unroll critical loops

vastly improves performance in SPEC benchmark
ComputationalRadiationPhysics · Oct 8, 2021 · 794f6d1 · 794f6d1
1 parent 23a5aaf
commit 794f6d1
Show file tree

Hide file tree

Showing 2 changed files with 4 additions and 0 deletions.
diff --git a/include/picongpu/algorithms/AssignedTrilinearInterpolation.hpp b/include/picongpu/algorithms/AssignedTrilinearInterpolation.hpp
@@ -68,12 +68,15 @@ namespace picongpu
             using type = typename ::pmacc::result_of::Functor<AssignedTrilinearInterpolation, T_Cursor>::type;
 
             auto result_z = type(0.0);
+#pragma unroll
             for(int z = T_begin; z <= T_end; ++z)
             {
                 auto result_y = type(0.0);
+#pragma unroll
                 for(int y = T_begin; y <= T_end; ++y)
                 {
                     auto result_x = type(0.0);
+#pragma unroll
                     for(int x = T_begin; x <= T_end; ++x)
                         /* a form factor is the "amount of particle" that is affected by this cell
                          * so we have to sum over: cell_value * form_factor

diff --git a/include/picongpu/algorithms/FieldToParticleInterpolation.hpp b/include/picongpu/algorithms/FieldToParticleInterpolation.hpp
@@ -76,6 +76,7 @@ namespace picongpu
             using Supports = typename pmacc::math::CT::make_Int<simDim, supp>::type;
 
             typename Cursor::ValueType result;
+#pragma unroll
             for(uint32_t i = 0; i < Cursor::ValueType::dim; i++)
             {
                 auto fieldComponent