Merge pull request #30 from PROBIC/parallelism-fixes

Parallelism & CI fixes

.github/workflows/make_release.yml

@@ -9,7 +9,7 @@ on:
 jobs:
   build_linux-x86_64:
     runs-on: ubuntu-latest
-    container: phusion/holy-build-box-64:3.0.2
+    container: phusion/holy-build-box:4.0.1
     steps:
       - name: Install wget
         id: install-wget
@@ -125,4 +125,4 @@ jobs:
           fail_on_unmatched_files: true
           generate_release_notes: true
           files: |
-            mSWEEP-*.tar.gz
+            mSWEEP-*.tar.gz

CMakeLists.txt

@@ -243,7 +243,7 @@ if (DEFINED CMAKE_TELESCOPE_LIBRARY AND DEFINED CMAKE_TELESCOPE_HEADERS)
 else()
   FetchContent_Declare(telescope
     GIT_REPOSITORY    https://github.com/tmaklin/telescope.git
-    GIT_TAG           v0.7.0-prerelease
+    GIT_TAG           v0.7.3
     PREFIX            "external"
     SOURCE_DIR        "${CMAKE_CURRENT_SOURCE_DIR}/external/telescope"
     BINARY_DIR        "${CMAKE_CURRENT_BINARY_DIR}/external/telescope"

include/Likelihood.hpp

@@ -135,7 +135,6 @@ class LL_WOR21 : public Likelihood<T> {
     const seamat::DenseMatrix<T> &precalc_lls_mat = this->precalc_lls(masked_group_sizes, n_masked_groups);
 
     this->log_likelihoods.resize(n_masked_groups, num_ecs, std::log(this->zero_inflation));
-#pragma omp parallel for schedule(static) shared(precalc_lls_mat)
     for (size_t j = 0; j < num_ecs; ++j) {
       size_t groups_pos = 0;
       for (size_t i = 0; i < n_groups; ++i) {

src/Sample.cpp

@@ -101,6 +101,7 @@ void Sample::dirichlet_kld(const std::vector<double> &log_ec_hit_counts) {
   size_t cols = this->get_probs().get_cols();
 
   std::vector<double> alphas(rows, 0.0);
+#pragma omp parallel for schedule(static)
   for (size_t i = 0; i < rows; ++i) {
     for (size_t j = 0; j < cols; ++j) {
       size_t num_hits = std::round(std::exp(log_ec_hit_counts[j]));
@@ -111,11 +112,13 @@ void Sample::dirichlet_kld(const std::vector<double> &log_ec_hit_counts) {
   }
 
   double alpha0 = 0.0;
+#pragma omp parallel for schedule(static) reduction(+:alpha0)
   for (size_t i = 0; i < rows; ++i) {
     alpha0 += alphas[i];
   }
 
   this->log_KLDs.resize(rows);
+#pragma omp parallel for schedule(static)
   for (size_t i = 0; i < rows; ++i) {
     double log_theta = std::log(alphas[i]) - std::log(alpha0);
     double alpha_k = alphas[rows - 1];