[PATCH] D154678: [InstCombine] Fold IEEE `fmul`/`fdiv` by Pow2 to `add`/`sub` of exp

[Noah Goldstein via Phabricator via llvm-commits]

goldstein.w.n added a comment.

In D154678#4486086 <https://reviews.llvm.org/D154678#4486086>, @arsenm wrote:

> In D154678#4481841 <https://reviews.llvm.org/D154678#4481841>, @goldstein.w.n wrote:
>
>>> Can we unconditionally turn fdiv into fmul if the divisor is known to be an exact power of 2?
>>
>> Not sure, best I can come up with just take reciprical.
>
> Seems to work for power of 2 known <= max_exp. Much better to do that rather than try to handle both

So you mean handle `fdiv` only and do the following:

  (fdiv C, (uitofp Pow2))
      -> (fmul C, (bitcast_to_FP (sub (FP_One), Log2(Pow2) << mantissa)))

?
Otherwise we still have the division to create the RCP. In this case
Seems we might as well also drop the `fmul`.

But I'm planning to abandon this patch as have implemented in the backend.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D154678/new/

https://reviews.llvm.org/D154678