small speed improvement by changing the block size

1 files changed, 1 insertions, 1 deletions

diff --git a/test.py b/test.py
index ec7c954..1c9761d 100644
--- a/test.py
+++ b/test.py
@@ -71,7 +71,7 @@ for i in range(100):
     rotate(np.int32(p.size), p_gpu, np.float32(np.pi/100), gpuarray.vec.make_float3(0,0,1), block=(256,1,1), grid=(width*height//256+1,1))
 
     t0 = time.time()
-    intersect(np.int32(x.size), x_gpu, p_gpu, np.int32(mesh.size//3), mesh_gpu, pixel_gpu, block=(256,1,1), grid=(width*height//256+1,1))
+    intersect(np.int32(x.size), x_gpu, p_gpu, np.int32(mesh.size//3), mesh_gpu, pixel_gpu, block=(128,1,1), grid=(width*height//128+1,1))
     cuda.Context.synchronize()
 
     elapsed = time.time() - t0