Hi,
I have a question about having multiple insns for one operation. Take
m68k port for example. It has many SImode move insns like so:
(define_insn "pushexthisi_const"
[(set (match_operand:SI 0 "push_operand" "=m")
(match_operand:SI 1 "const_int_operand" "J"))]
(define_insn ""
[(set (match_operand:SI 0 "nonimmediate_operand" "=a")
(match_operand:QI 1 "address_operand" "p"))]
(define_insn "movsi_const0"
[(set (match_operand:SI 0 "nonimmediate_operand" "=g")
(const_int 0))]
(define_insn ""
;; Notes: make sure no alternative allows g vs g.
;; We don't allow f-regs since fixed point cannot go in them.
[(set (match_operand:SI 0 "nonimmediate_operand" "=g,d,a<")
(match_operand:SI 1 "general_src_operand" "daymSKT,n,i"))]
"!TARGET_COLDFIRE"
Now, is it OK to have this many? IIRC, reload makes certain changes
to insns without re-recognizing modified ones. (Well, it used to at
least.) A few years ago, I ran into this kind of trouble in H8 port
having multiple add insns, which were cleaned up soon afterwards.
Aside from reload issues, are there other potential problems with
having multiple insns for one operation? Maybe maintainance reasons?
Having multiple move insns may make it difficult for people to figure
out which pattern a given insn matches because we would have to go
through a machine description to find the first matching insn. (By
the way, I do understand that it's OK to have multiple patterns for
different targets because values of TARGET_xxx don't change within one
run of cc1.)
FWIW, the second define_insn above could have "plus" in
operands[1], in which case it overlaps with other insns with plus.
Thanks in advance,
Kazu Hirata
