[LLVMdev] Tight overlapping loops and performance

[Evan Cheng]

On Mar 2, 2009, at 4:58 PM, Jonathan Turner wrote:

>
>
>> You're misreading the asm... nothing is touching memory. (BTW, "leal
>> -1(%eax), %eax" isn't a memory operation; it's just subtracting one
>> from %eax.) You might want to try reading the LLVM IR (which you can
>> generate with llvm-gcc -S -emit-llvm); it tends to be easier to read.
>
> I tried that, but I'm still learning LLVM. Seeing indvar, phi nodes,  
> tail
> calls on printfs, and nounwinds had me more confused than the asm.
>
>> A taken and non-taken branch have roughly the same cost on any
>> remotely recent x86 processor.
>
> I was wondering if that might be the case.
>
> The crux of the example still seems intact.  From LLVM SVN,  
> converted to asm via llc:
>
>
>        .align  4,0x90
> LBB1_1: ## loopto
>        cmpl    $1, %eax
>        leal    -1(%eax), %eax
>        cmove   %edx, %eax
>        incl    %ecx
>        cmpl    $999999999, %ecx
>        jne     LBB1_1  ## loopto
>
> LBB1_1: ## loopto
>        decl    %eax
>        jz      LBB1_3
>        decl    %ecx
>        jnz     LBB1_1  ## loopto
>

The main issue is incl updates the EFLAGS condition code register. But  
llvm x86 isn't taking advantage of that. This is a known issue,  
hopefully someone will find the time to implement before 2.6.

The second issue is the leal -1 can be turned (back) into a decl.  
Combine that with the optimization previously described, it can  
eliminate the first cmpl.

Feel free to file a bugzilla for this. I'm hopefully this will be  
fixed in the not too far future.

Thanks,

Evan

>
> Jonathan
>
> _________________________________________________________________
> Windows Live™ Contacts: Organize your contact list.
> http://windowslive.com/connect/post/marcusatmicrosoft.spaces.live.com-Blog-cns!503D1D86EBB2B53C!2285.entry?ocid=TXT_TAGLM_WL_UGC_Contacts_032009
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev