[PATCH] D28871: [CodeGenPrep]No negative cost in the ExtLd promotion

[Quentin Colombet via llvm-commits]

> On Jan 25, 2017, at 12:21 PM, Jun Bum Lim via Phabricator <reviews at reviews.llvm.org> wrote:
> 
> junbuml added a comment.
> 
> Hi Quentin,
> 
> I believe the test case I added shows the case where TotalCreatedInstsCost becomes negative :
> 
> - In the very first call of extLdPromotion(), CreatedInstsCost is 0, ExtCost is 1 for  the sext (%r), and NewCreatedInstsCost after promotion is 0 because the cost of sext becomes free after moving it up before %shl2.
> - Therefore, TotalCreatedInstsCost -= ExtCost becomes negative.
> 
> I think propagating the negative value of TotalCreatedInstsCost is also possible, but I think ceiling it to 0 is simple and conservative enough to handle this corner case.

That’s conservative, I agree, I was wondering whether it makes sense to propagate the “gain” by allowing negative values.
Put differently, I am fine with the current approach as long as we commit to revisit it in the near future.

> 
> This patch is a small particle from https://reviews.llvm.org/D27853, which should be rebased based on this.
> 
> Thanks,
> Jun
> 
> 
> https://reviews.llvm.org/D28871
> 
> 
>