unexpected insn sign extension

[" J S Hammond" <jshammond at my-Deja dot com>]

I have the following piece of code

main()
{
   short a,b;

   a = 2;
   b = a * ((short)101);
}

When I compile it using my gcc cross compiler
I get the following error message

test.c: In function `main':
test.c:7: internal error--unrecognizable insn:

(insn 14 13 16 (set (reg:HI 21)
        (mult:HI (sign_extend:HI (reg:QI 22))
            (sign_extend:HI (const_int 101)))) -1 (nil)
    (nil))
/cygwincore/gcc/toplev.c:1360: Internal compiler error in function fatal_insn

My first question is why would my multiply
try to sign extend the QI arguments to HI mode
when everything is in QI mode?

If I use another short variable instead of 
the typecast constant I get a multiply
statement that uses the QI mode multiply
(mulqi3).

I have defined mulqihi3 like so,
(define_insn "mulqihi3"
  [(set (match_operand:HI 0 "register_operand" "=r,r")
    (mult:HI (sign_extend:HI (match_operand:QI 1 "register_operand" "%0,0"))
             (sign_extend:HI (match_operand:QI 2 "general_operand" "r,i"))))]
  ""
  "@
     mr   r%0,r%2      ; mulqihi3 - case 0(1)
     mim  r%0,%2      ; mulqihi3 - case 1(1)")

My second question is why doesn't the 
unrecognized insn match case 1 of mulqihi3?
I can see that operands 0 & 1 are not the same
in my unregcnized insn like my mulqihi3 has,
but doesn't the comipler usually find a way around this?

Any help appreciated,

Jeff Hammond



--== Sent via Deja.com http://www.deja.com/ ==--
Before you buy.