Commit 768de824 authored by Doak, Peter W.'s avatar Doak, Peter W.
Browse files

more test udpates ahead of merge

parent a31c85e0

include/dca/linalg/matrixop.hpp

+5 −5
Original line number Diff line number Diff line
@@ -154,7 +154,7 @@ inline void copyRow(const Matrix<Scalar, device_name>& mat_x, int ix, 
//                0 <= i_y[i] < mat_y.nrRows() for 0 <= i < i_x.size().

template <typename Scalar, class Vec, class ALLOC>

inline void copyRows(const Matrix<Scalar, CPU, ALLOC>& mat_x, const Vec& i_x, Matrix<Scalar, CPU, ALLOC>& mat_y,

                     const Vec& i_y, int /*thread_id*/ = 0, int /*stream_id*/ = 0) {

                     const Vec& i_y, const int /*thread_id*/ = 0, const int /*stream_id*/ = 0) {

  assert(i_x.size() <= i_y.size());

  assert(mat_x.nrCols() == mat_y.nrCols());
@@ -165,8 +165,8 @@ inline void copyRows(const Matrix<Scalar, CPU, ALLOC>& mat_x, const Vec& i_x, Ma 
#ifdef DCA_HAVE_GPU

template <typename Scalar>

inline void copyRows(const Matrix<Scalar, GPU>& mat_x, const Vector<int, GPU>& i_x,

                     Matrix<Scalar, GPU>& mat_y, const Vector<int, GPU>& i_y, int thread_id = 0,

                     int stream_id = 0) {

                     Matrix<Scalar, GPU>& mat_y, const Vector<int, GPU>& i_y, const int thread_id,

                     const int stream_id) {

  assert(i_x.size() <= i_y.size());

  assert(mat_x.nrCols() == mat_y.nrCols());
@@ -179,9 +179,9 @@ inline void copyRows(const Matrix<Scalar, GPU>& mat_x, const Vector<int, GPU>& i 
// In/Out: mat_y

// Preconditions: mat_x.nrCols() == mat_y.nrCols()

//                0 <= i_x[i] < mat_x.nrRows() for 0 <= i < i_x.size().

template <typename Scalar, class ALLOC>

template <typename Scalar>

inline void copyRows(const Matrix<Scalar, GPU>& mat_x, const Vector<int, GPU>& i_x,

                     Matrix<Scalar, GPU>& mat_y, int thread_id = 0, int stream_id = 0) {

                     Matrix<Scalar, GPU>& mat_y, const int thread_id, const int stream_id) {

  assert(mat_x.nrCols() == mat_y.nrCols());



  blas::copyRows(mat_x.nrCols(), i_x.size(), i_x.ptr(), mat_x.ptr(), mat_x.leadingDimension(),

include/dca/linalg/vector.hpp

+7 −1
Original line number Diff line number Diff line
@@ -50,7 +50,13 @@ public: 
  Vector(size_t size, size_t capacity);

  Vector(const std::string& name, size_t size, size_t capacity);



  Vector(const ThisType& rhs, const std::string& name = default_name_);

  /** copy constructor except for name.

   *  this is strange but for historical reasons is kept.

   *  has needed to be explicit because with the `const ThisType&` somehow lead to an implicit conversion

   *  from an int to a Vector& argument that landed here.

   *  This occurred in Debug with 

   */

  explicit Vector(const ThisType& rhs, const std::string& name = default_name_);



  template <DeviceType device_name2, class Allocator2>

  Vector(const Vector<ScalarType, device_name2, Allocator2>& rhs,

include/dca/phys/dca_step/cluster_solver/ctint/walker/ctint_walker_gpu_submatrix.hpp

+1 −0
Original line number Diff line number Diff line
@@ -294,6 +294,7 @@ void CtintWalkerSubmatrixGpu<Parameters, DIST>::updateM() { 
    old_G.resizeNoCopy(std::make_pair(n_max_[s], gamma_size));

    old_M.resizeNoCopy(std::make_pair(gamma_size, n_max_[s]));



    assert(dca::linalg::util::getStream(thread_id_, s) == get_stream(s));

    move_indices_dev_[s].setAsync(move_indices_[s], get_stream(s));

    // Note: an event synchronization might be necessary if the order of operation is changed.

    linalg::matrixop::copyCols(G_dev_[s], move_indices_dev_[s], old_G, thread_id_, s);

include/dca/testing/minimalist_printer.hpp

+1 −1
Original line number Diff line number Diff line
@@ -13,7 +13,7 @@ 
#ifndef DCA_TESTING_MINIMALIST_PRINTER_HPP

#define DCA_TESTING_MINIMALIST_PRINTER_HPP



#include "gtest/gtest.h"

#include "dca/testing/gtest_h_w_warning_blocking.h"



namespace dca {

namespace testing {

src/phys/dca_step/cluster_solver/ctint/walker/walker_kernels.cu

+1 −1
Original line number Diff line number Diff line
@@ -43,7 +43,7 @@ __global__ void setRightSectorToIdKernel(Scalar* m, const int ldm, const int n0, 
  Scalar the_one{};

  the_one += 1.0;

  Scalar the_zero{};

  assert(the_one == (the_zero -= 1.0));    

  //assert(the_one == (the_zero -= 1.0));    

  m[i + ldm * j] = (i == j) ? the_one : the_zero;

}