Re: [PATCH-1, rs6000] Put constant into pseudo at expand when it needs two insns [PR86106]

[HAO CHEN GUI via Gcc-patches]

Hi Richard,

在 2023/3/16 15:57, Richard Biener 写道:
> I'm not sure if careful constraints massaging like adding magic letters to
> alternatives with constants to pessimize them for LRA, making them
> more expensive than spilling the constant to a register but avoid
> secondary reloads with spilling a register to the stack to make room
> for the constant, is possible - but in theory a special constraint modifier
> for this purpose could be invented.

I have made some tests on constraint modifiers. They all seems not work.
By checking the code, I found that the no reloading is always better than
reloading in LRA. So there is no way to spill the constant to register in
LRA.

      /* If this alternative can be made to work by reloading, and it
         needs less reloading than the others checked so far, record
         it as the chosen goal for reloading.  */
      if ((best_losers != 0 && losers == 0)
          || (((best_losers == 0 && losers == 0)
               || (best_losers != 0 && losers != 0))
              && (best_overall > overall
                  || (best_overall == overall
         ... // set goal_alt

Looking forward to your advice.

Thanks
Gui Haochen