[llvm] [SLP] Loop aware cost model/tree building (PR #187391)

[Alexey Bataev via llvm-commits]

alexey-bataev wrote:

> The function is _ZN8internal32ComplexMatrixVectorMultiplyNaiveIfLi27ELi51EE7ExecuteEv. My command line is
> 
> ```
> clang -c microbenchmarks_google3.ii -o microbenchmarks_google3.base.o --save-temps -fPIE --target=x86_64-grtev4-linux-gnu -fno-unique-section-names -maes -mcx16 -msse4.2 -mpclmul -mtune=haswell -mprefer-vector-width=128 -mprfchw -mavx -std=gnu++20 -Wno-deprecated-declarations -fno-strict-aliasing -ffp-contract=off -O3
> ```

https://github.com/llvm/llvm-project/pull/192801

https://github.com/llvm/llvm-project/pull/187391