test_loopy.py

    knl = lp.split_iname(knl, "k", 4, inner_tag="unr", slabs=(0, 1))
    knl = lp.set_loop_priority(knl, "I,k_outer,k_inner")

    knl = lp.preprocess_kernel(knl)
    knl = lp.get_one_scheduled_kernel(knl)
    print(lp.generate_body(knl))


def test_precompute_with_preexisting_inames(ctx_factory):
    ctx = ctx_factory()

    knl = lp.make_kernel(
        "{[e,i,j,k]: 0<=e<E and 0<=i,j,k<n}",
        """
        result[e,i] = sum(j, D1[i,j]*u[e,j])
        result2[e,i] = sum(k, D2[i,k]*u[e,k])
        """)

    knl = lp.add_and_infer_dtypes(knl, {
        "u": np.float32,
        "D1": np.float32,
        "D2": np.float32,
        })

    knl = lp.fix_parameters(knl, n=13)

    ref_knl = knl

    knl = lp.extract_subst(knl, "D1_subst", "D1[ii,jj]", parameters="ii,jj")
    knl = lp.extract_subst(knl, "D2_subst", "D2[ii,jj]", parameters="ii,jj")

    knl = lp.precompute(knl, "D1_subst", "i,j", default_tag="for",
            precompute_inames="ii,jj")
    knl = lp.precompute(knl, "D2_subst", "i,k", default_tag="for",
            precompute_inames="ii,jj")

    knl = lp.set_loop_priority(knl, "ii,jj,e,j,k")

    lp.auto_test_vs_ref(
            ref_knl, ctx, knl,
            parameters=dict(E=200))


def test_precompute_with_preexisting_inames_fail():
    knl = lp.make_kernel(
        "{[e,i,j,k]: 0<=e<E and 0<=i,j<n and 0<=k<2*n}",
        """
        result[e,i] = sum(j, D1[i,j]*u[e,j])
        result2[e,i] = sum(k, D2[i,k]*u[e,k])
        """)

    knl = lp.add_and_infer_dtypes(knl, {
        "u": np.float32,
        "D1": np.float32,
        "D2": np.float32,
        })

    knl = lp.fix_parameters(knl, n=13)

    knl = lp.extract_subst(knl, "D1_subst", "D1[ii,jj]", parameters="ii,jj")
    knl = lp.extract_subst(knl, "D2_subst", "D2[ii,jj]", parameters="ii,jj")

    knl = lp.precompute(knl, "D1_subst", "i,j", default_tag="for",
            precompute_inames="ii,jj")
    with pytest.raises(lp.LoopyError):
        lp.precompute(knl, "D2_subst", "i,k", default_tag="for",
                precompute_inames="ii,jj")


def test_vectorize(ctx_factory):
    ctx = ctx_factory()

    knl = lp.make_kernel(
        "{[i]: 0<=i<n}",
        """
        <> temp = 2*b[i]
        a[i] = temp
        """)
    knl = lp.add_and_infer_dtypes(knl, dict(b=np.float32))
    knl = lp.split_arg_axis(knl, [("a", 0), ("b", 0)], 4,
            split_kwargs=dict(slabs=(0, 1)))

    knl = lp.tag_data_axes(knl, "a,b", "c,vec")
    ref_knl = knl
    ref_knl = lp.tag_inames(ref_knl, {"i_inner": "unr"})

    knl = lp.tag_inames(knl, {"i_inner": "vec"})

    knl = lp.preprocess_kernel(knl)
    knl = lp.get_one_scheduled_kernel(knl)
    code, inf = lp.generate_code(knl)

    lp.auto_test_vs_ref(
            ref_knl, ctx, knl,
            parameters=dict(n=30))


def test_alias_temporaries(ctx_factory):
    ctx = ctx_factory()

    knl = lp.make_kernel(
        "{[i]: 0<=i<n}",
        """
        times2(i) := 2*a[i]
        times3(i) := 3*a[i]
        times4(i) := 4*a[i]

        x[i] = times2(i)
        y[i] = times3(i)
        z[i] = times4(i)
        """)

    knl = lp.add_and_infer_dtypes(knl, {"a": np.float32})

    ref_knl = knl

    knl = lp.split_iname(knl, "i", 16, outer_tag="g.0", inner_tag="l.0")

    knl = lp.precompute(knl, "times2", "i_inner")
    knl = lp.precompute(knl, "times3", "i_inner")
    knl = lp.precompute(knl, "times4", "i_inner")

    knl = lp.alias_temporaries(knl, ["times2_0", "times3_0", "times4_0"])

    lp.auto_test_vs_ref(
            ref_knl, ctx, knl,
            parameters=dict(n=30))


if __name__ == "__main__":
    if len(sys.argv) > 1:
        exec(sys.argv[1])
    else:
        from py.test.cmdline import main
        main([__file__])

# vim: foldmethod=marker