feat(ttm): add task level parallelism

bassoy · Nov 22, 2024 · 32053ff · 32053ff
1 parent fb3c0fa
commit 32053ff
Show file tree

Hide file tree

Showing 3 changed files with 247 additions and 86 deletions.
diff --git a/include/tlib/detail/tags.h b/include/tlib/detail/tags.h
@@ -25,13 +25,15 @@ namespace tlib::ttm::parallel_policy
 struct sequential_t          {}; // sequential loops and sequential gemm
 struct parallel_blas_t       {}; // multithreaded gemm
 struct parallel_loop_t       {}; // omp_for with single threaded gemm
+struct parallel_taskloop_t   {}; // omp_task for each loop with single threaded gemm 
 struct parallel_loop_blas_t  {}; // omp_for with multi-threaded gemm
 struct batched_gemm_t        {}; // multithreaded batched gemm with collapsed loops
 struct combined_t            {};
 
 inline constexpr sequential_t         sequential;
 inline constexpr parallel_blas_t      parallel_blas;
 inline constexpr parallel_loop_t      parallel_loop;
+inline constexpr parallel_taskloop_t  parallel_taskloop;
 inline constexpr parallel_loop_blas_t parallel_loop_blas;
 inline constexpr batched_gemm_t       batched_gemm;
 inline constexpr combined_t           combined;