[LLVMdev] Double spills with Greedy regalloc

Fri Jan 27 10:34:41 PST 2012

On Jan 27, 2012, at 10:20 AM, Borja Ferrer wrote:

> Hello,
> 
> I noticed the following interesting code sequence while compiling a piece of code with the backend I'm developing. Probably this issue is for Jakob, but anyways this is what I'm getting:
> 
>     STDWPtrQRr <fi#12>, 0, %R25R24; mem:ST2[FixedStack12](align=1)
>     STDWPtrQRr <fi#12>, 0, %R25R24; mem:ST2[FixedStack12](align=1)
>     STDWPtrQRr <fi#13>, 0, %R23R22; mem:ST2[FixedStack13](align=1)
>     STDWPtrQRr <fi#13>, 0, %R23R22; mem:ST2[FixedStack13](align=1)
>     STDWPtrQRr <fi#14>, 0, %R21R20; mem:ST2[FixedStack14](align=1)
>     STDWPtrQRr <fi#14>, 0, %R21R20; mem:ST2[FixedStack14](align=1)
>     STDWPtrQRr <fi#15>, 0, %R19R18; mem:ST2[FixedStack15](align=1)
>     STDWPtrQRr <fi#15>, 0, %R19R18; mem:ST2[FixedStack15](align=1)
> 
> Each STDWPtrQRr is a spill instruction, but notice that it's spilling everything twice to each frame index. I'm only getting this with the Greedy regalloc, PBQP will not duplicate the spills.
> Is there something here that I'm missing or could it be a bug? I haven't attached the C code now to not spam the message or the huge debug output of the regalloc, but I will supply any information as requested.

Did you implement isStoreToStackSlot() ?