Sprinkle in some OpenMP macros for SIMD parallelization.

libscran · Sep 1, 2024 · ed15765 · ed15765
1 parent d9b62af
commit ed15765
Show file tree

Hide file tree

Showing 2 changed files with 27 additions and 0 deletions.
diff --git a/include/scran_aggregate/aggregate_across_cells.hpp b/include/scran_aggregate/aggregate_across_cells.hpp
@@ -127,10 +127,16 @@ void compute_aggregate_by_row(
                 std::fill(tmp_sums.begin(), tmp_sums.end(), 0);
 
                 if constexpr(sparse_) {
+#ifdef _OPENMP
+                    #pragma omp simd
+#endif
                     for (Index_ j = 0; j < row.number; ++j) {
                         tmp_sums[factor[row.index[j]]] += row.value[j];
                     }
                 } else {
+#ifdef _OPENMP
+                    #pragma omp simd
+#endif
                     for (Index_ j = 0; j < NC; ++j) {
                         tmp_sums[factor[j]] += row[j];
                     }
@@ -146,10 +152,16 @@ void compute_aggregate_by_row(
                 std::fill(tmp_detected.begin(), tmp_detected.end(), 0);
 
                 if constexpr(sparse_) {
+#ifdef _OPENMP
+                    #pragma omp simd
+#endif
                     for (Index_ j = 0; j < row.number; ++j) {
                         tmp_detected[factor[row.index[j]]] += (row.value[j] > 0);
                     }
                 } else {
+#ifdef _OPENMP
+                    #pragma omp simd
+#endif
                     for (Index_ j = 0; j < NC; ++j) {
                         tmp_detected[factor[j]] += (row[j] > 0);
                     }
@@ -186,13 +198,19 @@ void compute_aggregate_by_column(
                 auto col = ext->fetch(vbuffer.data(), ibuffer.data());
                 if (buffers.sums.size()) {
                     auto& cursum = buffers.sums[current];
+#ifdef _OPENMP
+                    #pragma omp simd
+#endif
                     for (Index_ i = 0; i < col.number; ++i) {
                         cursum[col.index[i]] += col.value[i];
                     }
                 }
 
                 if (buffers.detected.size()) {
                     auto& curdetected = buffers.detected[current];
+#ifdef _OPENMP
+                    #pragma omp simd
+#endif
                     for (Index_ i = 0; i < col.number; ++i) {
                         curdetected[col.index[i]] += (col.value[i] > 0);
                     }
@@ -203,13 +221,19 @@ void compute_aggregate_by_column(
 
                 if (buffers.sums.size()) {
                     auto cursum = buffers.sums[current] + s;
+#ifdef _OPENMP
+                    #pragma omp simd
+#endif
                     for (Index_ i = 0; i < l; ++i) {
                         cursum[i] += col[i];
                     }
                 }
 
                 if (buffers.detected.size()) {
                     auto curdetected = buffers.detected[current] + s;
+#ifdef _OPENMP
+                    #pragma omp simd
+#endif
                     for (Index_ i = 0; i < l; ++i) {
                         curdetected[i] += (col[i] > 0);
                     }

diff --git a/include/scran_aggregate/combine_factors.hpp b/include/scran_aggregate/combine_factors.hpp
@@ -127,6 +127,9 @@ std::vector<std::vector<Factor_> > combine_factors_unused(size_t n, const std::v
         for (size_t f = nfac - 1; f > 0; --f) {
             const auto& finfo = factors[f - 1];
             auto ff = finfo.first;
+#ifdef _OPENMP
+            #pragma omp simd
+#endif
             for (size_t i = 0; i < n; ++i) {
                 combined[i] += mult * ff[i];
             }