llvm-reduce should convert array alloca to simple alloca

[arsenm]

I sometimes see testcases reduced with alloca remaining with a 0 sized array, e.g.
%alloc = alloca i32, i32 0

It would be nicer if the array size was folded out

[efriedma-quic]

Note that an alloca with the array size "folded out" is actually just alloca i32, i32 1 in disguise.

[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. Assign the issue to you.
2. Fix the issue locally.
3. Run the test suite locally.
   3.1) 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.
4. Create a git commit
5. Run git clang-format HEAD~1 to format your changes.
6. Submit the patch to Phabricator.
   6.1) Detailed instructions can be found here

For more instructions on how to submit a patch to LLVM, see our documentation.

If you have any further questions about this issue, don't hesitate to ask via a comment on this Github issue.

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

[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. Assign the issue to you.
2. Fix the issue locally.
3. Run the test suite locally.
   3.1) 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.
4. Create a git commit
5. Run git clang-format HEAD~1 to format your changes.
6. Submit the patch to Phabricator.
   6.1) Detailed instructions can be found here

For more instructions on how to submit a patch to LLVM, see our documentation.

If you have any further questions about this issue, don't hesitate to ask via a comment on this Github issue.

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

[eddiep24]

What advantage does

%alloc = alloca i32, i32 1

have over

%alloc = alloca i32, i32 0?

[arsenm]

size 1 allocations are canonical. Also size 0 allocation is an automatic path to undefined behavior, and something defined is preferable to undefined

[PranavReddyP16]

Hey @arsenm is this still free to work on? If so, can you please assign this to me? I'd love to take this up.

[PranavReddyP16]

Hi @arsenm @efriedma-quic just a quick question. Would we not want to remove the alloc and all further variable instances? I don't have any experience here but I can't think of a scenario where we would want to keep it

[efriedma-quic]

I think there's an existing transform that tries to replace uses of instructions, including alloca, with poison, or something like that. But that replacement might make the interesting-ness test fail. In that case, we need to consider whether we can simplify the alloca instruction itself.