[AMDGPU] Set hasSideEffects=0 for SALU psuedos

[amansharma612]

Fixes #128685

[github-actions]

Thank you for submitting a Pull Request (PR) to the LLVM Project!

This PR will be automatically labeled and the relevant teams will be notified.

If you wish to, you can add reviewers by using the "Reviewers" section on this page.

If this is not working for you, it is probably because you do not have write permissions for the repository. In which case you can instead tag reviewers by name in a comment by using @ followed by their GitHub username.

If you have received no comments on your PR for a week, you can request a review by "ping"ing the PR by adding a comment “Ping”. The common courtesy "ping" rate is once a week. Please remember that you are asking for valuable time from other developers.

If you have further questions, they may be answered by the LLVM GitHub User Guide.

You can also ask questions in a comment on this PR, on the LLVM Discord or on the forums.

[llvmbot]

@llvm/pr-subscribers-backend-amdgpu

Author: None (amansharma612)

Changes

Fixes #128685

---

Full diff: https://github.com/llvm/llvm-project/pull/134487.diff

2 Files Affected:

• (modified) llvm/lib/Target/AMDGPU/SIInstructions.td (+12-4)
• (modified) llvm/lib/Target/AMDGPU/SOPInstructions.td (+6-2)

diff --git a/llvm/lib/Target/AMDGPU/SIInstructions.td b/llvm/lib/Target/AMDGPU/SIInstructions.td
index 9051db0c01ed1..e78a29e36b358 100644
--- a/llvm/lib/Target/AMDGPU/SIInstructions.td
+++ b/llvm/lib/Target/AMDGPU/SIInstructions.td
@@ -340,19 +340,27 @@ def S_SUB_U64_PSEUDO : SPseudoInstSI <
 
 def S_ADD_CO_PSEUDO : SPseudoInstSI <
   (outs SReg_32:$sdst, SSrc_i1:$scc_out), (ins SSrc_b32:$src0, SSrc_b32:$src1, SSrc_i1:$scc_in)
->;
+>{
+    let hasSideEffects = 0;
+}
 
 def S_SUB_CO_PSEUDO : SPseudoInstSI <
   (outs SReg_32:$sdst, SSrc_i1:$scc_out), (ins SSrc_b32:$src0, SSrc_b32:$src1, SSrc_i1:$scc_in)
->;
+>{
+    let hasSideEffects = 0;
+}
 
 def S_UADDO_PSEUDO : SPseudoInstSI <
   (outs SReg_32:$sdst, SSrc_i1:$scc_out), (ins SSrc_b32:$src0, SSrc_b32:$src1)
->;
+>{
+    let hasSideEffects = 0;
+}
 
 def S_USUBO_PSEUDO : SPseudoInstSI <
   (outs SReg_32:$sdst, SSrc_i1:$scc_out), (ins SSrc_b32:$src0, SSrc_b32:$src1)
->;
+>{
+    let hasSideEffects = 0;
+}
 
 let OtherPredicates = [HasShaderCyclesHiLoRegisters] in
 def GET_SHADERCYCLESHILO : SPseudoInstSI<
diff --git a/llvm/lib/Target/AMDGPU/SOPInstructions.td b/llvm/lib/Target/AMDGPU/SOPInstructions.td
index 73f4655f735a2..be63bc4a745eb 100644
--- a/llvm/lib/Target/AMDGPU/SOPInstructions.td
+++ b/llvm/lib/Target/AMDGPU/SOPInstructions.td
@@ -665,12 +665,16 @@ let SubtargetPredicate = isGFX12Plus in {
   // The higher 32-bits of the inputs contain the sign extension bits.
   def S_MUL_I64_I32_PSEUDO : SPseudoInstSI <
     (outs SReg_64:$sdst), (ins SSrc_b64:$src0, SSrc_b64:$src1)
-  >;
+  >{
+      let hasSideEffects = 0;
+  }
 
   // The higher 32-bits of the inputs are zero.
   def S_MUL_U64_U32_PSEUDO : SPseudoInstSI <
     (outs SReg_64:$sdst), (ins SSrc_b64:$src0, SSrc_b64:$src1)
-  >;
+  >{
+    let hasSideEffects = 0;
+  }
 
 } // End SubtargetPredicate = isGFX12Plus
 

[kerbowa]

I'm curious why no side effects is not the default and this needs to be explicit for these pseudos in particular? Is there any motivating test/example where this impacts scheduling or some other optimization?

[arsenm]

I'm curious why no side effects is not the default

Wrong by default is a horrible design mistake LLVM has repeated in far too many places. This is the correct default

[arsenm]

Move >{ to previous line

[Pierre-vh]

nit: use let hasSideEffects=0 in { .. } to avoid repetition?

[arsenm]

Is there any motivating test/example where this impacts scheduling or some other optimization?

I doubt it, these are a SelectionDAG hack more than anything and expand immediately. It doesn't hurt to correct this though

[amansharma612]

Regression test mul.ll fails with the intended changes.

[arsenm]

Regression test mul.ll fails with the intended changes.

Probably just need to run update_llc_test_checks

[jayfoad]

Could also do this for S_INVERSE_BALLOT_U32/64

[arsenm]

LGTM but could also do this for the inverse ballots

[amansharma612]

LGTM but could also do this for the inverse ballots

@arsenm I've added the changes for inverse ballots already

[jayfoad]

Update the corresponding // End comment below.

[jayfoad]

LGTM, thanks.

[jayfoad]

mansharma612 force-pushed the salusideeffects branch from 926a934 to baddce9 3 minutes ago

To fix things in your PR it is almost always better to push an extra commit, instead of modifying your commit and force-pushing it. Force-pushing makes it harder for reviewers to see what you have changed.

[github-actions]

@amansharma612 Congratulations on having your first Pull Request (PR) merged into the LLVM Project!

Your changes will be combined with recent changes from other authors, then tested by our build bots. If there is a problem with a build, you may receive a report in an email or a comment on this PR.

Please check whether problems have been caused by your change specifically, as the builds can include changes from many authors. It is not uncommon for your change to be included in a build that fails due to someone else's changes, or infrastructure issues.

How to do this, and the rest of the post-merge process, is covered in detail here.

If your change does cause a problem, it may be reverted, or you can revert it yourself. This is a normal part of LLVM development. You can fix your changes and open a new PR to merge them again.

If you don't get any reports, no action is required from you. Your changes are working as expected, well done!