move patterns, emit_move_insn

[Meena]

> e.c:7: error: unrecognizable insn:
> (insn 7 6 8 3 e.c:4 (set (mem/c/i:HI (reg/f:HI 14) [0 i+0 S2 A16])
>         (const_int 3 [0x3])) -1 (nil))
> 
> I think GCC has to put immediate into memory. The arch can't do that
> directly. GCC should put the immediate in register and then register
> to memory.
> It succeded with old implementation (with a lot of define_insn). What
> should I change in the new implementation to get it right ?
> 

Consider the move pattern you have included 

(define_expand "movhi"
  [(set (match_operand:HI 0 "nonimmediate_operand" "")
        (match_operand:HI 1 "general_operand" "")
  )]
  ""
  ""
)

(define_insn "movhi1"
  [(set (match_operand:HI 0 "register_operand" "=r,r,m,r")
        (match_operand:HI 1 "general_operand" "r,m,r,i")
  )]
  ""
  "@
  movew\t%1,%0
  movew\t%1,%0
  movew\t%1,%0
  movew\t%1,%0"
)

Here, the last constraints caters the case for loading immediate operand into register. I think the problem could be with eliminating FP against SP.

I have also defined the move patterns as mentioned above and I am able to successfully compiled the test case as mentioned in e.c.

Please check.




On Tue, 2009-03-03 at 10:40 +0100, Florent DEFAY wrote:
> Thank you for your help.
> 
> Regarding your advice, I tried this implementation:
> 
> ___________________________________________________
> ;; movhi
> (define_expand "movhi"
>   [(set (match_operand:HI 0 "nonimmediate_operand" "")
>         (match_operand:HI 1 "general_operand" "")
>   )]
>   ""
>   ""
> )
> 
> (define_insn "movhi1"
>   [(set (match_operand:HI 0 "register_operand" "=r,r,m,r")
>         (match_operand:HI 1 "general_operand" "r,m,r,i")
>   )]
>   ""
>   "@
>   movew\t%1,%0
>   movew\t%1,%0
>   movew\t%1,%0
>   movew\t%1,%0"
> )
> ____________________________________________________
> 
> But I get this error then:
> 
> e.c:7: error: unrecognizable insn:
> (insn 7 6 8 3 e.c:4 (set (mem/c/i:HI (reg/f:HI 14) [0 i+0 S2 A16])
>         (const_int 3 [0x3])) -1 (nil))
> 
> I think GCC has to put immediate into memory. The arch can't do that
> directly. GCC should put the immediate in register and then register
> to memory.
> It succeded with old implementation (with a lot of define_insn). What
> should I change in the new implementation to get it right ?
> 
> Another question, I wonder how to deal with access modes ?
> For example, to obtain such an ASM instruction
> movew r1,8(r2)
> should it come from a define_insn like movhi1 and then treated by
> print_operand ?
> 
> Thank you.
> Regards.
> 
> Florent