[LLVMdev] Possible missed optimization?

[Borja Ferrer]

Hello, while testing trivial functions in my backend i noticed a suboptimal
way of assigning regs that had the following pattern, consider the following
function:

typedef unsigned short t;
t foo(t a, t b)
{
    t a4 = b^a^18;
    return a4;
}

Argument "a" is passed in R15:R14 and argument "b" is passed in R13:R12, the
return value is stored in R15:R14.

Producing the following asm code:
    xor    r15, r13  ; xor top part
    mov    r8, r14
    xor    r8, r12   ; xor bottom part
    movi    r14, 18
    xor    r14, r8  ; xor bottom part with imm value

However it could have produced:
    xor     r15, r13
    xor     r14, r12
    movi   r8, 18
    xor     r14, r8

This saves the first move instruction by using any scratch register when
moving the imm value instead of using r14 which is used as an incoming reg
and as the return value. Is this a missed optimization from LLVM or did i
miss something out?
Changing the register allocation order didnt work.

Thanks
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20100904/a6db132b/attachment.html>