merge with kernels from MH's master thesis

[LbmBenchmarkKernelsPublic.git] / src / BenchKernelD3Q19.c

diff --git a/src/BenchKernelD3Q19.c b/src/BenchKernelD3Q19.c
index 48529483ca31c1d3ffc52249ba2c7b76d125e6fa..62989c89e747ff32c74fee3906a507c32ce4efa1 100644 (file)
--- a/src/BenchKernelD3Q19.c
+++ b/src/BenchKernelD3Q19.c
@@ -101,6 +101,8 @@ void FNAME(D3Q19Kernel)(LatticeDesc * ld, KernelData * kernelData, CaseData * cd
        }
        #endif
 
+       X_KERNEL_START(kernelData);
+
        for (int iter = 0; iter < maxIterations; ++iter) {
 
                X_LIKWID_START("os");
@@ -125,7 +127,7 @@ void FNAME(D3Q19Kernel)(LatticeDesc * ld, KernelData * kernelData, CaseData * cd
                                        #pragma vector always
                                        #pragma simd
                                #endif
-                               for (int z = oZ; z < nZ + oZ; ++z) {
+                               for (int z = oZ; z < nZ + oZ; ++z) { // LOOP os
                                        #define I(x, y, z, dir) P_INDEX_5(gDims, (x), (y), (z), (dir))
 
 #ifdef PROP_MODEL_PUSH
@@ -320,6 +322,7 @@ void FNAME(D3Q19Kernel)(LatticeDesc * ld, KernelData * kernelData, CaseData * cd
                // Stop counters before bounce back. Else computing loop balance will be incorrect.
                X_LIKWID_STOP("os");
 
+
                // Fixup bounce back PDFs.
                #ifdef _OPENMP
                #pragma omp parallel for default(none) \
@@ -355,6 +358,8 @@ void FNAME(D3Q19Kernel)(LatticeDesc * ld, KernelData * kernelData, CaseData * cd
 
        } // for (int iter = 0; ...
 
+       X_KERNEL_END(kernelData);
+
        #ifdef VTK_OUTPUT
 
        if (cd->VtkOutput) {
@@ -445,6 +450,8 @@ void FNAME(D3Q19BlkKernel)(LatticeDesc * ld, KernelData * kernelData, CaseData *
        nThreads = omp_get_max_threads();
        #endif
 
+       X_KERNEL_START(kernelData);
+
        for (int iter = 0; iter < maxIterations; ++iter) {
 
                #ifdef _OPENMP
@@ -729,6 +736,8 @@ void FNAME(D3Q19BlkKernel)(LatticeDesc * ld, KernelData * kernelData, CaseData *
 
        } // for (int iter = 0; ...
 
+       X_KERNEL_END(kernelData);
+
        #ifdef VTK_OUTPUT
 
        if (cd->VtkOutput) {