Instconsistent sdiv and srem by pow2 codegen under Oz

[topperc]

Bugzilla Link	52481
Version	trunk
OS	All
CC	@topperc,@RKSimon,@phoebewang,@rotateright

Extended Description

There seems to be an ordering issue in our handling of sdiv/srem by constant handling under -Oz.

These two test cases produce different code just from reordering the div and rem

int foo(int x, int *rem) {
  *rem = x % 256;
  return x / 256;
}

int bar(int x, int *div) {
  *div = x / 256;
  return x % 256;
}

foo:                                    # @foo
        mov     ecx, 256
        mov     eax, edi
        cdq
        idiv    ecx
        mov     ecx, eax
        shl     ecx, 8
        sub     edi, ecx
        mov     dword ptr [rsi], edi
        ret
bar:                                    # @bar
        mov     eax, edi
        mov     ecx, 256
        cdq
        idiv    ecx
        mov     dword ptr [rsi], eax
        mov     eax, edx
        ret

bar is using the remainder from the idiv, while foo is using the quotient to compute the remainder.

[RKSimon]

This can be handled by folding sub(x, mul(divrem(x,y)[0], y)) to divrem(x,y)[1] - i.e. when we have folded to a divrem node that we can reuse.

https://alive2.llvm.org/ce/z/FRSI44

[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: Craig Topper (topperc)

| | | | --- | --- | | Bugzilla Link | [52481](https://llvm.org/bz52481) | | Version | trunk | | OS | All | | CC | @topperc,@RKSimon,@phoebewang,@rotateright |

Extended Description

There seems to be an ordering issue in our handling of sdiv/srem by constant handling under -Oz.

These two test cases produce different code just from reordering the div and rem

int foo(int x, int *rem) {
  *rem = x % 256;
  return x / 256;
}

int bar(int x, int *div) {
  *div = x / 256;
  return x % 256;
}

foo:                                    # @<!-- -->foo
        mov     ecx, 256
        mov     eax, edi
        cdq
        idiv    ecx
        mov     ecx, eax
        shl     ecx, 8
        sub     edi, ecx
        mov     dword ptr [rsi], edi
        ret
bar:                                    # @<!-- -->bar
        mov     eax, edi
        mov     ecx, 256
        cdq
        idiv    ecx
        mov     dword ptr [rsi], eax
        mov     eax, edx
        ret

bar is using the remainder from the idiv, while foo is using the quotient to compute the remainder.

[RKSimon]

This can be handled by folding sub(x, mul(divrem(x,y)[0], y)) to divrem(x,y)[1] - i.e. when we have folded to a divrem node that we can reuse.

https://alive2.llvm.org/ce/z/FRSI44

With the slight complication that this variant also needs to match the shl created from the mul :(