Parallelize AEILocalInterp with OpenMP

This leads to a slight change in behaviour. Currently, AEILocalInterp
traverses the list of points sequentially, and aborts when the first
error is encountered. After parallelisation, there is no fixed order
in which the points are traversed, and if several errors are
encountered, any one of the errors may be returned, not necessarily
the first. I am not aware of any thorn that would or should rely on
such an ordering.

This patch also adds "restrict" and "const" statements that may
improve performance as it gives the compiler more information about
dependencies between pointers.


git-svn-id: http://svn.aei.mpg.de/numrel/AEIThorns/AEILocalInterp/trunk@55 0f49ee68-0e4f-0410-9b9c-b2c123ded7ef

1 files changed, 32 insertions, 32 deletions

diff --git a/src/common/store.c b/src/common/store.c
index 3ed6003..1d414fe 100644
--- a/src/common/store.c
+++ b/src/common/store.c
@@ -17,48 +17,48 @@
 #define COEFF(mi)	Jacobian_ptr[Jacobian_mi_stride*mi]
 
 void AEILocalInterp_store_1dcube2
-      (fp factor, const struct coeffs_struct_1d_cube_size2 *coeffs,
-       fp Jacobian_ptr[],
+      (fp factor, const struct coeffs_struct_1d_cube_size2 *restrict const coeffs,
+       fp *restrict Jacobian_ptr,
        int Jacobian_mi_stride)
 {
 #include "1d.cube.size2/store-coeffs.c"
 }
 
 void AEILocalInterp_store_1dcube3
-      (fp factor, const struct coeffs_struct_1d_cube_size3 *coeffs,
-       fp Jacobian_ptr[],
+      (fp factor, const struct coeffs_struct_1d_cube_size3 *restrict const coeffs,
+       fp *restrict Jacobian_ptr,
        int Jacobian_mi_stride)
 {
 #include "1d.cube.size3/store-coeffs.c"
 }
 
 void AEILocalInterp_store_1dcube4
-      (fp factor, const struct coeffs_struct_1d_cube_size4 *coeffs,
-       fp Jacobian_ptr[],
+      (fp factor, const struct coeffs_struct_1d_cube_size4 *restrict const coeffs,
+       fp *restrict Jacobian_ptr,
        int Jacobian_mi_stride)
 {
 #include "1d.cube.size4/store-coeffs.c"
 }
 
 void AEILocalInterp_store_1dcube5
-      (fp factor, const struct coeffs_struct_1d_cube_size5 *coeffs,
-       fp Jacobian_ptr[],
+      (fp factor, const struct coeffs_struct_1d_cube_size5 *restrict const coeffs,
+       fp *restrict Jacobian_ptr,
        int Jacobian_mi_stride)
 {
 #include "1d.cube.size5/store-coeffs.c"
 }
 
 void AEILocalInterp_store_1dcube6
-      (fp factor, const struct coeffs_struct_1d_cube_size6 *coeffs,
-       fp Jacobian_ptr[],
+      (fp factor, const struct coeffs_struct_1d_cube_size6 *restrict const coeffs,
+       fp *restrict Jacobian_ptr,
        int Jacobian_mi_stride)
 {
 #include "1d.cube.size6/store-coeffs.c"
 }
 
 void AEILocalInterp_store_1dcube7
-      (fp factor, const struct coeffs_struct_1d_cube_size7 *coeffs,
-       fp Jacobian_ptr[],
+      (fp factor, const struct coeffs_struct_1d_cube_size7 *restrict const coeffs,
+       fp *restrict Jacobian_ptr,
        int Jacobian_mi_stride)
 {
 #include "1d.cube.size7/store-coeffs.c"
@@ -75,40 +75,40 @@ void AEILocalInterp_store_1dcube7
 				      + Jacobian_mj_stride*mj ]
 
 void AEILocalInterp_store_2dcube2
-      (fp factor, const struct coeffs_struct_2d_cube_size2 *coeffs,
-       fp Jacobian_ptr[],
+      (fp factor, const struct coeffs_struct_2d_cube_size2 *restrict const coeffs,
+       fp *restrict Jacobian_ptr,
        int Jacobian_mi_stride, int Jacobian_mj_stride)
 {
 #include "2d.cube.size2/store-coeffs.c"
 }
 
 void AEILocalInterp_store_2dcube3
-      (fp factor, const struct coeffs_struct_2d_cube_size3 *coeffs,
-       fp Jacobian_ptr[],
+      (fp factor, const struct coeffs_struct_2d_cube_size3 *restrict const coeffs,
+       fp *restrict Jacobian_ptr,
        int Jacobian_mi_stride, int Jacobian_mj_stride)
 {
 #include "2d.cube.size3/store-coeffs.c"
 }
 
 void AEILocalInterp_store_2dcube4
-      (fp factor, const struct coeffs_struct_2d_cube_size4 *coeffs,
-       fp Jacobian_ptr[],
+      (fp factor, const struct coeffs_struct_2d_cube_size4 *restrict const coeffs,
+       fp *restrict Jacobian_ptr,
        int Jacobian_mi_stride, int Jacobian_mj_stride)
 {
 #include "2d.cube.size4/store-coeffs.c"
 }
 
 void AEILocalInterp_store_2dcube5
-      (fp factor, const struct coeffs_struct_2d_cube_size5 *coeffs,
-       fp Jacobian_ptr[],
+      (fp factor, const struct coeffs_struct_2d_cube_size5 *restrict const coeffs,
+       fp *restrict Jacobian_ptr,
        int Jacobian_mi_stride, int Jacobian_mj_stride)
 {
 #include "2d.cube.size5/store-coeffs.c"
 }
 
 void AEILocalInterp_store_2dcube6
-      (fp factor, const struct coeffs_struct_2d_cube_size6 *coeffs,
-       fp Jacobian_ptr[],
+      (fp factor, const struct coeffs_struct_2d_cube_size6 *restrict const coeffs,
+       fp *restrict Jacobian_ptr,
        int Jacobian_mi_stride, int Jacobian_mj_stride)
 {
 #include "2d.cube.size6/store-coeffs.c"
@@ -126,40 +126,40 @@ void AEILocalInterp_store_2dcube6
 				      + Jacobian_mk_stride*mk ]
 
 void AEILocalInterp_store_3dcube2
-      (fp factor, const struct coeffs_struct_3d_cube_size2 *coeffs,
-       fp Jacobian_ptr[],
+      (fp factor, const struct coeffs_struct_3d_cube_size2 *restrict const coeffs,
+       fp *restrict Jacobian_ptr,
        int Jacobian_mi_stride, int Jacobian_mj_stride, int Jacobian_mk_stride)
 {
 #include "3d.cube.size2/store-coeffs.c"
 }
 
 void AEILocalInterp_store_3dcube3
-      (fp factor, const struct coeffs_struct_3d_cube_size3 *coeffs,
-       fp Jacobian_ptr[],
+      (fp factor, const struct coeffs_struct_3d_cube_size3 *restrict const coeffs,
+       fp *restrict Jacobian_ptr,
        int Jacobian_mi_stride, int Jacobian_mj_stride, int Jacobian_mk_stride)
 {
 #include "3d.cube.size3/store-coeffs.c"
 }
 
 void AEILocalInterp_store_3dcube4
-      (fp factor, const struct coeffs_struct_3d_cube_size4 *coeffs,
-       fp Jacobian_ptr[],
+      (fp factor, const struct coeffs_struct_3d_cube_size4 *restrict const coeffs,
+       fp *restrict Jacobian_ptr,
        int Jacobian_mi_stride, int Jacobian_mj_stride, int Jacobian_mk_stride)
 {
 #include "3d.cube.size4/store-coeffs.c"
 }
 
 void AEILocalInterp_store_3dcube5
-      (fp factor, const struct coeffs_struct_3d_cube_size5 *coeffs,
-       fp Jacobian_ptr[],
+      (fp factor, const struct coeffs_struct_3d_cube_size5 *restrict const coeffs,
+       fp *restrict Jacobian_ptr,
        int Jacobian_mi_stride, int Jacobian_mj_stride, int Jacobian_mk_stride)
 {
 #include "3d.cube.size5/store-coeffs.c"
 }
 
 void AEILocalInterp_store_3dcube6
-      (fp factor, const struct coeffs_struct_3d_cube_size6 *coeffs,
-       fp Jacobian_ptr[],
+      (fp factor, const struct coeffs_struct_3d_cube_size6 *restrict const coeffs,
+       fp *restrict Jacobian_ptr,
        int Jacobian_mi_stride, int Jacobian_mj_stride, int Jacobian_mk_stride)
 {
 #include "3d.cube.size6/store-coeffs.c"