[PATCH] D24681: Optimize patterns of vectorized interleaved memory accesses for X86.

[Elena Demikhovsky via llvm-commits]

delena added inline comments.


> X86InterleavedAccess.cpp:70
> +///   %v2 = shuffle %intrshuffvec3, %intrshuffvec4, <0, 4, 2, 6>;
> +///   %v3 = shuffle %intrshuffvec3, %intrshuffvec4, <1, 5, 3, 7>;
> +///

AVX512 probably has another set of shuffles

> X86InterleavedAccess.cpp:89
> +    Value *NewBasePtr =
> +        Builder.CreateGEP(nullptr, VecBasePtr, Builder.getInt32(Part));
> +    Instruction *NewLoad =

inbounds GEP?

> X86InterleavedAccess.cpp:143
> +
> +  return lower(LI, Shuffles, Indices, Factor);
> +}

It is not a good name for function. I think that you don't need additional function call here at all.

https://reviews.llvm.org/D24681