[llvm] [SLP]Fix fmul/fadd fusion costs and retry FMA seeds after all blocks (PR #226117)

[Alexey Bataev via llvm-commits]

alexey-bataev wrote:

> @alexey-bataev we are seeing a cumulative ~5% performance regression in AArch64 SPEC LBM between this and #227083. I'll try to get reproducers for one or both of them.
> 
> There seem to be regular SLP patches which cause AArch64 regressions of this size in SPEC and other important benchmarks. You have consistently responded quickly and fixed issues once we have provided reproducers, which I appreciate. However, the current workflow of regression → reproducer → fix → further regression → further reproducer is not sustainable. Producing these reproducers takes substantial engineering effort, and we are falling behind the rate at which regressions are being introduced.
> 
> Would it be possible for you to add some aarch64 performance regression checking before merging changes, specifically SPEC17 and/or SPEC26 would be very helpful as an initial set?
> 
> In the meantime I think it would make sense if we could agree to revert any problematic patches and work together on fixes before landing them again, as per the [developer policy](https://llvm.org/docs/DeveloperPolicy.html#patch-reversion-policy).
> 
> I am open to other suggestions also.

I cannot test all possible configs for all possible targets myself. I'm doing a testing, fixing possible corner cases, but cannot cover everything.
Instead, I can suggest adding you as a reviewer on the patches, so you could try to test them using your specific configs and report potential regresions.
Also not, the development is an iterative process, sometimes it is not acceptable to handle all corner cases in a single patch, it makes the patch too huge and hard to review, so some fixes better to be landed after the initial part of the patch

https://github.com/llvm/llvm-project/pull/226117