InstCombine should fold ldexp(x, zext(i1 y)) to fmul x, (select y, 2.0, 1.0)

[arsenm]

define float @src(float %x, i1 %bool) {
  %zext = zext i1 %bool to i32
  %ldexp = call float @llvm.ldexp.f32.i32(float %x, i32 %zext)
  ret float %ldexp
}

define float @tgt(float %x, i1 %bool) {
  %select = select i1 %bool, float 2.000000e+00, float 1.000000e+00
  %mul = fmul float %x, %select
  ret float %mul
}

I noticed the pattern while looking at builtin math library code

[llvmbot]

Hi!

This issue may be a good introductory issue for people new to working on LLVM. If you would like to work on this issue, your first steps are:

1. Check that no other contributor has already been assigned to this issue. If you believe that no one is actually working on it despite an assignment, ping the person. After one week without a response, the assignee may be changed.
2. In the comments of this issue, request for it to be assigned to you, or just create a pull request after following the steps below. Mention this issue in the description of the pull request.
3. Fix the issue locally.
4. Run the test suite locally. Remember that the subdirectories under test/ create fine-grained testing targets, so you can e.g. use make check-clang-ast to only run Clang's AST tests.
5. Create a Git commit.
6. Run git clang-format HEAD~1 to format your changes.
7. Open a pull request to the upstream repository on GitHub. Detailed instructions can be found in GitHub's documentation. Mention this issue in the description of the pull request.

If you have any further questions about this issue, don't hesitate to ask via a comment in the thread below.

[llvmbot]

@llvm/issue-subscribers-good-first-issue

Author: Matt Arsenault (arsenm)

``` define float @src(float %x, i1 %bool) { %zext = zext i1 %bool to i32 %ldexp = call float @llvm.ldexp.f32.i32(float %x, i32 %zext) ret float %ldexp }

define float @tgt(float %x, i1 %bool) {
%select = select i1 %bool, float 2.000000e+00, float 1.000000e+00
%mul = fmul float %x, %select
ret float %mul
}

I noticed the pattern while looking at builtin math library code 
</details>

[dtcxzyw]

I think -> select cond, (fadd x, x), x is better. BTW, a freeze node is need as this fold introduces more uses of X.
Godbolt: https://godbolt.org/z/6hMT6q4Pa.

[arsenm]

https://godbolt.org/z/6hMT6q4Pa

tgt3 should be using -0 as the fadd identity value, not +0

[arsenm]

I think -> select cond, (fadd x, x), x is better. BTW, a freeze node is need as this fold introduces more uses of X. Godbolt: https://godbolt.org/z/6hMT6q4Pa.

This seems to require an additional register

[aidenfoxivey]

Would it be alright if I assigned this issue to myself?

[aidenfoxivey]

Thanks @arsenm! Looking forward to taking a crack at it. I've never worked with instcombine yet, so this is pretty exciting.

[aidenfoxivey]

Just wanted to give you folks an update - I've written a bit of code and have some to push soon, but midterms are coming up so I might not be super responsive. If someone else wants this issue really badly / already has code for it, I'm happy to hand it over.

Not that I don't want to do it myself, just don't want to hog an issue.

[c8ef]

I would be happy to keep working on it.