diff options
author | Kyrylo Tkachov <kyrylo.tkachov@arm.com> | 2019-02-07 09:31:33 +0000 |
---|---|---|
committer | Kyrylo Tkachov <ktkachov@gcc.gnu.org> | 2019-02-07 09:31:33 +0000 |
commit | 3ed8fcfa3a82b0986b4c2090a3ea4b5fdfddeb3b (patch) | |
tree | 73c2608dc54faf47441cc1e91cdaa8edacc4aca4 | |
parent | dc51dece7ae902317e9131e153ef53d7bc09c564 (diff) | |
download | gcc-3ed8fcfa3a82b0986b4c2090a3ea4b5fdfddeb3b.zip gcc-3ed8fcfa3a82b0986b4c2090a3ea4b5fdfddeb3b.tar.gz gcc-3ed8fcfa3a82b0986b4c2090a3ea4b5fdfddeb3b.tar.bz2 |
[AArch64] Use neon_dot_q type for 128-bit [US]DOT instructions where appropriate
For the Dot Product instructions we have the scheduling types neon_dot and neon_dot_q for the 128-bit versions.
It seems that we're only using the former though, not assigning the neon_dot_q type anywhere.
This patch fixes that by adding the <q> mode attribute suffix to the type, similar to how we do it for other
types in aarch64-simd.md.
* config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>):
Use neon_dot<q> for type.
(aarch64_<sur>dot_lane<vsi2qi>): Likewise.
(aarch64_<sur>dot_laneq<vsi2qi>): Likewise.
From-SVN: r268610
-rw-r--r-- | gcc/ChangeLog | 7 | ||||
-rw-r--r-- | gcc/config/aarch64/aarch64-simd.md | 6 |
2 files changed, 10 insertions, 3 deletions
diff --git a/gcc/ChangeLog b/gcc/ChangeLog index 3b8a343..6d2067a 100644 --- a/gcc/ChangeLog +++ b/gcc/ChangeLog @@ -1,3 +1,10 @@ +2019-02-07 Kyrylo Tkachov <kyrylo.tkachov@arm.com> + + * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): + Use neon_dot<q> for type. + (aarch64_<sur>dot_lane<vsi2qi>): Likewise. + (aarch64_<sur>dot_laneq<vsi2qi>): Likewise. + 2019-02-06 Vladimir Makarov <vmakarov@redhat.com> PR rtl-optimization/89225 diff --git a/gcc/config/aarch64/aarch64-simd.md b/gcc/config/aarch64/aarch64-simd.md index 805d7a8..cae9a8f 100644 --- a/gcc/config/aarch64/aarch64-simd.md +++ b/gcc/config/aarch64/aarch64-simd.md @@ -503,7 +503,7 @@ DOTPROD)))] "TARGET_DOTPROD" "<sur>dot\\t%0.<Vtype>, %2.<Vdottype>, %3.<Vdottype>" - [(set_attr "type" "neon_dot")] + [(set_attr "type" "neon_dot<q>")] ) ;; These expands map to the Dot Product optab the vectorizer checks for. @@ -555,7 +555,7 @@ operands[4] = aarch64_endian_lane_rtx (V8QImode, INTVAL (operands[4])); return "<sur>dot\\t%0.<Vtype>, %2.<Vdottype>, %3.4b[%4]"; } - [(set_attr "type" "neon_dot")] + [(set_attr "type" "neon_dot<q>")] ) (define_insn "aarch64_<sur>dot_laneq<vsi2qi>" @@ -570,7 +570,7 @@ operands[4] = aarch64_endian_lane_rtx (V16QImode, INTVAL (operands[4])); return "<sur>dot\\t%0.<Vtype>, %2.<Vdottype>, %3.4b[%4]"; } - [(set_attr "type" "neon_dot")] + [(set_attr "type" "neon_dot<q>")] ) (define_expand "copysign<mode>3" |