Added OpenMP (Fortran 90) and C++ threads code samples for gemm.

gemm/gemm_openmp_fortran/CMakeLists.txt

+
+# ==================================================================================================
+# This file is part of the CodeVault project. The project is licensed under Apache Version 2.0.
+# CodeVault is part of the EU-project PRACE-4IP (WP7.3.C).
+#
+# Author(s):
+#  Mariusz Uchronski <mariusz.uchronski@pwr.edu.nl>
+#
+# ==================================================================================================
+
+cmake_minimum_required(VERSION 2.8.7 FATAL_ERROR)
+include(${CMAKE_CURRENT_SOURCE_DIR}/../../../cmake/common.cmake)
+
+# ==================================================================================================
+if ("${DWARF_PREFIX}" STREQUAL "")
+  set(DWARF_PREFIX dense_linear_algebra)
+endif()
+
+find_package(Common)
+find_package(OpenMP)
+
+enable_language (Fortran)
+
+set(NAME ${DWARF_PREFIX}_gemm_openmp_fortran)
+if (OPENMP_FOUND)
+  set (CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${OpenMP_CXX_FLAGS}")
+  add_executable(${NAME} src/gemm.f90)
+  install(TARGETS ${NAME} DESTINATION bin)
+else()
+  message("## Skipping '${NAME}_omp': no OpenMP support found")
+  dummy_install(${NAME} "OpenMP")
+endif()
+
+unset(NAME)

gemm/gemm_openmp_fortran/README.md

+=======
+README
+=======
+
+# 1. Code sample name
+gemm
+
+# 2. Description of the code sample package
+This example demonstrates the use of Fortran 90 with OpenMP for matrix-matrix multiplication
+
+# 3. Release date
+15 August 2016
+
+# 4. Version history
+1.0
+
+# 6. Copyright / License of the code sample
+Apache Version 2.0
+
+# 5. Contributor (s) / Maintainer(s)
+Mariusz Uchronski <mariusz.uchronski@pwr.edu.pl>
+
+# 7. Language(s)
+Fortran 90
+
+# 8. Parallelisation Implementation(s)
+OpenMP
+
+# 9. Level of the code sample complexity
+basic
+
+# 10. Instructions on how to compile the code
+Uses the CodeVault CMake infrastructure, see main README.md
+
+# 11. Instructions on how to run the code
+Just run compiled executable
+
+# 12. Sample input(s)
+Randomly generated on sample code execution
+
+# 13. Sample output(s)
+Not available

gemm/gemm_openmp_fortran/src/gemm.f90

+
+! =================================================================================================
+! This file is part of the CodeVault project. The project is licensed under Apache Version 2.0.
+! CodeVault is part of the EU-project PRACE-4IP (WP7.3.C).
+!
+! Author(s):
+!   Mariusz Uchronski <mariusz.uchronski@pwr.edu.pl>
+!
+! This example demonstrates the use of Fortran 90 with OpenMP for matrix-matrix multiplication.
+! The example is set-up to perform single precision matrix-matrix multiplication.
+!
+! See [http://openmp.org] for the full OpenMP documentation.
+!
+! =================================================================================================
+
+program gemm_openmp
+
+  implicit none
+  integer,parameter :: seed = 86456
+  integer, parameter :: n = 400
+  integer, parameter :: m = 600
+  integer, parameter :: k = 800
+  real, allocatable :: a(:)
+  real, allocatable :: b(:)
+  real, allocatable :: c(:)
+
+  allocate(a(n*m))
+  allocate(b(m*k))
+  allocate(c(n*k))
+
+  call fill_random(a, n, m)
+  call fill_random(b, m, k)
+
+  call run_gemm_openmp(a, b, c, n, m, k)
+
+  deallocate(a)
+  deallocate(b)
+  deallocate(c)
+
+
+end program gemm_openmp
+
+
+subroutine fill_random(a, n, m)
+
+  implicit none
+
+  integer, intent(in) :: n, m
+  real, dimension(n*m), intent(inout) :: a
+  integer :: i, j
+  do i=1, n
+    do j=1, m
+      a((i-1)*m+j) = rand()
+    end do
+  end do
+
+end subroutine fill_random
+
+subroutine print_array(a, n, m)
+
+  implicit none
+
+  integer, intent(in) :: n, m
+  real, dimension(n*m), intent(in) :: a
+  integer :: i, j
+  do i=1, n
+    do j=1, m
+      print '(F8.4,$)', a((i-1)*m+j)
+    end do
+    print *, ''
+  end do
+
+end subroutine print_array
+
+
+subroutine run_gemm_openmp(a, b, c, n, m, k)
+
+  integer, intent(in) :: n, m , k
+  real, dimension(n*m), intent(in) :: a
+  real, dimension(m*k), intent(in) :: b
+  real, dimension(n*k), intent(inout) :: c
+  integer :: i, j, l
+  real :: summ
+
+
+!$OMP PARALLEL DO DEFAULT (NONE) &
+!$OMP SHARED(n,m,k,a,b,c) PRIVATE(i,j,l,summ)
+  do i=1, n
+    do j=1, k
+      summ = 0
+      do l=1, m
+        summ = summ + a((i-1)*m+l) * b((l-1)*k+j)
+      end do
+      c((i-1)*k+j) = summ
+    end do
+  end do
+!$OMP END PARALLEL DO
+
+end subroutine

gemm/gemm_threads/CMakeLists.txt

+
+# ==================================================================================================
+# This file is part of the CodeVault project. The project is licensed under Apache Version 2.0.
+# CodeVault is part of the EU-project PRACE-4IP (WP7.3.C).
+#
+# Author(s):
+#  Mariusz Uchronski <mariusz.uchronski@pwr.edu.nl>
+#
+# ==================================================================================================
+
+cmake_minimum_required(VERSION 2.8.7 FATAL_ERROR)
+include(${CMAKE_CURRENT_SOURCE_DIR}/../../../cmake/common.cmake)
+
+# ==================================================================================================
+if ("${DWARF_PREFIX}" STREQUAL "")
+  set(DWARF_PREFIX gemm_threads)
+endif()
+
+find_package(Common)
+
+enable_language (C)
+
+set(NAME ${DWARF_PREFIX}_gemm_threads)
+
+set(CXX_FLAGS "${CXX_FLAGS} -Wall -Wno-comment -std=c++0x -g -pthread")
+set (CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${CXX_FLAGS}")
+add_executable(${NAME} src/gemm.cpp)
+install(TARGETS ${NAME} DESTINATION bin)
+
+unset(NAME)

gemm/gemm_threads/README.md

+=======
+README
+=======
+
+# 1. Code sample name
+gemm
+
+# 2. Description of the code sample package
+This example demonstrates the use of C++ threads for matrix-matrix multiplication.
+
+# 3. Release date
+15 August 2016
+
+# 4. Version history
+1.0
+
+# 6. Copyright / License of the code sample
+Apache Version 2.0
+
+# 5. Contributor (s) / Maintainer(s)
+Mariusz Uchronski <mariusz.uchronski@pwr.edu.pl>
+
+# 7. Language(s)
+C++ 11
+
+# 8. Parallelisation Implementation(s)
+OpenMP
+
+# 9. Level of the code sample complexity
+basic
+
+# 10. Instructions on how to compile the code
+Uses the CodeVault CMake infrastructure, see main README.md
+
+# 11. Instructions on how to run the code
+Just run compiled executable
+
+# 12. Sample input(s)
+Randomly generated on sample code execution
+
+# 13. Sample output(s)
+Not available

gemm/gemm_threads/src/gemm.cpp

+
+// =================================================================================================
+// This file is part of the CodeVault project. The project is licensed under Apache Version 2.0.
+// CodeVault is part of the EU-project PRACE-4IP (WP7.3.C).
+//
+// Author(s):
+//   Mariusz Uchronski <mariusz.uchronski@pwr.edu.pl>
+//
+// This example demonstrates the use of C++ threads for matrix-matrix multiplication.
+// The example is set-up to perform single precision matrix-matrix multiplication.
+//
+// See [http://www.cplusplus.com/reference/thread/thread/] for the full C++ threads documentation.
+//
+// =================================================================================================
+
+#include <thread>
+#include <random>
+#include <iostream>
+#include <vector>
+#include <algorithm>
+
+void fill_random(float *a, const int &n, const int &m)
+{
+  std::mt19937 e(static_cast<unsigned int>(time(nullptr)));
+  std::uniform_real_distribution<float> f;
+  for(int i=0; i<n; ++i)
+  {
+    for(int j=0; j<m; ++j)
+    {
+      a[i*m+j] = f(e);
+    }
+  }
+}
+
+void gemm_threads(const int &num_threads, const int &id,
+                  const int &n, const int &m, const int &k,
+                  float *a, float *b, float *c)
+{
+  const int part = n/num_threads;
+  const int begin = id*part;
+  const int end = (id+1)*part - 1;
+  for(int i=begin; i<=end; ++i)
+  {
+    for(int j=0; j<k; ++j)
+    {
+      float sum = 0.0;
+      for(int l=0; l<m; ++l)
+      {
+        sum = sum + a[i*m+l] * b[l*k+j];
+      }
+      c[i*k+j] = sum;
+    }
+  }
+}
+
+void print_data(float *a, const int &n, const int &m)
+{
+  for(int i=0; i<n; ++i)
+  {
+    for(int j=0; j<m; ++j)
+    {
+      std::cout << a[i*m+j] << " ";
+    }
+    std::cout << std::endl;
+  }
+}
+
+int main(int argc, char *argv[])
+{
+  const int n = 200;
+  const int m = 300;
+  const int k = 400;
+  const int num_threads = 4;
+  std::vector<std::thread> threads;
+
+  float *a = new float[n*m];
+  float *b = new float[m*k];
+  float *c = new float[n*k];
+
+  fill_random(a, n, m);
+  fill_random(b, m, k);
+
+  for(int i=0; i<num_threads; ++i)
+    threads.push_back(std::thread(gemm_threads, num_threads, i, n, m, k, a, b, c));
+  std::for_each(threads.begin(), threads.end(), std::mem_fn(&std::thread::join));
+
+  delete[] a;
+  delete[] b;
+  delete[] c;
+  return 0;
+}