This implements patterns combining vector element insertion of
vector element extraction to a VEC_PERM_EXPR of both vectors
when supported. Plus it adds the more generic identity transform
of inserting a piece of itself at the same position.
Richard - is there anything I can do to make this SVE aware?
I'd need to construct an identity permute and "insert" into
that permute that element from the other (or same) vector.
I suppose for most element positions that won't work but
at least inserting at [0] should? I'm mostly struggling
on how to use vec_perm_builder here when nelts is not constant,
since it's derived from vec<> can I simply start with
a single pattern with 1 stride and then insert by using []?
Bootstrap / regtest running on x86_64-unknown-linux-gnu.
Thanks,
Richard.
2020-05-07 Richard Biener <rguent...@suse.de>
PR tree-optimization/94865
* match.pd ((bit_insert @0 (BIT_FIELD_REF @0 ...) ...) -> @0):
New simplification.
((bit_insert @0 (BIT_FIELD_REF @1 ...) ...) -> (vec_perm @0 @1 ..):
Likewise.
* gcc.dg/tree-ssa/forwprop-39.c: New testcase.
* gcc.dg/tree-ssa/forwprop-40.c: Likewise.
---
gcc/match.pd | 42 +++++++++++++++++++++++++++++
gcc/testsuite/gcc.dg/tree-ssa/forwprop-39.c | 21 +++++++++++++++
gcc/testsuite/gcc.dg/tree-ssa/forwprop-40.c | 18 +++++++++++++
3 files changed, 81 insertions(+)
create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/forwprop-39.c
create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/forwprop-40.c
diff --git a/gcc/match.pd b/gcc/match.pd
index 9259dd4ddaa..4ce728d78c8 100644
--- a/gcc/match.pd
+++ b/gcc/match.pd
@@ -5838,6 +5838,48 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
wi::to_wide (@ipos) + isize))
(BIT_FIELD_REF @0 @rsize @rpos)))))
+/* Optimize a bit-insertion of a bit-extraction from the same object,
+ position and size. */
+(simplify
+ (bit_insert @0 (BIT_FIELD_REF@1 @0 @size @pos) @pos)
+ @0)
+
+/* Optimize an element insertion into a vector that is extracted from
+ another vector to a permutation of both vectors. */
+(simplify
+ (bit_insert @0 (BIT_FIELD_REF@1 @2 @size @rpos) @ipos)
+ (if (VECTOR_TYPE_P (type)
+ && types_match (type, TREE_TYPE (@2))
+ && single_use (@1))
+ (with
+ {
+ unsigned HOST_WIDE_INT nelts;
+ unsigned int elem, at, elemsz;
+ }
+ (if (TYPE_VECTOR_SUBPARTS (type).is_constant (&nelts)
+ && tree_fits_uhwi_p (@size)
+ && constant_multiple_p (tree_to_poly_uint64 (@rpos),
+ tree_to_poly_uint64 (@size), &elem)
+ && constant_multiple_p (tree_to_poly_uint64 (@ipos),
+ tree_to_poly_uint64 (@size), &at)
+ && constant_multiple_p (tree_to_poly_uint64 (@size),
+ tree_to_poly_uint64
+ (TYPE_SIZE (TREE_TYPE (type))), &elemsz))
+ (with
+ {
+ vec_perm_builder sel (nelts, nelts, 1);
+ for (unsigned i = 0; i < nelts; ++i)
+ sel.quick_push (i / elemsz == at
+ ? nelts + elem * elemsz + i % elemsz : i);
+ vec_perm_indices indices (sel, @0 == @2 ? 1 : 2, nelts);
+ }
+ (if (can_vec_perm_const_p (TYPE_MODE (TREE_TYPE (@0)), indices))
+ (vec_perm @0 @2 { vec_perm_indices_to_tree
+ (build_vector_type
+ (build_nonstandard_integer_type
+ (tree_to_uhwi (@size), 1),
+ nelts), indices); })))))))
+
(if (canonicalize_math_after_vectorization_p ())
(for fmas (FMA)
(simplify
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/forwprop-39.c
b/gcc/testsuite/gcc.dg/tree-ssa/forwprop-39.c
new file mode 100644
index 00000000000..f0212c373e2
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/forwprop-39.c
@@ -0,0 +1,21 @@
+/* { dg-do compile } */
+/* { dg-options "-O -fdump-tree-forwprop1 -fdump-tree-cddce1 -Wno-psabi -w" }
*/
+
+typedef double v2df __attribute__((vector_size(16)));
+
+v2df move_sd(v2df a, v2df b)
+{
+ v2df result = a;
+ result[1] = b[1];
+ return result;
+}
+
+v2df move_nnop(v2df a)
+{
+ v2df result = a;
+ result[1] = a[1];
+ return result;
+}
+
+/* { dg-final { scan-tree-dump-times "VEC_PERM_EXPR" 1 "forwprop1" } } */
+/* { dg-final { scan-tree-dump-times "BIT_FIELD_REF" 0 "cddce1" } } */
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/forwprop-40.c
b/gcc/testsuite/gcc.dg/tree-ssa/forwprop-40.c
new file mode 100644
index 00000000000..94329437f0d
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/forwprop-40.c
@@ -0,0 +1,18 @@
+/* { dg-do compile } */
+/* { dg-options "-O -fgimple -fdump-tree-forwprop1 -w -Wno-psabi" } */
+
+typedef int v4si __attribute__((vector_size(16)));
+typedef int v2si __attribute__((vector_size(8)));
+
+v4si __GIMPLE(ssa) bar (v4si a)
+{
+ v2si el;
+ v4si res;
+
+ __BB(2):
+ el_2 = __BIT_FIELD_REF <v2si> (a_1(D), 64u, 64u);
+ res_3 = __BIT_INSERT (a_1(D), el_2, 0u);
+ return res_3;
+}
+
+/* { dg-final { scan-tree-dump "VEC_PERM" "forwprop1" } } */
--
2.13.7
