diff options
| -rw-r--r-- | numpy/core/src/_simd/_simd.dispatch.c.src | 26 | ||||
| -rw-r--r-- | numpy/core/tests/test_simd.py | 21 |
2 files changed, 44 insertions, 3 deletions
diff --git a/numpy/core/src/_simd/_simd.dispatch.c.src b/numpy/core/src/_simd/_simd.dispatch.c.src index fabec069c..ab48db5b1 100644 --- a/numpy/core/src/_simd/_simd.dispatch.c.src +++ b/numpy/core/src/_simd/_simd.dispatch.c.src @@ -15,7 +15,8 @@ /**begin repeat * #sfx = u8, s8, u16, s16, u32, s32, u64, s64, f32, f64# * #bsfx = b8, b8, b16, b16, b32, b32, b64, b64, b32, b64# - * #esfx = u16, s8, u32,s16, u32, s32, u64, s64, f32, f64# + * #esfx = u16,s8, u32, s16, u32, s32, u64, s64, f32, f64# + * #size = 8, 8, 16, 16, 32, 32, 64, 64, 32, 64# * #expand_sup= 1, 0, 1, 0, 0, 0, 0, 0, 0, 0# * #simd_sup = 1, 1, 1, 1, 1, 1, 1, 1, 1, NPY_SIMD_F64# * #fp_only = 0, 0, 0, 0, 0, 0, 0, 0, 1, 1# @@ -232,6 +233,15 @@ err: /**end repeat1**/ #endif // @ncont_sup@ +/**************************** + * Lookup tables + ****************************/ +#if @size@ == 32 +SIMD_IMPL_INTRIN_2(lut32_@sfx@, v@sfx@, q@sfx@, vu@size@) +#endif +#if @size@ == 64 +SIMD_IMPL_INTRIN_2(lut16_@sfx@, v@sfx@, q@sfx@, vu@size@) +#endif /*************************** * Misc ***************************/ @@ -470,8 +480,9 @@ static PyMethodDef simd__intrinsics_methods[] = { /**begin repeat * #sfx = u8, s8, u16, s16, u32, s32, u64, s64, f32, f64# * #bsfx = b8, b8, b16, b16, b32, b32, b64, b64, b32, b64# - * #esfx = u16, s8, u32,s16, u32, s32, u64, s64, f32, f64# - * #expand_sup =1, 0, 1, 0, 0, 0, 0, 0, 0, 0# + * #esfx = u16,s8, u32, s16, u32, s32, u64, s64, f32, f64# + * #size = 8, 8, 16, 16, 32, 32, 64, 64, 32, 64# + * #expand_sup= 1, 0, 1, 0, 0, 0, 0, 0, 0, 0# * #simd_sup = 1, 1, 1, 1, 1, 1, 1, 1, 1, NPY_SIMD_F64# * #fp_only = 0, 0, 0, 0, 0, 0, 0, 0, 1, 1# * #sat_sup = 1, 1, 1, 1, 0, 0, 0, 0, 0, 0# @@ -509,6 +520,15 @@ SIMD_INTRIN_DEF(@intrin@_@sfx@) /**end repeat1**/ #endif // ncont_sup +/**************************** + * Lookup tables + ****************************/ +#if @size@ == 32 +SIMD_INTRIN_DEF(lut32_@sfx@) +#endif +#if @size@ == 64 +SIMD_INTRIN_DEF(lut16_@sfx@) +#endif /*************************** * Misc ***************************/ diff --git a/numpy/core/tests/test_simd.py b/numpy/core/tests/test_simd.py index 605baefe6..e4b5e0c8f 100644 --- a/numpy/core/tests/test_simd.py +++ b/numpy/core/tests/test_simd.py @@ -627,6 +627,27 @@ class _SIMD_ALL(_Test_Utility): assert storen_till[64:] == data_till assert storen_till[:64] == [127]*64 # detect overflow + @pytest.mark.parametrize("intrin, table_size, elsize", [ + ("self.lut32", 32, 32), + ("self.lut16", 16, 64) + ]) + def test_lut(self, intrin, table_size, elsize): + """ + Test lookup table intrinsics: + npyv_lut32_##sfx + npyv_lut16_##sfx + """ + if elsize != self._scalar_size(): + return + intrin = eval(intrin) + idx_itrin = getattr(self.npyv, f"setall_u{elsize}") + table = range(0, table_size) + for i in table: + broadi = self.setall(i) + idx = idx_itrin(i) + lut = intrin(table, idx) + assert lut == broadi + def test_misc(self): broadcast_zero = self.zero() assert broadcast_zero == [0] * self.nlanes |