Merge pull request #42 from Stoops-ML/0.3.2

FIX: OMP parallelised loops + add tests
Stoops-ML · Oct 10, 2024 · f3afbaf · f3afbaf
2 parents 7d3d440 + 22b87b5
commit f3afbaf
Show file tree

Hide file tree

Showing 19 changed files with 416 additions and 125 deletions.
diff --git a/include/distances.c b/include/distances.c
@@ -24,11 +24,11 @@ void HaversineDouble(const double* rrmStart,
     double mRadiusSphere,
     double* mDistance)
 {
-    int iPoint, iPointEnd, iPointStart;
+    int iPoint;
 #pragma omp parallel for if (nPoints > omp_get_num_procs() * THREADING_CORES_MULTIPLIER)
     for (iPoint = 0; iPoint < nPoints; ++iPoint) {
-        iPointEnd = iPoint * NCOORDSINPOINT;
-        iPointStart = iPointEnd * isArraysSizeEqual;
+        int iPointEnd = iPoint * NCOORDSINPOINT;
+        int iPointStart = iPointEnd * isArraysSizeEqual;
         mDistance[iPoint] = 2.0 * mRadiusSphere * asin(sqrt((1.0 - cos(rrmEnd[iPointEnd] - rrmStart[iPointStart]) + cos(rrmStart[iPointStart]) * cos(rrmEnd[iPointEnd]) * (1.0 - cos(rrmEnd[iPointEnd + 1] - rrmStart[iPointStart + 1]))) / 2.0));
     }
 }
@@ -53,11 +53,11 @@ void HaversineFloat(const float* rrmStart,
     float mRadiusSphere,
     float* mDistance)
 {
-    int iPoint, iPointEnd, iPointStart;
+    int iPoint;
 #pragma omp parallel for if (nPoints > omp_get_num_procs() * THREADING_CORES_MULTIPLIER)
     for (iPoint = 0; iPoint < nPoints; ++iPoint) {
-        iPointEnd = iPoint * NCOORDSINPOINT;
-        iPointStart = iPointEnd * isArraysSizeEqual;
+        int iPointEnd = iPoint * NCOORDSINPOINT;
+        int iPointStart = iPointEnd * isArraysSizeEqual;
         mDistance[iPoint] = (float)(2.0) * mRadiusSphere * asinf(sqrtf(((float)(1.0) - cosf(rrmEnd[iPointEnd] - rrmStart[iPointStart]) + cosf(rrmStart[iPointStart]) * cosf(rrmEnd[iPointEnd]) * ((float)(1.0) - cosf(rrmEnd[iPointEnd + 1] - rrmStart[iPointStart + 1]))) / (float)(2.0)));
     }
 }

diff --git a/include/helpers.c b/include/helpers.c
@@ -126,10 +126,10 @@ void XXM2YYMDouble(const double* rrmPoint,
     const double transform,
     double* ddmPoint)
 {
-    int iPoint, i;
+    int iPoint;
 #pragma omp parallel for if (nPoints > omp_get_num_procs() * THREADING_CORES_MULTIPLIER)
     for (iPoint = 0; iPoint < nPoints; ++iPoint) {
-        i = iPoint * NCOORDSINPOINT;
+        int i = iPoint * NCOORDSINPOINT;
         ddmPoint[i + 0] = rrmPoint[i + 0] * transform;
         ddmPoint[i + 1] = rrmPoint[i + 1] * transform;
         ddmPoint[i + 2] = rrmPoint[i + 2];
@@ -148,10 +148,10 @@ void XXM2YYMFloat(const float* rrmPoint,
     const float transform,
     float* ddmPoint)
 {
-    int iPoint, i;
+    int iPoint;
 #pragma omp parallel for if (nPoints > omp_get_num_procs() * THREADING_CORES_MULTIPLIER)
     for (iPoint = 0; iPoint < nPoints; ++iPoint) {
-        i = iPoint * NCOORDSINPOINT;
+        int i = iPoint * NCOORDSINPOINT;
         ddmPoint[i + 0] = rrmPoint[i + 0] * transform;
         ddmPoint[i + 1] = rrmPoint[i + 1] * transform;
         ddmPoint[i + 2] = rrmPoint[i + 2];
@@ -167,8 +167,9 @@ void WrapsFloat3(const float* val,
     int nPoints,
     float* boundedVal)
 {
+    int iPoint;
 #pragma omp parallel for if (nPoints > omp_get_num_procs() * THREADING_CORES_MULTIPLIER)
-    for (int iPoint = 0; iPoint < nPoints; ++iPoint) {
+    for (iPoint = 0; iPoint < nPoints; ++iPoint) {
         boundedVal[iPoint] = fmodf(val[iPoint] - minVal[iPoint], maxVal[iPoint] - minVal[iPoint]) + minVal[iPoint];
         if (boundedVal[iPoint] < minVal[iPoint])
             boundedVal[iPoint] += maxVal[iPoint] - minVal[iPoint];
@@ -184,8 +185,9 @@ void WrapsDouble3(const double* val,
     int nPoints,
     double* boundedVal)
 {
+    int iPoint;
 #pragma omp parallel for if (nPoints > omp_get_num_procs() * THREADING_CORES_MULTIPLIER)
-    for (int iPoint = 0; iPoint < nPoints; ++iPoint) {
+    for (iPoint = 0; iPoint < nPoints; ++iPoint) {
         boundedVal[iPoint] = fmod(val[iPoint] - minVal[iPoint], maxVal[iPoint] - minVal[iPoint]) + minVal[iPoint];
         if (boundedVal[iPoint] < minVal[iPoint])
             boundedVal[iPoint] += maxVal[iPoint] - minVal[iPoint];
@@ -201,8 +203,9 @@ void WrapsFloat1(const float* val,
     int nPoints,
     float* boundedVal)
 {
+    int iPoint;
 #pragma omp parallel for if (nPoints > omp_get_num_procs() * THREADING_CORES_MULTIPLIER)
-    for (int iPoint = 0; iPoint < nPoints; ++iPoint) {
+    for (iPoint = 0; iPoint < nPoints; ++iPoint) {
         boundedVal[iPoint] = fmodf(val[iPoint] - minVal, maxVal - minVal) + minVal;
         if (boundedVal[iPoint] < minVal)
             boundedVal[iPoint] += maxVal - minVal;
@@ -218,8 +221,9 @@ void WrapsDouble1(const double* val,
     int nPoints,
     double* boundedVal)
 {
+    int iPoint;
 #pragma omp parallel for if (nPoints > omp_get_num_procs() * THREADING_CORES_MULTIPLIER)
-    for (int iPoint = 0; iPoint < nPoints; ++iPoint) {
+    for (iPoint = 0; iPoint < nPoints; ++iPoint) {
         boundedVal[iPoint] = fmod(val[iPoint] - minVal, maxVal - minVal) + minVal;
         if (boundedVal[iPoint] < minVal)
             boundedVal[iPoint] += maxVal - minVal;