Hi,
When making a patch to adjust VECTOR_P8_VECTOR rs6000_vector
enum, I noticed that V1TImode's mode attribute in VI_unit
VECTOR_UNIT_ALTIVEC_P (V1TImode) is never true, since
VECTOR_UNIT_ALTIVEC_P checks if vector_unit[V1TImode] is
equal to VECTOR_ALTIVEC, but vector_unit[V1TImode] can only
be VECTOR_NONE or VECTOR_P8_VECTOR, there is no chance to be
VECTOR_ALTIVEC:
rs6000_vector_unit[V1TImode]
= (TARGET_P8_VECTOR) ? VECTOR_P8_VECTOR : VECTOR_NONE;
By checking all uses of VI_unit, the used mode iterator is
one of VI2, VI, VP_small and VP, none of them has V1TImode,
so the entry for V1TImode is useless. I guessed it was
designed to have one mode attribute to cover all integer
vector modes, but later we separated V1TI handlings to its
own patterns (those guarded with TARGET_VADDUQM). Anyway,
this patch is to remove this useless and confusing entry.
Bootstrapped and regtested on ppc64-linux P8/P9 and
ppc64le-linux P9/P10 two months ago, re-tested (also rebased)
on ppc64le-linux P10 again.
I'm going to push this next week if no objections.
BR,
Kewen
-----
gcc/ChangeLog:
* config/rs6000/altivec.md (mode attr for V1TI in VI_unit): Remove.
---
gcc/config/rs6000/altivec.md | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
diff --git a/gcc/config/rs6000/altivec.md b/gcc/config/rs6000/altivec.md
index 00dad4b91f1..687c3c0ac7e 100644
--- a/gcc/config/rs6000/altivec.md
+++ b/gcc/config/rs6000/altivec.md
@@ -227,8 +227,7 @@ (define_mode_attr VI_scalar [(V2DI "DI") (V4SI "SI") (V8HI
"HI") (V16QI "QI")])
(define_mode_attr VI_unit [(V16QI "VECTOR_UNIT_ALTIVEC_P (V16QImode)")
(V8HI "VECTOR_UNIT_ALTIVEC_P (V8HImode)")
(V4SI "VECTOR_UNIT_ALTIVEC_P (V4SImode)")
- (V2DI "VECTOR_UNIT_P8_VECTOR_P (V2DImode)")
- (V1TI "VECTOR_UNIT_ALTIVEC_P (V1TImode)")])
+ (V2DI "VECTOR_UNIT_P8_VECTOR_P (V2DImode)")])
;; Vector pack/unpack
(define_mode_iterator VP [V2DI V4SI V8HI])
--
2.43.5
