public inbox for gcc-cvs@sourceware.org
help / color / mirror / Atom feed
* [gcc r12-3141] Make xxsplti*, xpermx, xxeval be vecperm type.
@ 2021-08-25 19:12 Michael Meissner
0 siblings, 0 replies; only message in thread
From: Michael Meissner @ 2021-08-25 19:12 UTC (permalink / raw)
To: gcc-cvs
https://gcc.gnu.org/g:bb24717e5042b6e8a3847e780a8d215edb9c62f6
commit r12-3141-gbb24717e5042b6e8a3847e780a8d215edb9c62f6
Author: Michael Meissner <meissner@linux.ibm.com>
Date: Wed Aug 25 15:11:47 2021 -0400
Make xxsplti*, xpermx, xxeval be vecperm type.
I noticed that the built-functions for xxspltiw, xxspltidp, xxsplti32dx,
xxpermx, and xxeval all used the 'vecsimple' type. These instructions are
permute instructions (3 cycle latency) and should use 'vecperm' instead.
While I was at it, I changed the UNSPEC name for xxspltidp to be
UNSPEC_XXSPLTIDP instead of UNSPEC_XXSPLTID.
2021-08-25 Michael Meissner <meissner@linux.ibm.com>
gcc/
* config/rs6000/vsx.md (UNSPEC_XXSPLTIDP): Rename from
UNSPEC_XXSPLTID.
(xxspltiw_v4si): Use vecperm type attribute.
(xxspltiw_v4si_inst): Use vecperm type attribute.
(xxspltiw_v4sf_inst): Likewise.
(xxspltidp_v2df): Use vecperm type attribute. Use
UNSPEC_XXSPLTIDP instead of UNSPEC_XXSPLTID.
(xxspltidp_v2df_inst): Likewise.
(xxsplti32dx_v4si): Use vecperm type attribute.
(xxsplti32dx_v4si_inst): Likewise.
(xxsplti32dx_v4sf_inst): Likewise.
(xxblend_<mode>): Likewise.
(xxpermx): Likewise.
(xxpermx_inst): Likewise.
(xxeval): Likewise.
Diff:
---
gcc/config/rs6000/vsx.md | 26 +++++++++++++-------------
1 file changed, 13 insertions(+), 13 deletions(-)
diff --git a/gcc/config/rs6000/vsx.md b/gcc/config/rs6000/vsx.md
index e4ca6e94d49..bf033e31c1c 100644
--- a/gcc/config/rs6000/vsx.md
+++ b/gcc/config/rs6000/vsx.md
@@ -374,7 +374,7 @@
UNSPEC_VDIVEU
UNSPEC_XXEVAL
UNSPEC_XXSPLTIW
- UNSPEC_XXSPLTID
+ UNSPEC_XXSPLTIDP
UNSPEC_XXSPLTI32DX
UNSPEC_XXBLEND
UNSPEC_XXPERMX
@@ -6414,7 +6414,7 @@
UNSPEC_XXSPLTIW))]
"TARGET_POWER10"
"xxspltiw %x0,%1"
- [(set_attr "type" "vecsimple")
+ [(set_attr "type" "vecperm")
(set_attr "prefixed" "yes")])
(define_expand "xxspltiw_v4sf"
@@ -6434,14 +6434,14 @@
UNSPEC_XXSPLTIW))]
"TARGET_POWER10"
"xxspltiw %x0,%1"
- [(set_attr "type" "vecsimple")
+ [(set_attr "type" "vecperm")
(set_attr "prefixed" "yes")])
;; XXSPLTIDP built-in function support
(define_expand "xxspltidp_v2df"
[(set (match_operand:V2DF 0 "register_operand" )
(unspec:V2DF [(match_operand:SF 1 "const_double_operand")]
- UNSPEC_XXSPLTID))]
+ UNSPEC_XXSPLTIDP))]
"TARGET_POWER10"
{
long value = rs6000_const_f32_to_i32 (operands[1]);
@@ -6452,10 +6452,10 @@
(define_insn "xxspltidp_v2df_inst"
[(set (match_operand:V2DF 0 "register_operand" "=wa")
(unspec:V2DF [(match_operand:SI 1 "c32bit_cint_operand" "n")]
- UNSPEC_XXSPLTID))]
+ UNSPEC_XXSPLTIDP))]
"TARGET_POWER10"
"xxspltidp %x0,%1"
- [(set_attr "type" "vecsimple")
+ [(set_attr "type" "vecperm")
(set_attr "prefixed" "yes")])
;; XXSPLTI32DX built-in function support
@@ -6476,7 +6476,7 @@
GEN_INT (index), operands[3]));
DONE;
}
- [(set_attr "type" "vecsimple")])
+ [(set_attr "type" "vecperm")])
(define_insn "xxsplti32dx_v4si_inst"
[(set (match_operand:V4SI 0 "register_operand" "=wa")
@@ -6486,7 +6486,7 @@
UNSPEC_XXSPLTI32DX))]
"TARGET_POWER10"
"xxsplti32dx %x0,%2,%3"
- [(set_attr "type" "vecsimple")
+ [(set_attr "type" "vecperm")
(set_attr "prefixed" "yes")])
(define_expand "xxsplti32dx_v4sf"
@@ -6515,7 +6515,7 @@
UNSPEC_XXSPLTI32DX))]
"TARGET_POWER10"
"xxsplti32dx %x0,%2,%3"
- [(set_attr "type" "vecsimple")
+ [(set_attr "type" "vecperm")
(set_attr "prefixed" "yes")])
;; XXBLEND built-in function support
@@ -6527,7 +6527,7 @@
UNSPEC_XXBLEND))]
"TARGET_POWER10"
"xxblendv<VM3_char> %x0,%x1,%x2,%x3"
- [(set_attr "type" "vecsimple")
+ [(set_attr "type" "vecperm")
(set_attr "prefixed" "yes")])
;; XXPERMX built-in function support
@@ -6562,7 +6562,7 @@
DONE;
}
- [(set_attr "type" "vecsimple")])
+ [(set_attr "type" "vecperm")])
(define_insn "xxpermx_inst"
[(set (match_operand:V2DI 0 "register_operand" "+v")
@@ -6573,7 +6573,7 @@
UNSPEC_XXPERMX))]
"TARGET_POWER10"
"xxpermx %x0,%x1,%x2,%x3,%4"
- [(set_attr "type" "vecsimple")
+ [(set_attr "type" "vecperm")
(set_attr "prefixed" "yes")])
;; XXEVAL built-in function support
@@ -6586,6 +6586,6 @@
UNSPEC_XXEVAL))]
"TARGET_POWER10"
"xxeval %0,%1,%2,%3,%4"
- [(set_attr "type" "vecsimple")
+ [(set_attr "type" "vecperm")
(set_attr "prefixed" "yes")])
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2021-08-25 19:12 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-08-25 19:12 [gcc r12-3141] Make xxsplti*, xpermx, xxeval be vecperm type Michael Meissner
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).