *pap, RealScalar *px, int *incx) {
typedef void (*functype)(int, const Scalar *, const Scalar *, Scalar *, Scalar); typedef void (*functype)(int, const Scalar *, const Scalar *, Scalar *, Scalar);
using Eigen::ColMajor;
using Eigen::Lower;
using Eigen::OnTheLeft;
using Eigen::RowMajor;
using Eigen::UnitDiag;
using Eigen::Upper;
static const functype func[16] = { static const functype func[16] = {
// array index: NOTR | (UP << 2) | (NUNIT << 3) // array index: NOTR | (UP << 2) | (NUNIT << 3)
(internal::packed_triangular_matrix_vector_product<int, Upper | 0, Scalar, false, Scalar, false, ColMajor>::run), (Eigen::internal::packed_triangular_matrix_vector_product<int, Upper | 0, Scalar, false, Scalar, false,
ColMajor>::run),
// array index: TR | (UP << 2) | (NUNIT << 3) // array index: TR | (UP << 2) | (NUNIT << 3)
(internal::packed_triangular_matrix_vector_product<int, Lower | 0, Scalar, false, Scalar, false, RowMajor>::run), (Eigen::internal::packed_triangular_matrix_vector_product<int, Lower | 0, Scalar, false, Scalar, false,
RowMajor>::run),
// array index: ADJ | (UP << 2) | (NUNIT << 3) // array index: ADJ | (UP << 2) | (NUNIT << 3)
(internal::packed_triangular_matrix_vector_product<int, Lower | 0, Scalar, Conj, Scalar, false, RowMajor>::run), (Eigen::internal::packed_triangular_matrix_vector_product<int, Lower | 0, Scalar, Conj, Scalar, false,
RowMajor>::run),
0, 0,
// array index: NOTR | (LO << 2) | (NUNIT << 3) // array index: NOTR | (LO << 2) | (NUNIT << 3)
(internal::packed_triangular_matrix_vector_product<int, Lower | 0, Scalar, false, Scalar, false, ColMajor>::run), (Eigen::internal::packed_triangular_matrix_vector_product<int, Lower | 0, Scalar, false, Scalar, false,
ColMajor>::run),
// array index: TR | (LO << 2) | (NUNIT << 3) // array index: TR | (LO << 2) | (NUNIT << 3)
(internal::packed_triangular_matrix_vector_product<int, Upper | 0, Scalar, false, Scalar, false, RowMajor>::run), (Eigen::internal::packed_triangular_matrix_vector_product<int, Upper | 0, Scalar, false, Scalar, false,
RowMajor>::run),
// array index: ADJ | (LO << 2) | (NUNIT << 3) // array index: ADJ | (LO << 2) | (NUNIT << 3)
(internal::packed_triangular_matrix_vector_product<int, Upper | 0, Scalar, Conj, Scalar, false, RowMajor>::run), (Eigen::internal::packed_triangular_matrix_vector_product<int, Upper | 0, Scalar, Conj, Scalar, false,
RowMajor>::run),
0, 0,
// array index: NOTR | (UP << 2) | (UNIT << 3) // array index: NOTR | (UP << 2) | (UNIT << 3)
(internal::packed_triangular_matrix_vector_product<int, Upper | UnitDiag, Scalar, false, Scalar, false, (Eigen::internal::packed_triangular_matrix_vector_product<int, Upper | UnitDiag, Scalar, false, Scalar, false,
ColMajor>::run), ColMajor>::run),
// array index: TR | (UP << 2) | (UNIT << 3) // array index: TR | (UP << 2) | (UNIT << 3)
(internal::packed_triangular_matrix_vector_product<int, Lower | UnitDiag, Scalar, false, Scalar, false, (Eigen::internal::packed_triangular_matrix_vector_product<int, Lower | UnitDiag, Scalar, false, Scalar, false,
RowMajor>::run), RowMajor>::run),
// array index: ADJ | (UP << 2) | (UNIT << 3) // array index: ADJ | (UP << 2) | (UNIT << 3)
(internal::packed_triangular_matrix_vector_product<int, Lower | UnitDiag, Scalar, Conj, Scalar, false, (Eigen::internal::packed_triangular_matrix_vector_product<int, Lower | UnitDiag, Scalar, Conj, Scalar, false,
RowMajor>::run), RowMajor>::run),
0, 0,
// array index: NOTR | (LO << 2) | (UNIT << 3) // array index: NOTR | (LO << 2) | (UNIT << 3)
(internal::packed_triangular_matrix_vector_product<int, Lower | UnitDiag, Scalar, false, Scalar, false, (Eigen::internal::packed_triangular_matrix_vector_product<int, Lower | UnitDiag, Scalar, false, Scalar, false,
ColMajor>::run), ColMajor>::run),
// array index: TR | (LO << 2) | (UNIT << 3) // array index: TR | (LO << 2) | (UNIT << 3)
(internal::packed_triangular_matrix_vector_product<int, Upper | UnitDiag, Scalar, false, Scalar, false, (Eigen::internal::packed_triangular_matrix_vector_product<int, Upper | UnitDiag, Scalar, false, Scalar, false,
RowMajor>::run), RowMajor>::run),
// array index: ADJ | (LO << 2) | (UNIT << 3) // array index: ADJ | (LO << 2) | (UNIT << 3)
(internal::packed_triangular_matrix_vector_product<int, Upper | UnitDiag, Scalar, Conj, Scalar, false, (Eigen::internal::packed_triangular_matrix_vector_product<int, Upper | UnitDiag, Scalar, Conj, Scalar, false,
RowMajor>::run), RowMajor>::run),
0}; 0};
Scalar *ap = reinterpret_cast<Scalar *>(pap); Scalar *ap = reinterpret_cast<Scalar *>(pap);
@ -475,7 +521,7 @@ EIGEN_BLAS_FUNC(tpmv)(char *uplo, char *opa, char *diag, int *n, RealScalar *pap
if (*n == 0) return; if (*n == 0) return;
Scalar *actual_x = get_compact_vector(x, *n, *incx); Scalar *actual_x = get_compact_vector(x, *n, *incx);
Matrix<Scalar, Dynamic, 1> res(*n); Eigen::Matrix<Scalar, Eigen::Dynamic, 1> res(*n);
res.setZero(); res.setZero();
int code = OP(*opa) | (UPLO(*uplo) << 2) | (DIAG(*diag) << 3); int code = OP(*opa) | (UPLO(*uplo) << 2) | (DIAG(*diag) << 3);
@ -499,36 +545,50 @@ EIGEN_BLAS_FUNC(tpmv)(char *uplo, char *opa, char *diag, int *n, RealScalar *pap
*/ */
EIGEN_BLAS_FUNC(tpsv)(char *uplo, char *opa, char *diag, int *n, RealScalar *pap, RealScalar *px, int *incx) { EIGEN_BLAS_FUNC(tpsv)(char *uplo, char *opa, char *diag, int *n, RealScalar *pap, RealScalar *px, int *incx) {
typedef void (*functype)(int, const Scalar *, Scalar *); typedef void (*functype)(int, const Scalar *, Scalar *);
using Eigen::ColMajor;
using Eigen::Lower;
using Eigen::OnTheLeft;
using Eigen::RowMajor;
using Eigen::UnitDiag;
using Eigen::Upper;
static const functype func[16] = { static const functype func[16] = {
// array index: NOTR | (UP << 2) | (NUNIT << 3) // array index: NOTR | (UP << 2) | (NUNIT << 3)
(internal::packed_triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Upper | 0, false, ColMajor>::run), (Eigen::internal::packed_triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Upper | 0, false,
ColMajor>::run),
// array index: TR | (UP << 2) | (NUNIT << 3) // array index: TR | (UP << 2) | (NUNIT << 3)
(internal::packed_triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Lower | 0, false, RowMajor>::run), (Eigen::internal::packed_triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Lower | 0, false,
RowMajor>::run),
// array index: ADJ | (UP << 2) | (NUNIT << 3) // array index: ADJ | (UP << 2) | (NUNIT << 3)
(internal::packed_triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Lower | 0, Conj, RowMajor>::run), 0, (Eigen::internal::packed_triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Lower | 0, Conj, RowMajor>::run),
0,
// array index: NOTR | (LO << 2) | (NUNIT << 3) // array index: NOTR | (LO << 2) | (NUNIT << 3)
(internal::packed_triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Lower | 0, false, ColMajor>::run), (Eigen::internal::packed_triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Lower | 0, false,
ColMajor>::run),
// array index: TR | (LO << 2) | (NUNIT << 3) // array index: TR | (LO << 2) | (NUNIT << 3)
(internal::packed_triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Upper | 0, false, RowMajor>::run), (Eigen::internal::packed_triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Upper | 0, false,
RowMajor>::run),
// array index: ADJ | (LO << 2) | (NUNIT << 3) // array index: ADJ | (LO << 2) | (NUNIT << 3)
(internal::packed_triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Upper | 0, Conj, RowMajor>::run), 0, (Eigen::internal::packed_triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Upper | 0, Conj, RowMajor>::run),
0,
// array index: NOTR | (UP << 2) | (UNIT << 3) // array index: NOTR | (UP << 2) | (UNIT << 3)
(internal::packed_triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Upper | UnitDiag, false, (Eigen::internal::packed_triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Upper | UnitDiag, false,
ColMajor>::run), ColMajor>::run),
// array index: TR | (UP << 2) | (UNIT << 3) // array index: TR | (UP << 2) | (UNIT << 3)
(internal::packed_triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Lower | UnitDiag, false, (Eigen::internal::packed_triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Lower | UnitDiag, false,
RowMajor>::run), RowMajor>::run),
// array index: ADJ | (UP << 2) | (UNIT << 3) // array index: ADJ | (UP << 2) | (UNIT << 3)
(internal::packed_triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Lower | UnitDiag, Conj, RowMajor>::run), (Eigen::internal::packed_triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Lower | UnitDiag, Conj,
RowMajor>::run),
0, 0,
// array index: NOTR | (LO << 2) | (UNIT << 3) // array index: NOTR | (LO << 2) | (UNIT << 3)
(internal::packed_triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Lower | UnitDiag, false, (Eigen::internal::packed_triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Lower | UnitDiag, false,
ColMajor>::run), ColMajor>::run),
// array index: TR | (LO << 2) | (UNIT << 3) // array index: TR | (LO << 2) | (UNIT << 3)
(internal::packed_triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Upper | UnitDiag, false, (Eigen::internal::packed_triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Upper | UnitDiag, false,
RowMajor>::run), RowMajor>::run),
// array index: ADJ | (LO << 2) | (UNIT << 3) // array index: ADJ | (LO << 2) | (UNIT << 3)
(internal::packed_triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Upper | UnitDiag, Conj, RowMajor>::run), (Eigen::internal::packed_triangular_solve_vector<Scalar, Scalar, int, OnTheLeft, Upper | UnitDiag, Conj,
RowMajor>::run),
0}; 0};
Scalar *ap = reinterpret_cast<Scalar *>(pap); Scalar *ap = reinterpret_cast<Scalar *>(pap);

29
blas/level2_real_impl.h
View File

@ -14,11 +14,14 @@ EIGEN_BLAS_FUNC(symv)
(const char *uplo, const int *n, const RealScalar *palpha, const RealScalar *pa, const int *lda, const RealScalar *px, (const char *uplo, const int *n, const RealScalar *palpha, const RealScalar *pa, const int *lda, const RealScalar *px,
const int *incx, const RealScalar *pbeta, RealScalar *py, const int *incy) { const int *incx, const RealScalar *pbeta, RealScalar *py, const int *incy) {
typedef void (*functype)(int, const Scalar *, int, const Scalar *, Scalar *, Scalar); typedef void (*functype)(int, const Scalar *, int, const Scalar *, Scalar *, Scalar);
using Eigen::ColMajor;
using Eigen::Lower;
using Eigen::Upper;
static const functype func[2] = { static const functype func[2] = {
// array index: UP // array index: UP
(internal::selfadjoint_matrix_vector_product<Scalar, int, ColMajor, Upper, false, false>::run), (Eigen::internal::selfadjoint_matrix_vector_product<Scalar, int, ColMajor, Upper, false, false>::run),
// array index: LO // array index: LO
(internal::selfadjoint_matrix_vector_product<Scalar, int, ColMajor, Lower, false, false>::run), (Eigen::internal::selfadjoint_matrix_vector_product<Scalar, int, ColMajor, Lower, false, false>::run),
}; };
const Scalar *a = reinterpret_cast<const Scalar *>(pa); const Scalar *a = reinterpret_cast<const Scalar *>(pa);
@ -67,11 +70,14 @@ EIGEN_BLAS_FUNC(syr)
(const char *uplo, const int *n, const RealScalar *palpha, const RealScalar *px, const int *incx, RealScalar *pc, (const char *uplo, const int *n, const RealScalar *palpha, const RealScalar *px, const int *incx, RealScalar *pc,
const int *ldc) { const int *ldc) {
typedef void (*functype)(int, Scalar *, int, const Scalar *, const Scalar *, const Scalar &); typedef void (*functype)(int, Scalar *, int, const Scalar *, const Scalar *, const Scalar &);
using Eigen::ColMajor;
using Eigen::Lower;
using Eigen::Upper;
static const functype func[2] = { static const functype func[2] = {
// array index: UP // array index: UP
(selfadjoint_rank1_update<Scalar, int, ColMajor, Upper, false, Conj>::run), (Eigen::selfadjoint_rank1_update<Scalar, int, ColMajor, Upper, false, Conj>::run),
// array index: LO // array index: LO
(selfadjoint_rank1_update<Scalar, int, ColMajor, Lower, false, Conj>::run), (Eigen::selfadjoint_rank1_update<Scalar, int, ColMajor, Lower, false, Conj>::run),
}; };
const Scalar *x = reinterpret_cast<const Scalar *>(px); const Scalar *x = reinterpret_cast<const Scalar *>(px);
@ -109,9 +115,9 @@ EIGEN_BLAS_FUNC(syr2)
typedef void (*functype)(int, Scalar *, int, const Scalar *, const Scalar *, Scalar); typedef void (*functype)(int, Scalar *, int, const Scalar *, const Scalar *, Scalar);
static const functype func[2] = { static const functype func[2] = {
// array index: UP // array index: UP
(internal::rank2_update_selector<Scalar, int, Upper>::run), (Eigen::internal::rank2_update_selector<Scalar, int, Eigen::Upper>::run),
// array index: LO // array index: LO
(internal::rank2_update_selector<Scalar, int, Lower>::run), (Eigen::internal::rank2_update_selector<Scalar, int, Eigen::Lower>::run),
}; };
const Scalar *x = reinterpret_cast<const Scalar *>(px); const Scalar *x = reinterpret_cast<const Scalar *>(px);
@ -190,9 +196,9 @@ EIGEN_BLAS_FUNC(spr)(char *uplo, int *n, Scalar *palpha, Scalar *px, int *incx,
typedef void (*functype)(int, Scalar *, const Scalar *, Scalar); typedef void (*functype)(int, Scalar *, const Scalar *, Scalar);
static const functype func[2] = { static const functype func[2] = {
// array index: UP // array index: UP
(internal::selfadjoint_packed_rank1_update<Scalar, int, ColMajor, Upper, false, false>::run), (Eigen::internal::selfadjoint_packed_rank1_update<Scalar, int, Eigen::ColMajor, Eigen::Upper, false, false>::run),
// array index: LO // array index: LO
(internal::selfadjoint_packed_rank1_update<Scalar, int, ColMajor, Lower, false, false>::run), (Eigen::internal::selfadjoint_packed_rank1_update<Scalar, int, Eigen::ColMajor, Eigen::Lower, false, false>::run),
}; };
Scalar *x = reinterpret_cast<Scalar *>(px); Scalar *x = reinterpret_cast<Scalar *>(px);
@ -232,9 +238,9 @@ EIGEN_BLAS_FUNC(spr2)
typedef void (*functype)(int, Scalar *, const Scalar *, const Scalar *, Scalar); typedef void (*functype)(int, Scalar *, const Scalar *, const Scalar *, Scalar);
static const functype func[2] = { static const functype func[2] = {
// array index: UP // array index: UP
(internal::packed_rank2_update_selector<Scalar, int, Upper>::run), (Eigen::internal::packed_rank2_update_selector<Scalar, int, Eigen::Upper>::run),
// array index: LO // array index: LO
(internal::packed_rank2_update_selector<Scalar, int, Lower>::run), (Eigen::internal::packed_rank2_update_selector<Scalar, int, Eigen::Lower>::run),
}; };
Scalar *x = reinterpret_cast<Scalar *>(px); Scalar *x = reinterpret_cast<Scalar *>(px);
@ -299,7 +305,8 @@ EIGEN_BLAS_FUNC(ger)
Scalar *x_cpy = get_compact_vector(x, *m, *incx); Scalar *x_cpy = get_compact_vector(x, *m, *incx);
Scalar *y_cpy = get_compact_vector(y, *n, *incy); Scalar *y_cpy = get_compact_vector(y, *n, *incy);
internal::general_rank1_update<Scalar, int, ColMajor, false, false>::run(*m, *n, a, *lda, x_cpy, y_cpy, alpha); Eigen::internal::general_rank1_update<Scalar, int, Eigen::ColMajor, false, false>::run(*m, *n, a, *lda, x_cpy, y_cpy,
alpha);
if (x_cpy != x) delete[] x_cpy; if (x_cpy != x) delete[] x_cpy;
if (y_cpy != y) delete[] y_cpy; if (y_cpy != y) delete[] y_cpy;

391
blas/level3_impl.h
View File

@ -15,39 +15,43 @@ EIGEN_BLAS_FUNC(gemm)
const int *ldc) { const int *ldc) {
// std::cerr << "in gemm " << *opa << " " << *opb << " " << *m << " " << *n << " " << *k << " " << *lda << " " << // std::cerr << "in gemm " << *opa << " " << *opb << " " << *m << " " << *n << " " << *k << " " << *lda << " " <<
// *ldb << " " << *ldc << " " << *palpha << " " << *pbeta << "\n"; // *ldb << " " << *ldc << " " << *palpha << " " << *pbeta << "\n";
using Eigen::ColMajor;
using Eigen::DenseIndex;
using Eigen::Dynamic;
using Eigen::RowMajor;
typedef void (*functype)(DenseIndex, DenseIndex, DenseIndex, const Scalar *, DenseIndex, const Scalar *, DenseIndex, typedef void (*functype)(DenseIndex, DenseIndex, DenseIndex, const Scalar *, DenseIndex, const Scalar *, DenseIndex,
Scalar *, DenseIndex, DenseIndex, Scalar, Eigen::internal::level3_blocking<Scalar, Scalar> &, Scalar *, DenseIndex, DenseIndex, Scalar, Eigen::internal::level3_blocking<Scalar, Scalar> &,
Eigen::internal::GemmParallelInfo<DenseIndex> *); Eigen::internal::GemmParallelInfo<DenseIndex> *);
static const functype func[12] = { static const functype func[12] = {
// array index: NOTR | (NOTR << 2) // array index: NOTR | (NOTR << 2)
(internal::general_matrix_matrix_product<DenseIndex, Scalar, ColMajor, false, Scalar, ColMajor, false, ColMajor, (Eigen::internal::general_matrix_matrix_product<DenseIndex, Scalar, ColMajor, false, Scalar, ColMajor, false,
1>::run), ColMajor, 1>::run),
// array index: TR | (NOTR << 2) // array index: TR | (NOTR << 2)
(internal::general_matrix_matrix_product<DenseIndex, Scalar, RowMajor, false, Scalar, ColMajor, false, ColMajor, (Eigen::internal::general_matrix_matrix_product<DenseIndex, Scalar, RowMajor, false, Scalar, ColMajor, false,
1>::run), ColMajor, 1>::run),
// array index: ADJ | (NOTR << 2) // array index: ADJ | (NOTR << 2)
(internal::general_matrix_matrix_product<DenseIndex, Scalar, RowMajor, Conj, Scalar, ColMajor, false, ColMajor, (Eigen::internal::general_matrix_matrix_product<DenseIndex, Scalar, RowMajor, Conj, Scalar, ColMajor, false,
1>::run), ColMajor, 1>::run),
0, 0,
// array index: NOTR | (TR << 2) // array index: NOTR | (TR << 2)
(internal::general_matrix_matrix_product<DenseIndex, Scalar, ColMajor, false, Scalar, RowMajor, false, ColMajor, (Eigen::internal::general_matrix_matrix_product<DenseIndex, Scalar, ColMajor, false, Scalar, RowMajor, false,
1>::run), ColMajor, 1>::run),
// array index: TR | (TR << 2) // array index: TR | (TR << 2)
(internal::general_matrix_matrix_product<DenseIndex, Scalar, RowMajor, false, Scalar, RowMajor, false, ColMajor, (Eigen::internal::general_matrix_matrix_product<DenseIndex, Scalar, RowMajor, false, Scalar, RowMajor, false,
1>::run), ColMajor, 1>::run),
// array index: ADJ | (TR << 2) // array index: ADJ | (TR << 2)
(internal::general_matrix_matrix_product<DenseIndex, Scalar, RowMajor, Conj, Scalar, RowMajor, false, ColMajor, (Eigen::internal::general_matrix_matrix_product<DenseIndex, Scalar, RowMajor, Conj, Scalar, RowMajor, false,
1>::run), ColMajor, 1>::run),
0, 0,
// array index: NOTR | (ADJ << 2) // array index: NOTR | (ADJ << 2)
(internal::general_matrix_matrix_product<DenseIndex, Scalar, ColMajor, false, Scalar, RowMajor, Conj, ColMajor, (Eigen::internal::general_matrix_matrix_product<DenseIndex, Scalar, ColMajor, false, Scalar, RowMajor, Conj,
1>::run), ColMajor, 1>::run),
// array index: TR | (ADJ << 2) // array index: TR | (ADJ << 2)
(internal::general_matrix_matrix_product<DenseIndex, Scalar, RowMajor, false, Scalar, RowMajor, Conj, ColMajor, (Eigen::internal::general_matrix_matrix_product<DenseIndex, Scalar, RowMajor, false, Scalar, RowMajor, Conj,
1>::run), ColMajor, 1>::run),
// array index: ADJ | (ADJ << 2) // array index: ADJ | (ADJ << 2)
(internal::general_matrix_matrix_product<DenseIndex, Scalar, RowMajor, Conj, Scalar, RowMajor, Conj, ColMajor, (Eigen::internal::general_matrix_matrix_product<DenseIndex, Scalar, RowMajor, Conj, Scalar, RowMajor, Conj,
1>::run), ColMajor, 1>::run),
0}; 0};
const Scalar *a = reinterpret_cast<const Scalar *>(pa); const Scalar *a = reinterpret_cast<const Scalar *>(pa);
@ -86,7 +90,8 @@ EIGEN_BLAS_FUNC(gemm)
if (*k == 0) return; if (*k == 0) return;
internal::gemm_blocking_space<ColMajor, Scalar, Scalar, Dynamic, Dynamic, Dynamic> blocking(*m, *n, *k, 1, true); Eigen::internal::gemm_blocking_space<ColMajor, Scalar, Scalar, Dynamic, Dynamic, Dynamic> blocking(*m, *n, *k, 1,
true);
int code = OP(*opa) | (OP(*opb) << 2); int code = OP(*opa) | (OP(*opb) << 2);
func[code](*m, *n, *k, a, *lda, b, *ldb, c, 1, *ldc, alpha, blocking, 0); func[code](*m, *n, *k, a, *lda, b, *ldb, c, 1, *ldc, alpha, blocking, 0);
@ -97,76 +102,97 @@ EIGEN_BLAS_FUNC(trsm)
const RealScalar *palpha, const RealScalar *pa, const int *lda, RealScalar *pb, const int *ldb) { const RealScalar *palpha, const RealScalar *pa, const int *lda, RealScalar *pb, const int *ldb) {
// std::cerr << "in trsm " << *side << " " << *uplo << " " << *opa << " " << *diag << " " << *m << "," << *n << " " // std::cerr << "in trsm " << *side << " " << *uplo << " " << *opa << " " << *diag << " " << *m << "," << *n << " "
// << *palpha << " " << *lda << " " << *ldb<< "\n"; // << *palpha << " " << *lda << " " << *ldb<< "\n";
using Eigen::ColMajor;
using Eigen::DenseIndex;
using Eigen::Dynamic;
using Eigen::Lower;
using Eigen::OnTheLeft;
using Eigen::OnTheRight;
using Eigen::RowMajor;
using Eigen::UnitDiag;
using Eigen::Upper;
typedef void (*functype)(DenseIndex, DenseIndex, const Scalar *, DenseIndex, Scalar *, DenseIndex, DenseIndex, typedef void (*functype)(DenseIndex, DenseIndex, const Scalar *, DenseIndex, Scalar *, DenseIndex, DenseIndex,
Eigen::internal::level3_blocking<Scalar, Scalar> &); Eigen::internal::level3_blocking<Scalar, Scalar> &);
static const functype func[32] = { static const functype func[32] = {
// array index: NOTR | (LEFT << 2) | (UP << 3) | (NUNIT << 4) // array index: NOTR | (LEFT << 2) | (UP << 3) | (NUNIT << 4)
(internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheLeft, Upper | 0, false, ColMajor, ColMajor, 1>::run), (Eigen::internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheLeft, Upper | 0, false, ColMajor, ColMajor,
1>::run),
// array index: TR | (LEFT << 2) | (UP << 3) | (NUNIT << 4) // array index: TR | (LEFT << 2) | (UP << 3) | (NUNIT << 4)
(internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheLeft, Lower | 0, false, RowMajor, ColMajor, 1>::run), (Eigen::internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheLeft, Lower | 0, false, RowMajor, ColMajor,
1>::run),
// array index: ADJ | (LEFT << 2) | (UP << 3) | (NUNIT << 4) // array index: ADJ | (LEFT << 2) | (UP << 3) | (NUNIT << 4)
(internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheLeft, Lower | 0, Conj, RowMajor, ColMajor, 1>::run), (Eigen::internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheLeft, Lower | 0, Conj, RowMajor, ColMajor,
1>::run),
0, 0,
// array index: NOTR | (RIGHT << 2) | (UP << 3) | (NUNIT << 4) // array index: NOTR | (RIGHT << 2) | (UP << 3) | (NUNIT << 4)
(internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheRight, Upper | 0, false, ColMajor, ColMajor, 1>::run), (Eigen::internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheRight, Upper | 0, false, ColMajor, ColMajor,
1>::run),
// array index: TR | (RIGHT << 2) | (UP << 3) | (NUNIT << 4) // array index: TR | (RIGHT << 2) | (UP << 3) | (NUNIT << 4)
(internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheRight, Lower | 0, false, RowMajor, ColMajor, 1>::run), (Eigen::internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheRight, Lower | 0, false, RowMajor, ColMajor,
1>::run),
// array index: ADJ | (RIGHT << 2) | (UP << 3) | (NUNIT << 4) // array index: ADJ | (RIGHT << 2) | (UP << 3) | (NUNIT << 4)
(internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheRight, Lower | 0, Conj, RowMajor, ColMajor, 1>::run), (Eigen::internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheRight, Lower | 0, Conj, RowMajor, ColMajor,
1>::run),
0, 0,
// array index: NOTR | (LEFT << 2) | (LO << 3) | (NUNIT << 4) // array index: NOTR | (LEFT << 2) | (LO << 3) | (NUNIT << 4)
(internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheLeft, Lower | 0, false, ColMajor, ColMajor, 1>::run), (Eigen::internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheLeft, Lower | 0, false, ColMajor, ColMajor,
1>::run),
// array index: TR | (LEFT << 2) | (LO << 3) | (NUNIT << 4) // array index: TR | (LEFT << 2) | (LO << 3) | (NUNIT << 4)
(internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheLeft, Upper | 0, false, RowMajor, ColMajor, 1>::run), (Eigen::internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheLeft, Upper | 0, false, RowMajor, ColMajor,
1>::run),
// array index: ADJ | (LEFT << 2) | (LO << 3) | (NUNIT << 4) // array index: ADJ | (LEFT << 2) | (LO << 3) | (NUNIT << 4)
(internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheLeft, Upper | 0, Conj, RowMajor, ColMajor, 1>::run), (Eigen::internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheLeft, Upper | 0, Conj, RowMajor, ColMajor,
1>::run),
0, 0,
// array index: NOTR | (RIGHT << 2) | (LO << 3) | (NUNIT << 4) // array index: NOTR | (RIGHT << 2) | (LO << 3) | (NUNIT << 4)
(internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheRight, Lower | 0, false, ColMajor, ColMajor, 1>::run), (Eigen::internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheRight, Lower | 0, false, ColMajor, ColMajor,
1>::run),
// array index: TR | (RIGHT << 2) | (LO << 3) | (NUNIT << 4) // array index: TR | (RIGHT << 2) | (LO << 3) | (NUNIT << 4)
(internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheRight, Upper | 0, false, RowMajor, ColMajor, 1>::run), (Eigen::internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheRight, Upper | 0, false, RowMajor, ColMajor,
1>::run),
// array index: ADJ | (RIGHT << 2) | (LO << 3) | (NUNIT << 4) // array index: ADJ | (RIGHT << 2) | (LO << 3) | (NUNIT << 4)
(internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheRight, Upper | 0, Conj, RowMajor, ColMajor, 1>::run), (Eigen::internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheRight, Upper | 0, Conj, RowMajor, ColMajor,
1>::run),
0, 0,
// array index: NOTR | (LEFT << 2) | (UP << 3) | (UNIT << 4) // array index: NOTR | (LEFT << 2) | (UP << 3) | (UNIT << 4)
(internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheLeft, Upper | UnitDiag, false, ColMajor, ColMajor, (Eigen::internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheLeft, Upper | UnitDiag, false, ColMajor,
1>::run), ColMajor, 1>::run),
// array index: TR | (LEFT << 2) | (UP << 3) | (UNIT << 4) // array index: TR | (LEFT << 2) | (UP << 3) | (UNIT << 4)
(internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheLeft, Lower | UnitDiag, false, RowMajor, ColMajor, (Eigen::internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheLeft, Lower | UnitDiag, false, RowMajor,
1>::run), ColMajor, 1>::run),
// array index: ADJ | (LEFT << 2) | (UP << 3) | (UNIT << 4) // array index: ADJ | (LEFT << 2) | (UP << 3) | (UNIT << 4)
(internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheLeft, Lower | UnitDiag, Conj, RowMajor, ColMajor, (Eigen::internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheLeft, Lower | UnitDiag, Conj, RowMajor,
1>::run), ColMajor, 1>::run),
0, 0,
// array index: NOTR | (RIGHT << 2) | (UP << 3) | (UNIT << 4) // array index: NOTR | (RIGHT << 2) | (UP << 3) | (UNIT << 4)
(internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheRight, Upper | UnitDiag, false, ColMajor, ColMajor, (Eigen::internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheRight, Upper | UnitDiag, false, ColMajor,
1>::run), ColMajor, 1>::run),
// array index: TR | (RIGHT << 2) | (UP << 3) | (UNIT << 4) // array index: TR | (RIGHT << 2) | (UP << 3) | (UNIT << 4)
(internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheRight, Lower | UnitDiag, false, RowMajor, ColMajor, (Eigen::internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheRight, Lower | UnitDiag, false, RowMajor,
1>::run), ColMajor, 1>::run),
// array index: ADJ | (RIGHT << 2) | (UP << 3) | (UNIT << 4) // array index: ADJ | (RIGHT << 2) | (UP << 3) | (UNIT << 4)
(internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheRight, Lower | UnitDiag, Conj, RowMajor, ColMajor, (Eigen::internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheRight, Lower | UnitDiag, Conj, RowMajor,
1>::run), ColMajor, 1>::run),
0, 0,
// array index: NOTR | (LEFT << 2) | (LO << 3) | (UNIT << 4) // array index: NOTR | (LEFT << 2) | (LO << 3) | (UNIT << 4)
(internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheLeft, Lower | UnitDiag, false, ColMajor, ColMajor, (Eigen::internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheLeft, Lower | UnitDiag, false, ColMajor,
1>::run), ColMajor, 1>::run),
// array index: TR | (LEFT << 2) | (LO << 3) | (UNIT << 4) // array index: TR | (LEFT << 2) | (LO << 3) | (UNIT << 4)
(internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheLeft, Upper | UnitDiag, false, RowMajor, ColMajor, (Eigen::internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheLeft, Upper | UnitDiag, false, RowMajor,
1>::run), ColMajor, 1>::run),
// array index: ADJ | (LEFT << 2) | (LO << 3) | (UNIT << 4) // array index: ADJ | (LEFT << 2) | (LO << 3) | (UNIT << 4)
(internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheLeft, Upper | UnitDiag, Conj, RowMajor, ColMajor, (Eigen::internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheLeft, Upper | UnitDiag, Conj, RowMajor,
1>::run), ColMajor, 1>::run),
0, 0,
// array index: NOTR | (RIGHT << 2) | (LO << 3) | (UNIT << 4) // array index: NOTR | (RIGHT << 2) | (LO << 3) | (UNIT << 4)
(internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheRight, Lower | UnitDiag, false, ColMajor, ColMajor, (Eigen::internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheRight, Lower | UnitDiag, false, ColMajor,
1>::run), ColMajor, 1>::run),
// array index: TR | (RIGHT << 2) | (LO << 3) | (UNIT << 4) // array index: TR | (RIGHT << 2) | (LO << 3) | (UNIT << 4)
(internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheRight, Upper | UnitDiag, false, RowMajor, ColMajor, (Eigen::internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheRight, Upper | UnitDiag, false, RowMajor,
1>::run), ColMajor, 1>::run),
// array index: ADJ | (RIGHT << 2) | (LO << 3) | (UNIT << 4) // array index: ADJ | (RIGHT << 2) | (LO << 3) | (UNIT << 4)
(internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheRight, Upper | UnitDiag, Conj, RowMajor, ColMajor, (Eigen::internal::triangular_solve_matrix<Scalar, DenseIndex, OnTheRight, Upper | UnitDiag, Conj, RowMajor,
1>::run), ColMajor, 1>::run),
0}; 0};
const Scalar *a = reinterpret_cast<const Scalar *>(pa); const Scalar *a = reinterpret_cast<const Scalar *>(pa);
@ -197,12 +223,12 @@ EIGEN_BLAS_FUNC(trsm)
int code = OP(*opa) | (SIDE(*side) << 2) | (UPLO(*uplo) << 3) | (DIAG(*diag) << 4); int code = OP(*opa) | (SIDE(*side) << 2) | (UPLO(*uplo) << 3) | (DIAG(*diag) << 4);
if (SIDE(*side) == LEFT) { if (SIDE(*side) == LEFT) {
internal::gemm_blocking_space<ColMajor, Scalar, Scalar, Dynamic, Dynamic, Dynamic, 4> blocking(*m, *n, *m, 1, Eigen::internal::gemm_blocking_space<ColMajor, Scalar, Scalar, Dynamic, Dynamic, Dynamic, 4> blocking(*m, *n, *m, 1,
false); false);
func[code](*m, *n, a, *lda, b, 1, *ldb, blocking); func[code](*m, *n, a, *lda, b, 1, *ldb, blocking);
} else { } else {
internal::gemm_blocking_space<ColMajor, Scalar, Scalar, Dynamic, Dynamic, Dynamic, 4> blocking(*m, *n, *n, 1, Eigen::internal::gemm_blocking_space<ColMajor, Scalar, Scalar, Dynamic, Dynamic, Dynamic, 4> blocking(*m, *n, *n, 1,
false); false);
func[code](*n, *m, a, *lda, b, 1, *ldb, blocking); func[code](*n, *m, a, *lda, b, 1, *ldb, blocking);
} }
@ -216,89 +242,96 @@ EIGEN_BLAS_FUNC(trmm)
const RealScalar *palpha, const RealScalar *pa, const int *lda, RealScalar *pb, const int *ldb) { const RealScalar *palpha, const RealScalar *pa, const int *lda, RealScalar *pb, const int *ldb) {
// std::cerr << "in trmm " << *side << " " << *uplo << " " << *opa << " " << *diag << " " << *m << " " << *n << " " // std::cerr << "in trmm " << *side << " " << *uplo << " " << *opa << " " << *diag << " " << *m << " " << *n << " "
// << *lda << " " << *ldb << " " << *palpha << "\n"; // << *lda << " " << *ldb << " " << *palpha << "\n";
using Eigen::ColMajor;
using Eigen::DenseIndex;
using Eigen::Dynamic;
using Eigen::Lower;
using Eigen::RowMajor;
using Eigen::UnitDiag;
using Eigen::Upper;
typedef void (*functype)(DenseIndex, DenseIndex, DenseIndex, const Scalar *, DenseIndex, const Scalar *, DenseIndex, typedef void (*functype)(DenseIndex, DenseIndex, DenseIndex, const Scalar *, DenseIndex, const Scalar *, DenseIndex,
Scalar *, DenseIndex, DenseIndex, const Scalar &, Scalar *, DenseIndex, DenseIndex, const Scalar &,
internal::level3_blocking<Scalar, Scalar> &); Eigen::internal::level3_blocking<Scalar, Scalar> &);
static const functype func[32] = { static const functype func[32] = {
// array index: NOTR | (LEFT << 2) | (UP << 3) | (NUNIT << 4) // array index: NOTR | (LEFT << 2) | (UP << 3) | (NUNIT << 4)
(internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Upper | 0, true, ColMajor, false, ColMajor, false, (Eigen::internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Upper | 0, true, ColMajor, false, ColMajor,
ColMajor, 1>::run), false, ColMajor, 1>::run),
// array index: TR | (LEFT << 2) | (UP << 3) | (NUNIT << 4) // array index: TR | (LEFT << 2) | (UP << 3) | (NUNIT << 4)
(internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Lower | 0, true, RowMajor, false, ColMajor, false, (Eigen::internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Lower | 0, true, RowMajor, false, ColMajor,
ColMajor, 1>::run), false, ColMajor, 1>::run),
// array index: ADJ | (LEFT << 2) | (UP << 3) | (NUNIT << 4) // array index: ADJ | (LEFT << 2) | (UP << 3) | (NUNIT << 4)
(internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Lower | 0, true, RowMajor, Conj, ColMajor, false, (Eigen::internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Lower | 0, true, RowMajor, Conj, ColMajor,
ColMajor, 1>::run), false, ColMajor, 1>::run),
0, 0,
// array index: NOTR | (RIGHT << 2) | (UP << 3) | (NUNIT << 4) // array index: NOTR | (RIGHT << 2) | (UP << 3) | (NUNIT << 4)
(internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Upper | 0, false, ColMajor, false, ColMajor, (Eigen::internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Upper | 0, false, ColMajor, false,
false, ColMajor, 1>::run), ColMajor, false, ColMajor, 1>::run),
// array index: TR | (RIGHT << 2) | (UP << 3) | (NUNIT << 4) // array index: TR | (RIGHT << 2) | (UP << 3) | (NUNIT << 4)
(internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Lower | 0, false, ColMajor, false, RowMajor, (Eigen::internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Lower | 0, false, ColMajor, false,
false, ColMajor, 1>::run), RowMajor, false, ColMajor, 1>::run),
// array index: ADJ | (RIGHT << 2) | (UP << 3) | (NUNIT << 4) // array index: ADJ | (RIGHT << 2) | (UP << 3) | (NUNIT << 4)
(internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Lower | 0, false, ColMajor, false, RowMajor, Conj, (Eigen::internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Lower | 0, false, ColMajor, false,
ColMajor, 1>::run), RowMajor, Conj, ColMajor, 1>::run),
0, 0,
// array index: NOTR | (LEFT << 2) | (LO << 3) | (NUNIT << 4) // array index: NOTR | (LEFT << 2) | (LO << 3) | (NUNIT << 4)
(internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Lower | 0, true, ColMajor, false, ColMajor, false, (Eigen::internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Lower | 0, true, ColMajor, false, ColMajor,
ColMajor, 1>::run), false, ColMajor, 1>::run),
// array index: TR | (LEFT << 2) | (LO << 3) | (NUNIT << 4) // array index: TR | (LEFT << 2) | (LO << 3) | (NUNIT << 4)
(internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Upper | 0, true, RowMajor, false, ColMajor, false, (Eigen::internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Upper | 0, true, RowMajor, false, ColMajor,
ColMajor, 1>::run), false, ColMajor, 1>::run),
// array index: ADJ | (LEFT << 2) | (LO << 3) | (NUNIT << 4) // array index: ADJ | (LEFT << 2) | (LO << 3) | (NUNIT << 4)
(internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Upper | 0, true, RowMajor, Conj, ColMajor, false, (Eigen::internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Upper | 0, true, RowMajor, Conj, ColMajor,
ColMajor, 1>::run), false, ColMajor, 1>::run),
0, 0,
// array index: NOTR | (RIGHT << 2) | (LO << 3) | (NUNIT << 4) // array index: NOTR | (RIGHT << 2) | (LO << 3) | (NUNIT << 4)
(internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Lower | 0, false, ColMajor, false, ColMajor, (Eigen::internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Lower | 0, false, ColMajor, false,
false, ColMajor, 1>::run), ColMajor, false, ColMajor, 1>::run),
// array index: TR | (RIGHT << 2) | (LO << 3) | (NUNIT << 4) // array index: TR | (RIGHT << 2) | (LO << 3) | (NUNIT << 4)
(internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Upper | 0, false, ColMajor, false, RowMajor, (Eigen::internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Upper | 0, false, ColMajor, false,
false, ColMajor, 1>::run), RowMajor, false, ColMajor, 1>::run),
// array index: ADJ | (RIGHT << 2) | (LO << 3) | (NUNIT << 4) // array index: ADJ | (RIGHT << 2) | (LO << 3) | (NUNIT << 4)
(internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Upper | 0, false, ColMajor, false, RowMajor, Conj, (Eigen::internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Upper | 0, false, ColMajor, false,
ColMajor, 1>::run), RowMajor, Conj, ColMajor, 1>::run),
0, 0,
// array index: NOTR | (LEFT << 2) | (UP << 3) | (UNIT << 4) // array index: NOTR | (LEFT << 2) | (UP << 3) | (UNIT << 4)
(internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Upper | UnitDiag, true, ColMajor, false, ColMajor, (Eigen::internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Upper | UnitDiag, true, ColMajor, false,
false, ColMajor, 1>::run), ColMajor, false, ColMajor, 1>::run),
// array index: TR | (LEFT << 2) | (UP << 3) | (UNIT << 4) // array index: TR | (LEFT << 2) | (UP << 3) | (UNIT << 4)
(internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Lower | UnitDiag, true, RowMajor, false, ColMajor, (Eigen::internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Lower | UnitDiag, true, RowMajor, false,
false, ColMajor, 1>::run), ColMajor, false, ColMajor, 1>::run),
// array index: ADJ | (LEFT << 2) | (UP << 3) | (UNIT << 4) // array index: ADJ | (LEFT << 2) | (UP << 3) | (UNIT << 4)
(internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Lower | UnitDiag, true, RowMajor, Conj, ColMajor, (Eigen::internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Lower | UnitDiag, true, RowMajor, Conj,
false, ColMajor, 1>::run), ColMajor, false, ColMajor, 1>::run),
0, 0,
// array index: NOTR | (RIGHT << 2) | (UP << 3) | (UNIT << 4) // array index: NOTR | (RIGHT << 2) | (UP << 3) | (UNIT << 4)
(internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Upper | UnitDiag, false, ColMajor, false, (Eigen::internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Upper | UnitDiag, false, ColMajor, false,
ColMajor, false, ColMajor, 1>::run), ColMajor, false, ColMajor, 1>::run),
// array index: TR | (RIGHT << 2) | (UP << 3) | (UNIT << 4) // array index: TR | (RIGHT << 2) | (UP << 3) | (UNIT << 4)
(internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Lower | UnitDiag, false, ColMajor, false, (Eigen::internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Lower | UnitDiag, false, ColMajor, false,
RowMajor, false, ColMajor, 1>::run), RowMajor, false, ColMajor, 1>::run),
// array index: ADJ | (RIGHT << 2) | (UP << 3) | (UNIT << 4) // array index: ADJ | (RIGHT << 2) | (UP << 3) | (UNIT << 4)
(internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Lower | UnitDiag, false, ColMajor, false, (Eigen::internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Lower | UnitDiag, false, ColMajor, false,
RowMajor, Conj, ColMajor, 1>::run), RowMajor, Conj, ColMajor, 1>::run),
0, 0,
// array index: NOTR | (LEFT << 2) | (LO << 3) | (UNIT << 4) // array index: NOTR | (LEFT << 2) | (LO << 3) | (UNIT << 4)
(internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Lower | UnitDiag, true, ColMajor, false, ColMajor, (Eigen::internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Lower | UnitDiag, true, ColMajor, false,
false, ColMajor, 1>::run), ColMajor, false, ColMajor, 1>::run),
// array index: TR | (LEFT << 2) | (LO << 3) | (UNIT << 4) // array index: TR | (LEFT << 2) | (LO << 3) | (UNIT << 4)
(internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Upper | UnitDiag, true, RowMajor, false, ColMajor, (Eigen::internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Upper | UnitDiag, true, RowMajor, false,
false, ColMajor, 1>::run), ColMajor, false, ColMajor, 1>::run),
// array index: ADJ | (LEFT << 2) | (LO << 3) | (UNIT << 4) // array index: ADJ | (LEFT << 2) | (LO << 3) | (UNIT << 4)
(internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Upper | UnitDiag, true, RowMajor, Conj, ColMajor, (Eigen::internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Upper | UnitDiag, true, RowMajor, Conj,
false, ColMajor, 1>::run), ColMajor, false, ColMajor, 1>::run),
0, 0,
// array index: NOTR | (RIGHT << 2) | (LO << 3) | (UNIT << 4) // array index: NOTR | (RIGHT << 2) | (LO << 3) | (UNIT << 4)
(internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Lower | UnitDiag, false, ColMajor, false, (Eigen::internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Lower | UnitDiag, false, ColMajor, false,
ColMajor, false, ColMajor, 1>::run), ColMajor, false, ColMajor, 1>::run),
// array index: TR | (RIGHT << 2) | (LO << 3) | (UNIT << 4) // array index: TR | (RIGHT << 2) | (LO << 3) | (UNIT << 4)
(internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Upper | UnitDiag, false, ColMajor, false, (Eigen::internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Upper | UnitDiag, false, ColMajor, false,
RowMajor, false, ColMajor, 1>::run), RowMajor, false, ColMajor, 1>::run),
// array index: ADJ | (RIGHT << 2) | (LO << 3) | (UNIT << 4) // array index: ADJ | (RIGHT << 2) | (LO << 3) | (UNIT << 4)
(internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Upper | UnitDiag, false, ColMajor, false, (Eigen::internal::product_triangular_matrix_matrix<Scalar, DenseIndex, Upper | UnitDiag, false, ColMajor, false,
RowMajor, Conj, ColMajor, 1>::run), RowMajor, Conj, ColMajor, 1>::run),
0}; 0};
const Scalar *a = reinterpret_cast<const Scalar *>(pa); const Scalar *a = reinterpret_cast<const Scalar *>(pa);
@ -329,16 +362,16 @@ EIGEN_BLAS_FUNC(trmm)
if (*m == 0 || *n == 0) return; if (*m == 0 || *n == 0) return;
// FIXME find a way to avoid this copy // FIXME find a way to avoid this copy
Matrix<Scalar, Dynamic, Dynamic, ColMajor> tmp = matrix(b, *m, *n, *ldb); Eigen::Matrix<Scalar, Dynamic, Dynamic, ColMajor> tmp = matrix(b, *m, *n, *ldb);
matrix(b, *m, *n, *ldb).setZero(); matrix(b, *m, *n, *ldb).setZero();
if (SIDE(*side) == LEFT) { if (SIDE(*side) == LEFT) {
internal::gemm_blocking_space<ColMajor, Scalar, Scalar, Dynamic, Dynamic, Dynamic, 4> blocking(*m, *n, *m, 1, Eigen::internal::gemm_blocking_space<ColMajor, Scalar, Scalar, Dynamic, Dynamic, Dynamic, 4> blocking(*m, *n, *m, 1,
false); false);
func[code](*m, *n, *m, a, *lda, tmp.data(), tmp.outerStride(), b, 1, *ldb, alpha, blocking); func[code](*m, *n, *m, a, *lda, tmp.data(), tmp.outerStride(), b, 1, *ldb, alpha, blocking);
} else { } else {
internal::gemm_blocking_space<ColMajor, Scalar, Scalar, Dynamic, Dynamic, Dynamic, 4> blocking(*m, *n, *n, 1, Eigen::internal::gemm_blocking_space<ColMajor, Scalar, Scalar, Dynamic, Dynamic, Dynamic, 4> blocking(*m, *n, *n, 1,
false); false);
func[code](*m, *n, *n, tmp.data(), tmp.outerStride(), a, *lda, b, 1, *ldb, alpha, blocking); func[code](*m, *n, *n, tmp.data(), tmp.outerStride(), a, *lda, b, 1, *ldb, alpha, blocking);
} }
} }
@ -383,9 +416,15 @@ EIGEN_BLAS_FUNC(symm)
if (*m == 0 || *n == 0) return; if (*m == 0 || *n == 0) return;
int size = (SIDE(*side) == LEFT) ? (*m) : (*n); int size = (SIDE(*side) == LEFT) ? (*m) : (*n);
using Eigen::ColMajor;
using Eigen::DenseIndex;
using Eigen::Dynamic;
using Eigen::Lower;
using Eigen::RowMajor;
using Eigen::Upper;
#if ISCOMPLEX #if ISCOMPLEX
// FIXME add support for symmetric complex matrix // FIXME add support for symmetric complex matrix
Matrix<Scalar, Dynamic, Dynamic, ColMajor> matA(size, size); Eigen::Matrix<Scalar, Dynamic, Dynamic, ColMajor> matA(size, size);
if (UPLO(*uplo) == UP) { if (UPLO(*uplo) == UP) {
matA.triangularView<Upper>() = matrix(a, size, size, *lda); matA.triangularView<Upper>() = matrix(a, size, size, *lda);
matA.triangularView<Lower>() = matrix(a, size, size, *lda).transpose(); matA.triangularView<Lower>() = matrix(a, size, size, *lda).transpose();
@ -398,24 +437,29 @@ EIGEN_BLAS_FUNC(symm)
else if (SIDE(*side) == RIGHT) else if (SIDE(*side) == RIGHT)
matrix(c, *m, *n, *ldc) += alpha * matrix(b, *m, *n, *ldb) * matA; matrix(c, *m, *n, *ldc) += alpha * matrix(b, *m, *n, *ldb) * matA;
#else #else
internal::gemm_blocking_space<ColMajor, Scalar, Scalar, Dynamic, Dynamic, Dynamic> blocking(*m, *n, size, 1, false); Eigen::internal::gemm_blocking_space<ColMajor, Scalar, Scalar, Dynamic, Dynamic, Dynamic> blocking(*m, *n, size, 1,
false);
if (SIDE(*side) == LEFT) if (SIDE(*side) == LEFT)
if (UPLO(*uplo) == UP) if (UPLO(*uplo) == UP)
internal::product_selfadjoint_matrix<Scalar, DenseIndex, RowMajor, true, false, ColMajor, false, false, ColMajor, Eigen::internal::product_selfadjoint_matrix<Scalar, DenseIndex, RowMajor, true, false, ColMajor, false, false,
1>::run(*m, *n, a, *lda, b, *ldb, c, 1, *ldc, alpha, blocking); ColMajor, 1>::run(*m, *n, a, *lda, b, *ldb, c, 1, *ldc, alpha,
blocking);
else if (UPLO(*uplo) == LO) else if (UPLO(*uplo) == LO)
internal::product_selfadjoint_matrix<Scalar, DenseIndex, ColMajor, true, false, ColMajor, false, false, ColMajor, Eigen::internal::product_selfadjoint_matrix<Scalar, DenseIndex, ColMajor, true, false, ColMajor, false, false,
1>::run(*m, *n, a, *lda, b, *ldb, c, 1, *ldc, alpha, blocking); ColMajor, 1>::run(*m, *n, a, *lda, b, *ldb, c, 1, *ldc, alpha,
blocking);
else else
return; return;
else if (SIDE(*side) == RIGHT) else if (SIDE(*side) == RIGHT)
if (UPLO(*uplo) == UP) if (UPLO(*uplo) == UP)
internal::product_selfadjoint_matrix<Scalar, DenseIndex, ColMajor, false, false, RowMajor, true, false, ColMajor, Eigen::internal::product_selfadjoint_matrix<Scalar, DenseIndex, ColMajor, false, false, RowMajor, true, false,
1>::run(*m, *n, b, *ldb, a, *lda, c, 1, *ldc, alpha, blocking); ColMajor, 1>::run(*m, *n, b, *ldb, a, *lda, c, 1, *ldc, alpha,
blocking);
else if (UPLO(*uplo) == LO) else if (UPLO(*uplo) == LO)
internal::product_selfadjoint_matrix<Scalar, DenseIndex, ColMajor, false, false, ColMajor, true, false, ColMajor, Eigen::internal::product_selfadjoint_matrix<Scalar, DenseIndex, ColMajor, false, false, ColMajor, true, false,
1>::run(*m, *n, b, *ldb, a, *lda, c, 1, *ldc, alpha, blocking); ColMajor, 1>::run(*m, *n, b, *ldb, a, *lda, c, 1, *ldc, alpha,
blocking);
else else
return; return;
else else
@ -430,29 +474,35 @@ EIGEN_BLAS_FUNC(syrk)
const int *lda, const RealScalar *pbeta, RealScalar *pc, const int *ldc) { const int *lda, const RealScalar *pbeta, RealScalar *pc, const int *ldc) {
// std::cerr << "in syrk " << *uplo << " " << *op << " " << *n << " " << *k << " " << *palpha << " " << *lda << " " // std::cerr << "in syrk " << *uplo << " " << *op << " " << *n << " " << *k << " " << *palpha << " " << *lda << " "
// << *pbeta << " " << *ldc << "\n"; // << *pbeta << " " << *ldc << "\n";
using Eigen::ColMajor;
using Eigen::DenseIndex;
using Eigen::Dynamic;
using Eigen::Lower;
using Eigen::RowMajor;
using Eigen::Upper;
#if !ISCOMPLEX #if !ISCOMPLEX
typedef void (*functype)(DenseIndex, DenseIndex, const Scalar *, DenseIndex, const Scalar *, DenseIndex, Scalar *, typedef void (*functype)(DenseIndex, DenseIndex, const Scalar *, DenseIndex, const Scalar *, DenseIndex, Scalar *,
DenseIndex, DenseIndex, const Scalar &, internal::level3_blocking<Scalar, Scalar> &); DenseIndex, DenseIndex, const Scalar &, Eigen::internal::level3_blocking<Scalar, Scalar> &);
static const functype func[8] = { static const functype func[8] = {
// array index: NOTR | (UP << 2) // array index: NOTR | (UP << 2)
(internal::general_matrix_matrix_triangular_product<DenseIndex, Scalar, ColMajor, false, Scalar, RowMajor, Conj, (Eigen::internal::general_matrix_matrix_triangular_product<DenseIndex, Scalar, ColMajor, false, Scalar, RowMajor,
ColMajor, 1, Upper>::run), Conj, ColMajor, 1, Upper>::run),
// array index: TR | (UP << 2) // array index: TR | (UP << 2)
(internal::general_matrix_matrix_triangular_product<DenseIndex, Scalar, RowMajor, false, Scalar, ColMajor, Conj, (Eigen::internal::general_matrix_matrix_triangular_product<DenseIndex, Scalar, RowMajor, false, Scalar, ColMajor,
ColMajor, 1, Upper>::run), Conj, ColMajor, 1, Upper>::run),
// array index: ADJ | (UP << 2) // array index: ADJ | (UP << 2)
(internal::general_matrix_matrix_triangular_product<DenseIndex, Scalar, RowMajor, Conj, Scalar, ColMajor, false, (Eigen::internal::general_matrix_matrix_triangular_product<DenseIndex, Scalar, RowMajor, Conj, Scalar, ColMajor,
ColMajor, 1, Upper>::run), false, ColMajor, 1, Upper>::run),
0, 0,
// array index: NOTR | (LO << 2) // array index: NOTR | (LO << 2)
(internal::general_matrix_matrix_triangular_product<DenseIndex, Scalar, ColMajor, false, Scalar, RowMajor, Conj, (Eigen::internal::general_matrix_matrix_triangular_product<DenseIndex, Scalar, ColMajor, false, Scalar, RowMajor,
ColMajor, 1, Lower>::run), Conj, ColMajor, 1, Lower>::run),
// array index: TR | (LO << 2) // array index: TR | (LO << 2)
(internal::general_matrix_matrix_triangular_product<DenseIndex, Scalar, RowMajor, false, Scalar, ColMajor, Conj, (Eigen::internal::general_matrix_matrix_triangular_product<DenseIndex, Scalar, RowMajor, false, Scalar, ColMajor,
ColMajor, 1, Lower>::run), Conj, ColMajor, 1, Lower>::run),
// array index: ADJ | (LO << 2) // array index: ADJ | (LO << 2)
(internal::general_matrix_matrix_triangular_product<DenseIndex, Scalar, RowMajor, Conj, Scalar, ColMajor, false, (Eigen::internal::general_matrix_matrix_triangular_product<DenseIndex, Scalar, RowMajor, Conj, Scalar, ColMajor,
ColMajor, 1, Lower>::run), false, ColMajor, 1, Lower>::run),
0}; 0};
#endif #endif
@ -508,7 +558,8 @@ EIGEN_BLAS_FUNC(syrk)
alpha * matrix(a, *k, *n, *lda).transpose() * matrix(a, *k, *n, *lda); alpha * matrix(a, *k, *n, *lda).transpose() * matrix(a, *k, *n, *lda);
} }
#else #else
internal::gemm_blocking_space<ColMajor, Scalar, Scalar, Dynamic, Dynamic, Dynamic> blocking(*n, *n, *k, 1, false); Eigen::internal::gemm_blocking_space<ColMajor, Scalar, Scalar, Dynamic, Dynamic, Dynamic> blocking(*n, *n, *k, 1,
false);
int code = OP(*op) | (UPLO(*uplo) << 2); int code = OP(*op) | (UPLO(*uplo) << 2);
func[code](*n, *k, a, *lda, a, *lda, c, 1, *ldc, alpha, blocking); func[code](*n, *k, a, *lda, a, *lda, c, 1, *ldc, alpha, blocking);
@ -546,6 +597,8 @@ EIGEN_BLAS_FUNC(syr2k)
info = 12; info = 12;
if (info) return xerbla_(SCALAR_SUFFIX_UP "SYR2K", &info); if (info) return xerbla_(SCALAR_SUFFIX_UP "SYR2K", &info);
using Eigen::Lower;
using Eigen::Upper;
if (beta != Scalar(1)) { if (beta != Scalar(1)) {
if (UPLO(*uplo) == UP) if (UPLO(*uplo) == UP)
if (beta == Scalar(0)) if (beta == Scalar(0))
@ -621,16 +674,25 @@ EIGEN_BLAS_FUNC(hemm)
if (*m == 0 || *n == 0) return; if (*m == 0 || *n == 0) return;
using Eigen::ColMajor;
using Eigen::DenseIndex;
using Eigen::Dynamic;
using Eigen::RowMajor;
using Eigen::Upper;
int size = (SIDE(*side) == LEFT) ? (*m) : (*n); int size = (SIDE(*side) == LEFT) ? (*m) : (*n);
internal::gemm_blocking_space<ColMajor, Scalar, Scalar, Dynamic, Dynamic, Dynamic> blocking(*m, *n, size, 1, false); Eigen::internal::gemm_blocking_space<ColMajor, Scalar, Scalar, Dynamic, Dynamic, Dynamic> blocking(*m, *n, size, 1,
false);
if (SIDE(*side) == LEFT) { if (SIDE(*side) == LEFT) {
if (UPLO(*uplo) == UP) if (UPLO(*uplo) == UP)
internal::product_selfadjoint_matrix<Scalar, DenseIndex, RowMajor, true, Conj, ColMajor, false, false, ColMajor, Eigen::internal::product_selfadjoint_matrix<Scalar, DenseIndex, RowMajor, true, Conj, ColMajor, false, false,
1>::run(*m, *n, a, *lda, b, *ldb, c, 1, *ldc, alpha, blocking); ColMajor, 1>::run(*m, *n, a, *lda, b, *ldb, c, 1, *ldc, alpha,
blocking);
else if (UPLO(*uplo) == LO) else if (UPLO(*uplo) == LO)
internal::product_selfadjoint_matrix<Scalar, DenseIndex, ColMajor, true, false, ColMajor, false, false, ColMajor, Eigen::internal::product_selfadjoint_matrix<Scalar, DenseIndex, ColMajor, true, false, ColMajor, false, false,
1>::run(*m, *n, a, *lda, b, *ldb, c, 1, *ldc, alpha, blocking); ColMajor, 1>::run(*m, *n, a, *lda, b, *ldb, c, 1, *ldc, alpha,
blocking);
else else
return; return;
} else if (SIDE(*side) == RIGHT) { } else if (SIDE(*side) == RIGHT) {
@ -642,8 +704,9 @@ EIGEN_BLAS_FUNC(hemm)
RowMajor,true,Conj, ColMajor, 1> RowMajor,true,Conj, ColMajor, 1>
::run(*m, *n, b, *ldb, a, *lda, c, 1, *ldc, alpha, blocking);*/ ::run(*m, *n, b, *ldb, a, *lda, c, 1, *ldc, alpha, blocking);*/
else if (UPLO(*uplo) == LO) else if (UPLO(*uplo) == LO)
internal::product_selfadjoint_matrix<Scalar, DenseIndex, ColMajor, false, false, ColMajor, true, false, ColMajor, Eigen::internal::product_selfadjoint_matrix<Scalar, DenseIndex, ColMajor, false, false, ColMajor, true, false,
1>::run(*m, *n, b, *ldb, a, *lda, c, 1, *ldc, alpha, blocking); ColMajor, 1>::run(*m, *n, b, *ldb, a, *lda, c, 1, *ldc, alpha,
blocking);
else else
return; return;
} else { } else {
@ -658,25 +721,32 @@ EIGEN_BLAS_FUNC(herk)
const int *lda, const RealScalar *pbeta, RealScalar *pc, const int *ldc) { const int *lda, const RealScalar *pbeta, RealScalar *pc, const int *ldc) {
// std::cerr << "in herk " << *uplo << " " << *op << " " << *n << " " << *k << " " << *palpha << " " << *lda << " " // std::cerr << "in herk " << *uplo << " " << *op << " " << *n << " " << *k << " " << *palpha << " " << *lda << " "
// << *pbeta << " " << *ldc << "\n"; // << *pbeta << " " << *ldc << "\n";
using Eigen::ColMajor;
using Eigen::DenseIndex;
using Eigen::Dynamic;
using Eigen::Lower;
using Eigen::RowMajor;
using Eigen::StrictlyLower;
using Eigen::StrictlyUpper;
using Eigen::Upper;
typedef void (*functype)(DenseIndex, DenseIndex, const Scalar *, DenseIndex, const Scalar *, DenseIndex, Scalar *, typedef void (*functype)(DenseIndex, DenseIndex, const Scalar *, DenseIndex, const Scalar *, DenseIndex, Scalar *,
DenseIndex, DenseIndex, const Scalar &, Eigen::internal::level3_blocking<Scalar, Scalar> &); DenseIndex, DenseIndex, const Scalar &, Eigen::internal::level3_blocking<Scalar, Scalar> &);
static const functype func[8] = { static const functype func[8] = {
// array index: NOTR | (UP << 2) // array index: NOTR | (UP << 2)
(internal::general_matrix_matrix_triangular_product<DenseIndex, Scalar, ColMajor, false, Scalar, RowMajor, Conj, (Eigen::internal::general_matrix_matrix_triangular_product<DenseIndex, Scalar, ColMajor, false, Scalar, RowMajor,
ColMajor, 1, Upper>::run), Conj, ColMajor, 1, Upper>::run),
0, 0,
// array index: ADJ | (UP << 2) // array index: ADJ | (UP << 2)
(internal::general_matrix_matrix_triangular_product<DenseIndex, Scalar, RowMajor, Conj, Scalar, ColMajor, false, (Eigen::internal::general_matrix_matrix_triangular_product<DenseIndex, Scalar, RowMajor, Conj, Scalar, ColMajor,
ColMajor, 1, Upper>::run), false, ColMajor, 1, Upper>::run),
0, 0,
// array index: NOTR | (LO << 2) // array index: NOTR | (LO << 2)
(internal::general_matrix_matrix_triangular_product<DenseIndex, Scalar, ColMajor, false, Scalar, RowMajor, Conj, (Eigen::internal::general_matrix_matrix_triangular_product<DenseIndex, Scalar, ColMajor, false, Scalar, RowMajor,
ColMajor, 1, Lower>::run), Conj, ColMajor, 1, Lower>::run),
0, 0,
// array index: ADJ | (LO << 2) // array index: ADJ | (LO << 2)
(internal::general_matrix_matrix_triangular_product<DenseIndex, Scalar, RowMajor, Conj, Scalar, ColMajor, false, (Eigen::internal::general_matrix_matrix_triangular_product<DenseIndex, Scalar, RowMajor, Conj, Scalar, ColMajor,
ColMajor, 1, Lower>::run), false, ColMajor, 1, Lower>::run),
0}; 0};
const Scalar *a = reinterpret_cast<const Scalar *>(pa); const Scalar *a = reinterpret_cast<const Scalar *>(pa);
@ -722,7 +792,8 @@ EIGEN_BLAS_FUNC(herk)
} }
if (*k > 0 && alpha != RealScalar(0)) { if (*k > 0 && alpha != RealScalar(0)) {
internal::gemm_blocking_space<ColMajor, Scalar, Scalar, Dynamic, Dynamic, Dynamic> blocking(*n, *n, *k, 1, false); Eigen::internal::gemm_blocking_space<ColMajor, Scalar, Scalar, Dynamic, Dynamic, Dynamic> blocking(*n, *n, *k, 1,
false);
func[code](*n, *k, a, *lda, a, *lda, c, 1, *ldc, alpha, blocking); func[code](*n, *k, a, *lda, a, *lda, c, 1, *ldc, alpha, blocking);
matrix(c, *n, *n, *ldc).diagonal().imag().setZero(); matrix(c, *n, *n, *ldc).diagonal().imag().setZero();
} }
@ -759,6 +830,10 @@ EIGEN_BLAS_FUNC(her2k)
info = 12; info = 12;
if (info) return xerbla_(SCALAR_SUFFIX_UP "HER2K", &info); if (info) return xerbla_(SCALAR_SUFFIX_UP "HER2K", &info);
using Eigen::Lower;
using Eigen::StrictlyLower;
using Eigen::StrictlyUpper;
using Eigen::Upper;
if (beta != RealScalar(1)) { if (beta != RealScalar(1)) {
if (UPLO(*uplo) == UP) if (UPLO(*uplo) == UP)
if (beta == Scalar(0)) if (beta == Scalar(0))
@ -783,20 +858,20 @@ EIGEN_BLAS_FUNC(her2k)
if (UPLO(*uplo) == UP) { if (UPLO(*uplo) == UP) {
matrix(c, *n, *n, *ldc).triangularView<Upper>() += matrix(c, *n, *n, *ldc).triangularView<Upper>() +=
alpha * matrix(a, *n, *k, *lda) * matrix(b, *n, *k, *ldb).adjoint() + alpha * matrix(a, *n, *k, *lda) * matrix(b, *n, *k, *ldb).adjoint() +
numext::conj(alpha) * matrix(b, *n, *k, *ldb) * matrix(a, *n, *k, *lda).adjoint(); Eigen::numext::conj(alpha) * matrix(b, *n, *k, *ldb) * matrix(a, *n, *k, *lda).adjoint();
} else if (UPLO(*uplo) == LO) } else if (UPLO(*uplo) == LO)
matrix(c, *n, *n, *ldc).triangularView<Lower>() += matrix(c, *n, *n, *ldc).triangularView<Lower>() +=
alpha * matrix(a, *n, *k, *lda) * matrix(b, *n, *k, *ldb).adjoint() + alpha * matrix(a, *n, *k, *lda) * matrix(b, *n, *k, *ldb).adjoint() +
numext::conj(alpha) * matrix(b, *n, *k, *ldb) * matrix(a, *n, *k, *lda).adjoint(); Eigen::numext::conj(alpha) * matrix(b, *n, *k, *ldb) * matrix(a, *n, *k, *lda).adjoint();
} else if (OP(*op) == ADJ) { } else if (OP(*op) == ADJ) {
if (UPLO(*uplo) == UP) if (UPLO(*uplo) == UP)
matrix(c, *n, *n, *ldc).triangularView<Upper>() += matrix(c, *n, *n, *ldc).triangularView<Upper>() +=
alpha * matrix(a, *k, *n, *lda).adjoint() * matrix(b, *k, *n, *ldb) + alpha * matrix(a, *k, *n, *lda).adjoint() * matrix(b, *k, *n, *ldb) +
numext::conj(alpha) * matrix(b, *k, *n, *ldb).adjoint() * matrix(a, *k, *n, *lda); Eigen::numext::conj(alpha) * matrix(b, *k, *n, *ldb).adjoint() * matrix(a, *k, *n, *lda);
else if (UPLO(*uplo) == LO) else if (UPLO(*uplo) == LO)
matrix(c, *n, *n, *ldc).triangularView<Lower>() += matrix(c, *n, *n, *ldc).triangularView<Lower>() +=
alpha * matrix(a, *k, *n, *lda).adjoint() * matrix(b, *k, *n, *ldb) + alpha * matrix(a, *k, *n, *lda).adjoint() * matrix(b, *k, *n, *ldb) +
numext::conj(alpha) * matrix(b, *k, *n, *ldb).adjoint() * matrix(a, *k, *n, *lda); Eigen::numext::conj(alpha) * matrix(b, *k, *n, *ldb).adjoint() * matrix(a, *k, *n, *lda);
} }
} }

12
lapack/cholesky.inc
View File

@ -28,9 +28,9 @@ EIGEN_LAPACK_FUNC(potrf)(char *uplo, int *n, RealScalar *pa, int *lda, int *info
MatrixType A(a, *n, *n, *lda); MatrixType A(a, *n, *n, *lda);
int ret; int ret;
if (UPLO(*uplo) == UP) if (UPLO(*uplo) == UP)
ret = int(internal::llt_inplace<Scalar, Upper>::blocked(A)); ret = int(Eigen::internal::llt_inplace<Scalar, Eigen::Upper>::blocked(A));
else else
ret = int(internal::llt_inplace<Scalar, Lower>::blocked(A)); ret = int(Eigen::internal::llt_inplace<Scalar, Eigen::Lower>::blocked(A));
if (ret >= 0) *info = ret + 1; if (ret >= 0) *info = ret + 1;
} }
@ -61,10 +61,10 @@ EIGEN_LAPACK_FUNC(potrs)(char *uplo, int *n, int *nrhs, RealScalar *pa, int *lda
MatrixType B(b, *n, *nrhs, *ldb); MatrixType B(b, *n, *nrhs, *ldb);
if (UPLO(*uplo) == UP) { if (UPLO(*uplo) == UP) {
A.triangularView<Upper>().adjoint().solveInPlace(B); A.triangularView<Eigen::Upper>().adjoint().solveInPlace(B);
A.triangularView<Upper>().solveInPlace(B); A.triangularView<Eigen::Upper>().solveInPlace(B);
} else { } else {
A.triangularView<Lower>().solveInPlace(B); A.triangularView<Eigen::Lower>().solveInPlace(B);
A.triangularView<Lower>().adjoint().solveInPlace(B); A.triangularView<Eigen::Lower>().adjoint().solveInPlace(B);
} }
} }

5
lapack/eigenvalues.inc
View File

@ -47,9 +47,10 @@ EIGEN_LAPACK_FUNC(syev)
mat = matrix(a, *n, *n, *lda); mat = matrix(a, *n, *n, *lda);
bool computeVectors = *jobz == 'V' || *jobz == 'v'; bool computeVectors = *jobz == 'V' || *jobz == 'v';
SelfAdjointEigenSolver<PlainMatrixType> eig(mat, computeVectors ? ComputeEigenvectors : EigenvaluesOnly); Eigen::SelfAdjointEigenSolver<PlainMatrixType> eig(
mat, computeVectors ? Eigen::ComputeEigenvectors : Eigen::EigenvaluesOnly);
if (eig.info() == NoConvergence) { if (eig.info() == Eigen::NoConvergence) {
make_vector(w, *n).setZero(); make_vector(w, *n).setZero();
if (computeVectors) matrix(a, *n, *n, *lda).setIdentity(); if (computeVectors) matrix(a, *n, *n, *lda).setIdentity();
//*info = 1; //*info = 1;

2
lapack/lu.inc
View File

@ -62,6 +62,8 @@ EIGEN_LAPACK_FUNC(getrs)
MatrixType lu(a, *n, *n, *lda); MatrixType lu(a, *n, *n, *lda);
MatrixType B(b, *n, *nrhs, *ldb); MatrixType B(b, *n, *nrhs, *ldb);
using Eigen::UnitLower;
using Eigen::Upper;
for (int i = 0; i < *n; ++i) ipiv[i]--; for (int i = 0; i < *n; ++i) ipiv[i]--;
if (OP(*trans) == NOTR) { if (OP(*trans) == NOTR) {
B = PivotsType(ipiv, *n) * B; B = PivotsType(ipiv, *n) * B;

18
lapack/svd.inc
View File

@ -56,12 +56,12 @@ EIGEN_LAPACK_FUNC(gesdd)
PlainMatrixType mat(*m, *n); PlainMatrixType mat(*m, *n);
mat = matrix(a, *m, *n, *lda); mat = matrix(a, *m, *n, *lda);
int option = *jobz == 'A' ? ComputeFullU | ComputeFullV int option = *jobz == 'A' ? Eigen::ComputeFullU | Eigen::ComputeFullV
: *jobz == 'S' ? ComputeThinU | ComputeThinV : *jobz == 'S' ? Eigen::ComputeThinU | Eigen::ComputeThinV
: *jobz == 'O' ? ComputeThinU | ComputeThinV : *jobz == 'O' ? Eigen::ComputeThinU | Eigen::ComputeThinV
: 0; : 0;
BDCSVD<PlainMatrixType> svd(mat, option); Eigen::BDCSVD<PlainMatrixType> svd(mat, option);
make_vector(s, diag_size) = svd.singularValues().head(diag_size); make_vector(s, diag_size) = svd.singularValues().head(diag_size);
@ -119,14 +119,14 @@ EIGEN_LAPACK_FUNC(gesvd)
PlainMatrixType mat(*m, *n); PlainMatrixType mat(*m, *n);
mat = matrix(a, *m, *n, *lda); mat = matrix(a, *m, *n, *lda);
int option = (*jobu == 'A' ? ComputeFullU int option = (*jobu == 'A' ? Eigen::ComputeFullU
: *jobu == 'S' || *jobu == 'O' ? ComputeThinU : *jobu == 'S' || *jobu == 'O' ? Eigen::ComputeThinU
: 0) | : 0) |
(*jobv == 'A' ? ComputeFullV (*jobv == 'A' ? Eigen::ComputeFullV
: *jobv == 'S' || *jobv == 'O' ? ComputeThinV : *jobv == 'S' || *jobv == 'O' ? Eigen::ComputeThinV
: 0); : 0);
JacobiSVD<PlainMatrixType> svd(mat, option); Eigen::JacobiSVD<PlainMatrixType> svd(mat, option);
make_vector(s, diag_size) = svd.singularValues().head(diag_size); make_vector(s, diag_size) = svd.singularValues().head(diag_size);
{ {