Merge pull request #28 from hcq9102/sender_cholesky

sender_choleskey

apps/choleskey/CMakeLists.txt

@@ -7,7 +7,15 @@ target_include_directories(
           ${ARGPARSE_INCLUDE_DIR} ${MDSPAN_INCLUDE_DIR})
 
 add_executable(choleskey_stdpar choleskey_stdpar.cpp)
+target_link_libraries(choleskey_stdpar stdexec)
 target_include_directories(
   choleskey_stdpar
   PRIVATE ${CMAKE_BINARY_DIR} ${CMAKE_CURRENT_LIST_DIR}/../../include
           ${ARGPARSE_INCLUDE_DIR} ${MDSPAN_INCLUDE_DIR})
+
+add_executable(choleskey_stdpar_snd choleskey_stdpar_snd.cpp)
+target_link_libraries(choleskey_stdpar_snd stdexec)
+target_include_directories(
+  choleskey_stdpar_snd
+  PRIVATE ${CMAKE_BINARY_DIR} ${CMAKE_CURRENT_LIST_DIR}/../../include
+          ${ARGPARSE_INCLUDE_DIR} ${MDSPAN_INCLUDE_DIR})

apps/choleskey/choleskey_stdpar_snd.cpp

@@ -0,0 +1,210 @@
+/*
+ * MIT License
+ *
+ * Copyright (c) 2023 Chuanqiu He 
+ * Copyright (c) 2023 Weile Wei 
+ * Copyright (c) 2023 The Regents of the University of California,
+ * through Lawrence Berkeley National Laboratory (subject to receipt of any
+ * required approvals from the U.S. Dept. of Energy).All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+//
+// This example provides a stdexec(senders/receivers) implementation for choleskey decomposition code.
+#include <algorithm>
+#include <exec/any_sender_of.hpp>
+#include <experimental/mdspan>
+#include <iostream>
+#include <numeric>
+#include <stdexec/execution.hpp>
+#include <vector>
+#include "argparse/argparse.hpp"
+#include "commons.hpp"
+#include "exec/static_thread_pool.hpp"
+
+#include "matrixutil.hpp"
+// using namespace stdexec;
+
+using namespace std;
+
+struct solver {
+
+  using view_2d = std::extents<int, std::dynamic_extent, std::dynamic_extent>;
+
+  template <typename T>
+  std::vector<std::vector<T>> Cholesky_Decomposition(std::vector<T>& vec, int n,
+                                                     int np) {
+
+    // test here first, scheduler from a thread pool
+    exec::static_thread_pool pool(np);
+    stdexec::scheduler auto sch = pool.get_scheduler();
+    stdexec::sender auto begin = stdexec::schedule(sch);
+
+    std::vector<std::vector<T>> lower(n, std::vector<T>(n, 0));
+
+    auto matrix_ms =
+        std::mdspan<T, view_2d, std::layout_right>(vec.data(), n, n);
+
+    auto multiplier_lambda = [=](auto a, auto b) {
+      return a * b;
+    };
+
+    for (int i = 0; i < matrix_ms.extent(0); i++) {
+      for (int j = 0; j <= i; j++) {
+        // avoid over parallelize
+        if (j == 0) {
+          np = 1;
+        } else if (j > 0 && np > j) {
+          np = j;
+        }
+
+        if (j == i)  // summation for diagonals
+        {
+
+          if (i == 0 && j == 0) {
+            lower[j][j] = std::sqrt(matrix_ms(i, j));
+          } else {
+
+            std::vector<T> sum_vec(np);  // sub res for each piece
+            int size = j;  // there are j elements need to be calculated(power)
+
+            stdexec::sender auto send1 =
+                stdexec::bulk(begin, np,
+                              [&](int piece) {
+                                int start = piece * size / np;
+                                int chunk_size = size / np;
+                                int remaining = size % np;
+                                chunk_size += (piece == np - 1) ? remaining : 0;
+
+                                sum_vec[piece] = std::transform_reduce(
+                                    std::execution::par,
+                                    counting_iterator(start),
+                                    counting_iterator(start + chunk_size), 0,
+                                    std ::plus{}, [=](int val) {
+                                      return lower[j][val] * lower[j][val];
+                                    });
+                              }) |
+                stdexec::then([&sum_vec]() {
+                  return std::reduce(std::execution::par, sum_vec.begin(),
+                                     sum_vec.end());
+                });
+
+            auto [sum1] = stdexec::sync_wait(std::move(send1)).value();
+
+            lower[j][j] = std::sqrt(matrix_ms(i, j) - sum1);
+          }
+
+        } else {
+          // Evaluating L(i, j) using L(j, j)
+
+          if (j == 0) {
+            lower[i][j] = (matrix_ms(i, j)) / lower[j][j];
+          } else {
+
+            std::vector<T> sum_vec(np);  // sub res for each piece
+            int size_nondiag = j;
+
+            stdexec::sender auto send2 =
+                stdexec::bulk(
+                    begin, np,
+                    [&](int piece) {
+                      int start = piece * size_nondiag / np;
+                      int chunk_size = size_nondiag / np;
+                      int remaining = size_nondiag % np;
+                      chunk_size += (piece == np - 1) ? remaining : 0;
+
+                      sum_vec[piece] = std::transform_reduce(
+                          std::execution::par, counting_iterator(start),
+                          counting_iterator(start + chunk_size), 0,
+                          std ::plus{},
+                          [=](int k) { return lower[j][k] * lower[i][k]; });
+                    }) |
+                stdexec::then([&sum_vec]() {
+                  return std::reduce(std::execution::par, sum_vec.begin(),
+                                     sum_vec.end());
+                });
+
+            auto [sum2] = stdexec::sync_wait(std::move(send2)).value();
+
+            lower[i][j] = (matrix_ms(i, j) - sum2) / lower[j][j];
+          }
+        }
+      }
+    }
+    return lower;
+  }
+};
+
+///////////////////////////////////////////////////////////////////////////////
+int benchmark(args_params_t const& args) {
+
+  std::uint64_t nd = args.nd;  // Number of matrix dimension.
+  std::uint64_t np = args.np;  // Number of parallel partitions.
+
+  std::vector<int> inputMatrix = generate_pascal_matrix<int>(nd);
+
+  // Create the solver object
+  solver solve;
+
+  // Measure execution time.
+  Timer timer;
+
+  // start decomposation
+  auto res_matrix = solve.Cholesky_Decomposition(inputMatrix, nd, np);
+
+  // Print the final results
+  if (args.results) {
+    // Displaying Lower Triangular and its Transpose
+    cout << setw(6) << " Lower Triangular" << setw(30) << "Transpose" << endl;
+    for (int i = 0; i < nd; i++) {
+      // Lower Triangular
+      for (int j = 0; j < nd; j++)
+        cout << setw(6) << res_matrix[i][j] << "\t";
+      cout << "\t";
+
+      // Transpose of Lower Triangular
+      for (int j = 0; j < nd; j++)
+        cout << setw(6) << res_matrix[j][i] << "\t";
+      cout << endl;
+    }
+  }
+
+  if (args.time) {
+    std::cout << "Duration: " << time << " ms."
+              << "\n";
+  }
+
+  return 0;
+}
+
+// Driver Code for testing
+int main(int argc, char* argv[]) {
+
+  // parse params
+  args_params_t args = argparse::parse<args_params_t>(argc, argv);
+  // see if help wanted
+  if (args.help) {
+    args.print();  // prints all variables
+    return 0;
+  }
+
+  benchmark(args);
+
+  return 0;
+}

apps/choleskey/matrixutil.hpp

@@ -35,7 +35,7 @@ struct args_params_t : public argparse::Args {
   std::uint64_t& nd =
       kwarg("nd", "Number of input(positive definition) matrix dimension(<=18)")
           .set_default(10);
-
+  std::uint64_t& np = kwarg("np", "Number of partitions").set_default(4);
   bool& help = flag("h, help", "print help");
   bool& time = kwarg("t, time", "print time").set_default(true);
 };