[llvm-commits] [llvm-testresults] lab-mini-01__O3-plain__clang_DEV__x86_64 test results

[Jakob Stoklund Olesen]

On May 25, 2012, at 8:28 AM, Duncan Sands <baldrick at free.fr> wrote:

> The reason for the big slowdown is in the different codegen:
> 
> # phi213.i is in %ebx
> 
> +       leaq    1(%rdx), %rsi
> +       xorl    %esi, %ebx
>         xorl    (%rax,%rdx,4), %ebx
> -       incq    %rdx
> -       xorl    %edx, %ebx
> -       cmpl    $500001, %edx           # imm = 0x7A121
> +       cmpl    $500001, %esi           # imm = 0x7A121
> +       movq    %rsi, %rdx
> 
> I'm not sure why this codegen difference arises.  Any suggestions?
> 
> If there is a fairly generic explanation for the different codegen, maybe the
> rank function can be tweaked to force the more effective order.

The leaq is created by the 2-addr instruction pass. It is converting the incq to a 3-addr instruction because there are other uses of the pre-increment value.

One might argue that LSR should have fixed this by making the load use the post-increment value of the induction variable. Maybe it already did place the load before the increment, and isel decided to fold it into an xor anyway?

/jakob