Skip to content

S
SWE
Commit f9ef8e12 authored by drossostum's avatar drossostum
Browse files
Options

Intel Cilk Array 

Signed-off-by: 's avatardrossostum <drevangel@mytum.de>
parent cb8cace4
Show whitespace changes
Inline Side-by-side
Showing with 37 additions and 29 deletions
src/blocks/SWE_WavePropagationBlock.cpp
View file @ f9ef8e12
...@@ -36,6 +36,8 @@ ...@@ -36,6 +36,8 @@
#include <string> #include <string>
#include <limits> #include <limits>
//#define vlength 4
#ifdef LOOP_OPENMP #ifdef LOOP_OPENMP
#include <omp.h> #include <omp.h>
#endif #endif
...@@ -149,7 +151,7 @@ SWE_WavePropagationBlock::computeNumericalFluxes () ...@@ -149,7 +151,7 @@ SWE_WavePropagationBlock::computeNumericalFluxes ()
float l_maxWaveSpeed = (float) 0.; float l_maxWaveSpeed = (float) 0.;
// solver::Hybrid<float> wavePropagationSolver; // solver::Hybrid<float> wavePropagationSolver;
#if WAVE_PROPAGATION_SOLVER==4 #if WAVE_PROPAGATION_SOLVER==4
solver::FWaveVec<float> wavePropagationSolver; solver::FWaveVec wavePropagationSolver;
#else // WAVE_PROPAGATION_SOLVER==4 #else // WAVE_PROPAGATION_SOLVER==4
//#pragma message "augmented Riemann solver was hardcoded set for OpenMP!" //#pragma message "augmented Riemann solver was hardcoded set for OpenMP!"
solver::AugRie_SIMD wavePropagationSolver; solver::AugRie_SIMD wavePropagationSolver;
...@@ -186,17 +188,19 @@ SWE_WavePropagationBlock::computeNumericalFluxes () ...@@ -186,17 +188,19 @@ SWE_WavePropagationBlock::computeNumericalFluxes ()
#if WAVE_PROPAGATION_SOLVER==4 and defined VECTORIZE #if WAVE_PROPAGATION_SOLVER==4 and defined VECTORIZE
// Vectorization is currently only possible for the FWaveVec solver // Vectorization is currently only possible for the FWaveVec solver
// Vectorize the inner loop // Vectorize the inner loop
#pragma simd //#pragma simd
#endif // WAVE_PROPAGATION_SOLVER==4 and defined VECTORIZE #endif // WAVE_PROPAGATION_SOLVER==4 and defined VECTORIZE
for (; j < end_ny_1_1; ++j) { for (; j < end_ny_1_1; j+=4) {
//{
//E.Drossos - maxEdgeSpeed takes the max edge speed value for all cells [j; j+vlength] in the vectorized block
float maxEdgeSpeed; float maxEdgeSpeed;
wavePropagationSolver.computeNetUpdates ( wavePropagationSolver.computeNetUpdates (
h[i - 1][j], h[i][j], &h[i - 1][j], &h[i][j],
hu[i - 1][j], hu[i][j], &hu[i - 1][j], &hu[i][j],
b[i - 1][j], b[i][j], &b[i - 1][j], &b[i][j],
hNetUpdatesLeft[i - 1][j - 1], hNetUpdatesRight[i - 1][j - 1], &hNetUpdatesLeft[i - 1][j - 1], &hNetUpdatesRight[i - 1][j - 1],
huNetUpdatesLeft[i - 1][j - 1], huNetUpdatesRight[i - 1][j - 1], &huNetUpdatesLeft[i - 1][j - 1], &huNetUpdatesRight[i - 1][j - 1],
maxEdgeSpeed maxEdgeSpeed
); );
...@@ -248,17 +252,19 @@ SWE_WavePropagationBlock::computeNumericalFluxes () ...@@ -248,17 +252,19 @@ SWE_WavePropagationBlock::computeNumericalFluxes ()
#if WAVE_PROPAGATION_SOLVER==4 and defined VECTORIZE #if WAVE_PROPAGATION_SOLVER==4 and defined VECTORIZE
// Vectorization is currently only possible for the FWaveVec solver // Vectorization is currently only possible for the FWaveVec solver
// Vectorize the inner loop // Vectorize the inner loop
#pragma simd //#pragma simd
#endif // WAVE_PROPAGATION_SOLVER==4 #endif // WAVE_PROPAGATION_SOLVER==4
for (; j < end_ny_1_2; j++) { for (; j < end_ny_1_2; j+=4) {
//{
//E.Drossos - maxEdgeSpeed takes the max edge speed value for all cells [j; j+vlength] in the vectorized block
float maxEdgeSpeed; float maxEdgeSpeed;
wavePropagationSolver.computeNetUpdates ( wavePropagationSolver.computeNetUpdates (
h[i][j - 1], h[i][j], &h[i][j - 1], &h[i][j],
hv[i][j - 1], hv[i][j], &hv[i][j - 1], &hv[i][j],
b[i][j - 1], b[i][j], &b[i][j - 1], &b[i][j],
hNetUpdatesBelow[i - 1][j - 1], hNetUpdatesAbove[i - 1][j - 1], &hNetUpdatesBelow[i - 1][j - 1], &hNetUpdatesAbove[i - 1][j - 1],
hvNetUpdatesBelow[i - 1][j - 1], hvNetUpdatesAbove[i - 1][j - 1], &hvNetUpdatesBelow[i - 1][j - 1], &hvNetUpdatesAbove[i - 1][j - 1],
maxEdgeSpeed maxEdgeSpeed
); );
...@@ -314,20 +320,22 @@ SWE_WavePropagationBlock::computeNumericalFluxes () ...@@ -314,20 +320,22 @@ SWE_WavePropagationBlock::computeNumericalFluxes ()
#if WAVE_PROPAGATION_SOLVER==4 and defined VECTORIZE #if WAVE_PROPAGATION_SOLVER==4 and defined VECTORIZE
// Vectorization is currently only possible for the FWaveVec solver // Vectorization is currently only possible for the FWaveVec solver
// Vectorize the inner loop // Vectorize the inner loop
#pragma simd //#pragma simd
#endif // WAVE_PROPAGATION_SOLVER==4 and defined VECTORIZE #endif // WAVE_PROPAGATION_SOLVER==4 and defined VECTORIZE
#ifdef LOOP_OPENMP #ifdef LOOP_OPENMP
#pragma omp for schedule(static) nowait #pragma omp for schedule(static) nowait
#endif #endif
for (j = 1; j < end_ny_1_1; ++j) { for (j = 1; j < end_ny_1_1; j+=4) {
//{
//E.Drossos - maxEdgeSpeed takes the max edge speed value for all cells [j; j+vlength] in the vectorized block
float maxEdgeSpeed; float maxEdgeSpeed;
wavePropagationSolver.computeNetUpdates ( wavePropagationSolver.computeNetUpdates (
h[i - 1][j], h[i][j], &h[i - 1][j], &h[i][j],
hu[i - 1][j], hu[i][j], &hu[i - 1][j], &hu[i][j],
b[i - 1][j], b[i][j], &b[i - 1][j], &b[i][j],
hNetUpdatesLeft[i - 1][j - 1], hNetUpdatesRight[i - 1][j - 1], &hNetUpdatesLeft[i - 1][j - 1], &hNetUpdatesRight[i - 1][j - 1],
huNetUpdatesLeft[i - 1][j - 1], huNetUpdatesRight[i - 1][j - 1], &huNetUpdatesLeft[i - 1][j - 1], &huNetUpdatesRight[i - 1][j - 1],
maxEdgeSpeed maxEdgeSpeed
); );
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment