[llvm] [polly] [SCEV] Introduce SCEVUse, use it instead of const SCEV * (NFCI). (PR #91961)

[Florian Hahn via llvm-commits]

fhahn wrote:

Before landing, just to recap on the compile-time impact of the change and follow-up changes needed

* This patch https://llvm-compile-time-tracker.com/compare.php?from=0d2c7d71744108fcfcbfb6258994da2c008314ba&to=2a95465d6980bfb3ebc45de18a55da97abef97d8&stat=instructions:u
```
stage1-O3 	stage1-ReleaseThinLTO 	stage1-ReleaseLTO-g 	stage1-O0-g 	stage1-aarch64-O3 	stage1-aarch64-O0-g 	stage2-O3 	stage2-O0-g 	stage2-clang
 (+0.01%)      (+0.03%)            (+0.01%)                   (-0.00%)              (+0.03%)           (-0.01%)           (+0.04%)      (+0.02%)      (+0.05%)
```

* Adding a canonical SCEV pointer (just initialized to itself for now) and used for compares (https://github.com/llvm/llvm-project/pull/185040): changes in the noise

* Creating canonical SCEV expressions for each SCEV at construction (https://github.com/llvm/llvm-project/pull/185042#issuecomment-4012855669) https://llvm-compile-time-tracker.com/compare.php?from=10021f0211c94d3a1d37434803acb48c08acea81&to=1bdee8bd0efaf88945213ec130af0a3a6a88dcbc&stat=instructions:u
```
stage1-O3 	stage1-ReleaseThinLTO 	stage1-ReleaseLTO-g 	stage1-O0-g 	stage1-aarch64-O3 	stage1-aarch64-O0-g 	stage2-O3 	stage2-O0-g 	stage2-clang
 (+0.06%)                (+0.05%)             (+0.02%)          (+0.01%)             (+0.04%)          (-0.01%)          (-0.03%)    (-0.01%)             (+0.01%)
```

Note that this is still with the flags never being != 0, so there will be a bit more overhead once some flags are non-zero

https://github.com/llvm/llvm-project/pull/91961