Update expected instructions for LLVM 22

[nikic]

No description provided.

[rustbot]

r? @folkertdev

rustbot has assigned @folkertdev.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[nikic]

First i686 difference is due to this: https://llvm.godbolt.org/z/oGW1d66zY

Codegen for _mm256_set_pd has changed (which is fine, it does not correspond to a specific instruction), though llvm-mca seems to think that the new form is slower. (I have some doubts that it's correct about that though.)

[nikic]

Second i686 difference: https://llvm.godbolt.org/z/41Td7oaqq

Codegen for _mm_set_ps changed, though here llvm-mca agrees that the new form is better.

[nikic]

The aarch64 failures are all variants of this: https://llvm.godbolt.org/z/PnGdxY3rf

Where zip2 is now used instead of a lane move.

[folkertdev]

https://clang.godbolt.org/z/jhPP8hdPq

for powerpc I think the new output, actually emitting vrfin, is correct.

[nikic]

I filed llvm/llvm-project#178632 to check on the vperm2f128 change, as that seems the only potentially problematic one.

[folkertdev]

Nice, thanks!