examples/python/vector-types.cl

+#define lid(N) ((int) get_local_id(N))
+#define gid(N) ((int) get_group_id(N))
+#define int_floor_div_pos_b(a,b) (                 ( (a) - ( ((a)<0) ? ((b)-1) : 0 )  ) / (b)                 )
+
+__kernel void __attribute__ ((reqd_work_group_size(1, 1, 1))) loopy_kernel(__global float4 const *__restrict__ a, int const n, __global float4 *__restrict__ out)
+{
+  /* bulk slab for 'i_outer' */
+  for (int i_outer = 0; i_outer <= -2 + int_floor_div_pos_b(3 + n, 4); ++i_outer)
+    out[i_outer] = 2.0f * a[i_outer];
+  /* final slab for 'i_outer' */
+  {
+    int const i_outer = -1 + n + -1 * int_floor_div_pos_b(3 * n, 4);
+
+    if (-1 + n >= 0)
+    {
+      if (-1 + -4 * i_outer + n >= 0)
+        out[i_outer].s0 = 2.0f * a[i_outer].s0;
+      if (-1 + -4 * i_outer + -1 + n >= 0)
+        out[i_outer].s1 = 2.0f * a[i_outer].s1;
+      if (-1 + -4 * i_outer + -1 * 2 + n >= 0)
+        out[i_outer].s2 = 2.0f * a[i_outer].s2;
+      if (-1 + -4 * i_outer + -1 * 3 + n >= 0)
+        out[i_outer].s3 = 2.0f * a[i_outer].s3;
+    }
+  }
+}

examples/python/vector-types.py

+import numpy as np
+
+import pyopencl as cl
+import pyopencl.array
+
+import loopy as lp
+from loopy.version import LOOPY_USE_LANGUAGE_VERSION_2018_2  # noqa: F401
+
+
+ctx = cl.create_some_context()
+queue = cl.CommandQueue(ctx)
+
+n = 15 * 10**6
+a = cl.array.arange(queue, n, dtype=np.float32)
+
+knl = lp.make_kernel(
+        "{ [i]: 0<=i<n }",
+        "out[i] = 2*a[i]")
+
+knl = lp.set_options(knl, write_code=True)
+knl = lp.split_iname(knl, "i", 4, slabs=(0, 1), inner_tag="vec")
+knl = lp.split_array_axis(knl, "a,out", axis_nr=0, count=4)
+knl = lp.tag_array_axes(knl, "a,out", "C,vec")
+
+knl(queue, a=a.reshape(-1, 4), n=n)

loopy/__main__.py

+from __future__ import annotations
+
+import loopy.cli
+
+
+loopy.cli.main()

loopy/frontend/fortran/diagnostic.py

-from __future__ import division, with_statement
+from __future__ import annotations
+
 
 __copyright__ = "Copyright (C) 2009 Andreas Kloeckner"