Undo this change (the problem was actually in reload):
Fri Jan 23 23:28:59 1998 J"orn Rennecke <amylaar@cygnus.co.uk> * sh.md (movqi_i+1): New peephole. From-SVN: r17663
This commit is contained in:
parent
2a66ec2b80
commit
2993ce4c34
2 changed files with 6 additions and 12 deletions
|
@ -1,3 +1,9 @@
|
|||
Thu Feb 5 01:45:19 1998 J"orn Rennecke <amylaar@cygnus.co.uk>
|
||||
Undo this change (the problem was actually in reload):
|
||||
Fri Jan 23 23:28:59 1998 J"orn Rennecke <amylaar@cygnus.co.uk>
|
||||
|
||||
* sh.md (movqi_i+1): New peephole.
|
||||
|
||||
Tue Feb 3 01:11:12 1998 Jeffrey A Law (law@cygnus.com)
|
||||
|
||||
* jump.c (jump_optimize): Lose calls to modified_in_p they are
|
||||
|
|
|
@ -1885,18 +1885,6 @@
|
|||
lds %1,%0"
|
||||
[(set_attr "type" "move,load,store,move,move,move")])
|
||||
|
||||
;; For Big-endian, reload now sometimes generates something like
|
||||
;; (insn 2107 2104 2099 (set (reg:QI 3 r3)
|
||||
;; (subreg:QI (mem:SI (reg:SI 0 r0)) 0)) -1 (nil)
|
||||
;; (nil))
|
||||
;; To avoid clutter in the movqi pattern, we express this as a peephole.
|
||||
|
||||
(define_peephole
|
||||
[(set (match_operand:QI 0 "arith_reg_operand" "r")
|
||||
(subreg:QI (match_operand:SI 1 "memory_operand" "m") 0))]
|
||||
""
|
||||
"mov.l %1,%0")
|
||||
|
||||
(define_expand "movqi"
|
||||
[(set (match_operand:QI 0 "general_operand" "")
|
||||
(match_operand:QI 1 "general_operand" ""))]
|
||||
|
|
Loading…
Add table
Reference in a new issue