Hello!
I am trying to implement the following insns for my back end: smulhssi3,
smulhshi3, smulhsqi3
According to [1], the operation should be equivalent to:
narrow op0, op1, op2;
op0 = (narrow) (((wide) op1 * (wide) op2) >> (N / 2 - 1));
...so I tried to write a corresponding matching pattern, like so:
(define_insn "smulhshi3"
[(set (match_operand:HI 0 "register_operand" "=r")
(truncate:HI
(ashiftrt:SI
(mult:SI
(sign_extend:SI (match_operand:HI 1 "register_operand" "r"))
(sign_extend:SI (match_operand:HI 2 "register_operand" "r")))
(const_int 15))))]
"TARGET_PACKED_OPS"
"mulq.h\\t%0, %1, %2")
However, I am unable to trigger this code path. I have tried with the
following C code:
short mulq(short op1, short op2) {
return (short) (((int) op1 * (int) op2) >> (32 / 2 - 1));
}
But I just get the regular 4-instruction result (2x sign extend, 1x mul,
1x shift).
Are there any builtins that I need to use, or anything that I need to
enable for my back end? Or should I use another pattern?
I also have similar problems with avgsi3_floor and friends.
Regards,
Marcus
[1] https://gcc.gnu.org/onlinedocs/gccint/Standard-Names.html
