Fixes in tests

Author: szymonlopaciuk

tests/test_capi.py

@@ -659,7 +659,11 @@ class Cells(xo.Struct):
     int MAX_CELLS = 3;
 
     GPUKERN
-    uint8_t loop_over(Cells cells, uint64_t* out_counts, uint64_t* out_vals)
+    uint8_t loop_over(
+        Cells cells,
+        GPUGLMEM uint64_t* out_counts,
+        GPUGLMEM uint64_t* out_vals
+    )
     {
         uint8_t success = 1;
         int64_t num_cells = Cells_len_ids(cells);
@@ -668,8 +672,6 @@ class Cells(xo.Struct):
             int64_t id = Cells_get_ids(cells, i);
             int64_t count = Cells_len1_particles(cells, i);
 
-            printf("Cell ID: %lld\n Particles (count %lld): ", id, count);
-
             if (i >= MAX_CELLS) {
                 success = 0;
                 continue;
@@ -682,7 +684,6 @@ class Cells(xo.Struct):
 
             VECTORIZE_OVER(j, num_particles);
                 int64_t val = ArrNInt64_get(particles, j);
-                printf("%lld ", val);
 
                 if (j >= MAX_PARTICLES) {
                     success = 0;
@@ -691,7 +692,6 @@ class Cells(xo.Struct):
 
                 out_vals[i * MAX_PARTICLES + j] = val;
             END_VECTORIZE;
-            printf("\n");
         }
         fflush(stdout);
         return success;

tests/test_common.py

@@ -16,8 +16,7 @@ def test_common_atomicadd(test_context):
     #include "xobjects/headers/common.h"
     #include "xobjects/headers/atomicadd.h"
 
-    GPUKERN
-    double test_atomic_add()
+    GPUKERN void test_atomic_add(GPUGLMEM double* out)
     {
         int iterations = 1000;
         double sum = 0;
@@ -28,7 +27,7 @@ def test_common_atomicadd(test_context):
             #endif
             atomicAdd(&sum, 1.0);
         END_VECTORIZE;
-        return sum;
+        *out = sum;
     }
     """
 
@@ -45,15 +44,15 @@ def test_common_atomicadd(test_context):
         sources=[src],
         kernels={
             "test_atomic_add": xo.Kernel(
-                args=[],
+                args=[xo.Arg(xo.Float64, pointer=True, name="out")],
                 n_threads=n_threads,
-                ret=xo.Arg(xo.Float64),
             )
         },
     )
 
     expected = 1000
-    result = test_context.kernels.test_atomic_add()
+    result = np.array([0], dtype=np.float64)
+    test_context.kernels.test_atomic_add(out=result)
 
     assert result == expected