X-Git-Url: http://git.rrze.uni-erlangen.de/gitweb/?p=LbmBenchmarkKernelsPublic.git;a=blobdiff_plain;f=src%2FBenchKernelD3Q19ListPullSplitNt.c;h=05d1d0a5e59a223c3b41669c67055ab32ded894a;hp=dfab54afbdc574c3936d146e28e9925e479a1088;hb=8cafd9ea08a6b1103eab29811227a7ae536dffa6;hpb=109880839321408644c94a34eb31208460b9f46d

diff --git a/src/BenchKernelD3Q19ListPullSplitNt.c b/src/BenchKernelD3Q19ListPullSplitNt.c
index dfab54a..05d1d0a 100644
--- a/src/BenchKernelD3Q19ListPullSplitNt.c
+++ b/src/BenchKernelD3Q19ListPullSplitNt.c
@@ -29,6 +29,7 @@
 #include "Memory.h"
 #include "Vtk.h"
 #include "Vector.h"
+#include "LikwidIf.h"
 
 #include <inttypes.h>
 #include <math.h>
@@ -54,8 +55,8 @@ void FNAME(KernelPullSplitNt1S)(LatticeDesc * ld, KernelData * kernelData, CaseD
 	Assert(kernelData != NULL);
 	Assert(cd != NULL);
 
-	Assert(cd->Omega > 0.0);
-	Assert(cd->Omega < 2.0);
+	Assert(cd->Omega > F(0.0));
+	Assert(cd->Omega < F(2.0));
 
 	KernelData        * kd   = (KernelData *)kernelData;
 	KernelDataList    * kdl  = KDL(kernelData);
@@ -64,16 +65,16 @@ void FNAME(KernelPullSplitNt1S)(LatticeDesc * ld, KernelData * kernelData, CaseD
 	PdfT omega = cd->Omega;
 	const PdfT omegaEven = omega;
 
-	PdfT magicParam = 1.0 / 12.0;
-	const PdfT omegaOdd = 1.0 / (0.5 + magicParam / (1.0 / omega - 0.5));
+	PdfT magicParam = F(1.0) / F(12.0);
+	const PdfT omegaOdd = F(1.0) / (F(0.5) + magicParam / (F(1.0) / omega - F(0.5)));
 
 
-	const PdfT w_0 = 1.0 /  3.0;
-	const PdfT w_1 = 1.0 / 18.0;
-	const PdfT w_2 = 1.0 / 36.0;
+	const PdfT w_0 = F(1.0) / F( 3.0);
+	const PdfT w_1 = F(1.0) / F(18.0);
+	const PdfT w_2 = F(1.0) / F(36.0);
 
-	const PdfT w_1_x3 = w_1 * 3.0;	const PdfT w_1_nine_half = w_1 * 9.0 / 2.0;
-	const PdfT w_2_x3 = w_2 * 3.0;	const PdfT w_2_nine_half = w_2 * 9.0 / 2.0;
+	const PdfT w_1_x3 = w_1 * F(3.0);	const PdfT w_1_nine_half = w_1 * F(9.0) / F(2.0);
+	const PdfT w_2_x3 = w_2 * F(3.0);	const PdfT w_2_nine_half = w_2 * F(9.0) / F(2.0);
 
 	const VPDFT vw_1_x3 = VSET(w_1_x3);
 	const VPDFT vw_2_x3 = VSET(w_2_x3);
@@ -84,7 +85,7 @@ void FNAME(KernelPullSplitNt1S)(LatticeDesc * ld, KernelData * kernelData, CaseD
 	const VPDFT vomegaEven = VSET(omegaEven);
 	const VPDFT vomegaOdd  = VSET(omegaOdd);
 
-	const VPDFT voneHalf = VSET(0.5);
+	const VPDFT voneHalf = VSET(F(0.5));
 
 	// uint32_t nConsecNodes = kdlr->nConsecNodes;
 	// uint32_t * consecNodes = kdlr->ConsecNodes;
@@ -118,6 +119,10 @@ void FNAME(KernelPullSplitNt1S)(LatticeDesc * ld, KernelData * kernelData, CaseD
 		KernelStatistics(kd, ld, cd, 0);
 	#endif
 
+	X_KERNEL_START(kernelData);
+
+	X_LIKWID_START("list-pull-split-nt-1s");
+
 	#ifdef _OPENMP
 		#pragma omp parallel default(none) \
 			shared(nFluid, nCells, kd, kdl, adjList, src, dst, \
@@ -184,6 +189,7 @@ void FNAME(KernelPullSplitNt1S)(LatticeDesc * ld, KernelData * kernelData, CaseD
 
 		for(int iter = 0; iter < maxIterations; ++iter) {
 
+
 #if 1
 			#define INDEX_START	blIndexStart
 			#define INDEX_STOP  blIndexVec
@@ -201,6 +207,8 @@ void FNAME(KernelPullSplitNt1S)(LatticeDesc * ld, KernelData * kernelData, CaseD
 			#define INDEX_STOP	blIndexStop
 			#include "BenchKernelD3Q19ListPullSplitNt1SScalar.h"
 #endif
+
+
 			#pragma omp barrier
 
 			#pragma omp single
@@ -235,6 +243,11 @@ void FNAME(KernelPullSplitNt1S)(LatticeDesc * ld, KernelData * kernelData, CaseD
 		MemFree((void **)&tmpArray);
 	}
 
+
+	X_LIKWID_STOP("list-pull-split-nt-1s");
+
+	X_KERNEL_END(kernelData);
+
 #ifdef VTK_OUTPUT
 	if (cd->VtkOutput) {
 		kd->PdfsActive = src;
@@ -257,8 +270,8 @@ void FNAME(KernelPullSplitNt2S)(LatticeDesc * ld, KernelData * kernelData, CaseD
 	Assert(kernelData != NULL);
 	Assert(cd != NULL);
 
-	Assert(cd->Omega > 0.0);
-	Assert(cd->Omega < 2.0);
+	Assert(cd->Omega > F(0.0));
+	Assert(cd->Omega < F(2.0));
 
 	KernelData        * kd   = (KernelData *)kernelData;
 	KernelDataList    * kdl  = KDL(kernelData);
@@ -267,16 +280,15 @@ void FNAME(KernelPullSplitNt2S)(LatticeDesc * ld, KernelData * kernelData, CaseD
 	PdfT omega = cd->Omega;
 	const PdfT omegaEven = omega;
 
-	PdfT magicParam = 1.0 / 12.0;
-	const PdfT omegaOdd = 1.0 / (0.5 + magicParam / (1.0 / omega - 0.5));
+	PdfT magicParam = F(1.0) / F(12.0);
+	const PdfT omegaOdd = F(1.0) / (F(0.5) + magicParam / (F(1.0) / omega - F(0.5)));
 
+	const PdfT w_0 = F(1.0) / F( 3.0);
+	const PdfT w_1 = F(1.0) / F(18.0);
+	const PdfT w_2 = F(1.0) / F(36.0);
 
-	const PdfT w_0 = 1.0 /  3.0;
-	const PdfT w_1 = 1.0 / 18.0;
-	const PdfT w_2 = 1.0 / 36.0;
-
-	const PdfT w_1_x3 = w_1 * 3.0;	const PdfT w_1_nine_half = w_1 * 9.0 / 2.0;
-	const PdfT w_2_x3 = w_2 * 3.0;	const PdfT w_2_nine_half = w_2 * 9.0 / 2.0;
+	const PdfT w_1_x3 = w_1 * F(3.0);	const PdfT w_1_nine_half = w_1 * F(9.0) / F(2.0);
+	const PdfT w_2_x3 = w_2 * F(3.0);	const PdfT w_2_nine_half = w_2 * F(9.0) / F(2.0);
 
 	const VPDFT vw_1_x3 = VSET(w_1_x3);
 	const VPDFT vw_2_x3 = VSET(w_2_x3);
@@ -287,7 +299,7 @@ void FNAME(KernelPullSplitNt2S)(LatticeDesc * ld, KernelData * kernelData, CaseD
 	const VPDFT vomegaEven = VSET(omegaEven);
 	const VPDFT vomegaOdd  = VSET(omegaOdd);
 
-	const VPDFT voneHalf = VSET(0.5);
+	const VPDFT voneHalf = VSET(F(0.5));
 
 	// uint32_t nConsecNodes = kdlr->nConsecNodes;
 	// uint32_t * consecNodes = kdlr->ConsecNodes;
@@ -321,6 +333,12 @@ void FNAME(KernelPullSplitNt2S)(LatticeDesc * ld, KernelData * kernelData, CaseD
 		KernelStatistics(kd, ld, cd, 0);
 	#endif
 
+
+	X_KERNEL_START(kernelData);
+
+	X_LIKWID_START("list-pull-split-nt-2s");
+
+
 	#ifdef _OPENMP
 		#pragma omp parallel default(none) \
 			shared(nFluid, nCells, kd, kdl, adjList, src, dst, \
@@ -406,6 +424,7 @@ void FNAME(KernelPullSplitNt2S)(LatticeDesc * ld, KernelData * kernelData, CaseD
 #endif
 			#pragma omp barrier
 
+
 			#pragma omp single
 			{
 				#ifdef VERIFICATION
@@ -438,6 +457,10 @@ void FNAME(KernelPullSplitNt2S)(LatticeDesc * ld, KernelData * kernelData, CaseD
 		MemFree((void **)&tmpArray);
 	}
 
+	X_LIKWID_STOP("list-pull-split-nt-2s");
+
+	X_KERNEL_END(kernelData);
+
 #ifdef VTK_OUTPUT
 	if (cd->VtkOutput) {
 		kd->PdfsActive = src;