* [PATCH][AArch64] Update patterns to support FP zero
@ 2015-10-08 13:16 Wilco Dijkstra
2015-10-13 7:56 ` James Greenhalgh
0 siblings, 1 reply; 2+ messages in thread
From: Wilco Dijkstra @ 2015-10-08 13:16 UTC (permalink / raw)
To: GCC Patches
This patch improves support for instructions that allow FP zero immediate. All FP compares generated
by various patterns should use aarch64_fp_compare_operand. LDP/STP uses aarch64_reg_or_fp_zero.
Passes regression on AArch64.
OK for commit?
ChangeLog:
2015-10-08 Wilco Dijkstra <wdijkstr@arm.com>
* gcc/config/aarch64/aarch64.md (cbranch<mode>4):
Use aarch64_fp_compare_operand.
(store_pairsf): Use aarch64_reg_or_fp_zero.
(store_pairdf): Likewise.
(cstore<mode>4): Use aarch64_fp_compare_operand.
(cmov<mode>6): Likewise.
* gcc/config/aarch64/aarch64-ldpstp.md: Use aarch64_reg_or_fp_zero.
---
gcc/config/aarch64/aarch64-ldpstp.md | 12 ++++++------
gcc/config/aarch64/aarch64.md | 14 +++++++-------
2 files changed, 13 insertions(+), 13 deletions(-)
diff --git a/gcc/config/aarch64/aarch64-ldpstp.md b/gcc/config/aarch64/aarch64-ldpstp.md
index 8d6d882..54cf34c 100644
--- a/gcc/config/aarch64/aarch64-ldpstp.md
+++ b/gcc/config/aarch64/aarch64-ldpstp.md
@@ -80,9 +80,9 @@
(define_peephole2
[(set (match_operand:GPF 0 "aarch64_mem_pair_operand" "")
- (match_operand:GPF 1 "register_operand" ""))
+ (match_operand:GPF 1 "aarch64_reg_or_fp_zero" ""))
(set (match_operand:GPF 2 "memory_operand" "")
- (match_operand:GPF 3 "register_operand" ""))]
+ (match_operand:GPF 3 "aarch64_reg_or_fp_zero" ""))]
"aarch64_operands_ok_for_ldpstp (operands, false, <MODE>mode)"
[(parallel [(set (match_dup 0) (match_dup 1))
(set (match_dup 2) (match_dup 3))])]
@@ -308,13 +308,13 @@
(define_peephole2
[(match_scratch:DI 8 "r")
(set (match_operand:GPF 0 "memory_operand" "")
- (match_operand:GPF 1 "aarch64_reg_or_zero" ""))
+ (match_operand:GPF 1 "aarch64_reg_or_fp_zero" ""))
(set (match_operand:GPF 2 "memory_operand" "")
- (match_operand:GPF 3 "aarch64_reg_or_zero" ""))
+ (match_operand:GPF 3 "aarch64_reg_or_fp_zero" ""))
(set (match_operand:GPF 4 "memory_operand" "")
- (match_operand:GPF 5 "aarch64_reg_or_zero" ""))
+ (match_operand:GPF 5 "aarch64_reg_or_fp_zero" ""))
(set (match_operand:GPF 6 "memory_operand" "")
- (match_operand:GPF 7 "aarch64_reg_or_zero" ""))
+ (match_operand:GPF 7 "aarch64_reg_or_fp_zero" ""))
(match_dup 8)]
"aarch64_operands_adjust_ok_for_ldpstp (operands, false, <MODE>mode)"
[(const_int 0)]
diff --git a/gcc/config/aarch64/aarch64.md b/gcc/config/aarch64/aarch64.md
index f2d1be1..67ce01b 100644
--- a/gcc/config/aarch64/aarch64.md
+++ b/gcc/config/aarch64/aarch64.md
@@ -240,7 +240,7 @@
(define_expand "cbranch<mode>4"
[(set (pc) (if_then_else (match_operator 0 "aarch64_comparison_operator"
[(match_operand:GPF 1 "register_operand" "")
- (match_operand:GPF 2 "aarch64_reg_or_zero" "")])
+ (match_operand:GPF 2 "aarch64_fp_compare_operand" "")])
(label_ref (match_operand 3 "" ""))
(pc)))]
""
@@ -1336,9 +1336,9 @@
;; fairly lax checking on the second memory operation.
(define_insn "store_pairsf"
[(set (match_operand:SF 0 "aarch64_mem_pair_operand" "=Ump,Ump")
- (match_operand:SF 1 "register_operand" "w,*r"))
+ (match_operand:SF 1 "aarch64_reg_or_fp_zero" "w,*rY"))
(set (match_operand:SF 2 "memory_operand" "=m,m")
- (match_operand:SF 3 "register_operand" "w,*r"))]
+ (match_operand:SF 3 "aarch64_reg_or_fp_zero" "w,*rY"))]
"rtx_equal_p (XEXP (operands[2], 0),
plus_constant (Pmode,
XEXP (operands[0], 0),
@@ -1352,9 +1352,9 @@
(define_insn "store_pairdf"
[(set (match_operand:DF 0 "aarch64_mem_pair_operand" "=Ump,Ump")
- (match_operand:DF 1 "register_operand" "w,*r"))
+ (match_operand:DF 1 "aarch64_reg_or_fp_zero" "w,*rY"))
(set (match_operand:DF 2 "memory_operand" "=m,m")
- (match_operand:DF 3 "register_operand" "w,*r"))]
+ (match_operand:DF 3 "aarch64_reg_or_fp_zero" "w,*rY"))]
"rtx_equal_p (XEXP (operands[2], 0),
plus_constant (Pmode,
XEXP (operands[0], 0),
@@ -2901,7 +2901,7 @@
[(set (match_operand:SI 0 "register_operand" "")
(match_operator:SI 1 "aarch64_comparison_operator"
[(match_operand:GPF 2 "register_operand" "")
- (match_operand:GPF 3 "register_operand" "")]))]
+ (match_operand:GPF 3 "aarch64_fp_compare_operand" "")]))]
""
"
operands[2] = aarch64_gen_compare_reg (GET_CODE (operands[1]), operands[2],
@@ -2971,7 +2971,7 @@
(if_then_else:GPF
(match_operator 1 "aarch64_comparison_operator"
[(match_operand:GPF 2 "register_operand" "")
- (match_operand:GPF 3 "register_operand" "")])
+ (match_operand:GPF 3 "aarch64_fp_compare_operand" "")])
(match_operand:GPF 4 "register_operand" "")
(match_operand:GPF 5 "register_operand" "")))]
""
--
1.9.1
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: [PATCH][AArch64] Update patterns to support FP zero
2015-10-08 13:16 [PATCH][AArch64] Update patterns to support FP zero Wilco Dijkstra
@ 2015-10-13 7:56 ` James Greenhalgh
0 siblings, 0 replies; 2+ messages in thread
From: James Greenhalgh @ 2015-10-13 7:56 UTC (permalink / raw)
To: Wilco Dijkstra; +Cc: GCC Patches
On Thu, Oct 08, 2015 at 02:16:16PM +0100, Wilco Dijkstra wrote:
> This patch improves support for instructions that allow FP zero immediate. All FP compares generated
> by various patterns should use aarch64_fp_compare_operand. LDP/STP uses aarch64_reg_or_fp_zero.
> Passes regression on AArch64.
>
> OK for commit?
OK.
Thanks,
James
>
> ChangeLog:
> 2015-10-08 Wilco Dijkstra <wdijkstr@arm.com>
>
> * gcc/config/aarch64/aarch64.md (cbranch<mode>4):
> Use aarch64_fp_compare_operand.
> (store_pairsf): Use aarch64_reg_or_fp_zero.
> (store_pairdf): Likewise.
> (cstore<mode>4): Use aarch64_fp_compare_operand.
> (cmov<mode>6): Likewise.
> * gcc/config/aarch64/aarch64-ldpstp.md: Use aarch64_reg_or_fp_zero.
>
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2015-10-13 7:56 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2015-10-08 13:16 [PATCH][AArch64] Update patterns to support FP zero Wilco Dijkstra
2015-10-13 7:56 ` James Greenhalgh
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).