[llvm] [TableGen] Rework `EmitIntrinsicToBuiltinMap` (PR #104681)

[Rahul Joshi via llvm-commits]

jurahul wrote:

> I don't think the size matters so much, whatever is fastest is best

I tested both with the benchmarks and there is no clear winner. One benchmark is faster with one and another one is faster with the other. So I suggest we get the space savings which are measurable. Thanks @arsenm again for the review. I'll let @Artem-B take a look as well.

https://github.com/llvm/llvm-project/pull/104681