[PowerPC] fix lowering of SPILL_CRBIT on pwr9 and pwr10

[pmur]

If a copy exists between creation of a crbit and a spill, machine-cp
may delete the copy since it seems unaware of the relation between a cr
and crbit. A fix was previously made for the generic ppc64 lowering. It
should be applied to the pwr9 and pwr10 variants too.

Likewise, relax and extend the pwr8 test to verify pwr9 and pwr10
codegen too.

This fixes #143989.

[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-powerpc

Author: Paul Murphy (pmur)

Changes

If a copy exists between creation of a crbit and a spill, machine-cp
may delete the copy since it seems unaware of the relation between a cr
and crbit. A fix was previously made for the generic ppc64 lowering. It
should be applied to the pwr9 and pwr10 variants too.

Likewise, relax and extend the pwr8 test to verify pwr9 and pwr10
codegen too.

This fixes #143989.

---

Full diff: https://github.com/llvm/llvm-project/pull/146424.diff

2 Files Affected:

• (modified) llvm/lib/Target/PowerPC/PPCRegisterInfo.cpp (+11-4)
• (modified) llvm/test/CodeGen/PowerPC/NoCRFieldRedefWhenSpillingCRBIT.mir (+7-1)

diff --git a/llvm/lib/Target/PowerPC/PPCRegisterInfo.cpp b/llvm/lib/Target/PowerPC/PPCRegisterInfo.cpp
index 76dca4794e050..a9165705c2f59 100644
--- a/llvm/lib/Target/PowerPC/PPCRegisterInfo.cpp
+++ b/llvm/lib/Target/PowerPC/PPCRegisterInfo.cpp
@@ -1105,24 +1105,31 @@ void PPCRegisterInfo::lowerCRBitSpilling(MachineBasicBlock::iterator II,
     // On Power10, we can use SETNBC to spill all CR bits. SETNBC will set all
     // bits (specifically, it produces a -1 if the CR bit is set). Ultimately,
     // the bit that is of importance to us is bit 32 (bit 0 of a 32-bit
-    // register), and SETNBC will set this.
+    // register), and SETNBC will set this. Also, in order to preserve the kill
+    // flag on the CR bit, we add it as an implicit use.
     if (Subtarget.isISA3_1()) {
       BuildMI(MBB, II, dl, TII.get(LP64 ? PPC::SETNBC8 : PPC::SETNBC), Reg)
-          .addReg(SrcReg, RegState::Undef);
+          .addReg(SrcReg, RegState::Undef)
+          .addReg(SrcReg, RegState::Implicit |
+                              getKillRegState(MI.getOperand(0).isKill()));
       break;
     }
 
     // On Power9, we can use SETB to extract the LT bit. This only works for
     // the LT bit since SETB produces -1/1/0 for LT/GT/<neither>. So the value
     // of the bit we care about (32-bit sign bit) will be set to the value of
-    // the LT bit (regardless of the other bits in the CR field).
+    // the LT bit (regardless of the other bits in the CR field). Also, in
+    // order to preserve the kill flag on the CR bit, we add it as an implicit
+    // use.
     if (Subtarget.isISA3_0()) {
       if (SrcReg == PPC::CR0LT || SrcReg == PPC::CR1LT ||
           SrcReg == PPC::CR2LT || SrcReg == PPC::CR3LT ||
           SrcReg == PPC::CR4LT || SrcReg == PPC::CR5LT ||
           SrcReg == PPC::CR6LT || SrcReg == PPC::CR7LT) {
         BuildMI(MBB, II, dl, TII.get(LP64 ? PPC::SETB8 : PPC::SETB), Reg)
-          .addReg(getCRFromCRBit(SrcReg), RegState::Undef);
+            .addReg(getCRFromCRBit(SrcReg), RegState::Undef)
+            .addReg(SrcReg, RegState::Implicit |
+                                getKillRegState(MI.getOperand(0).isKill()));
         break;
       }
     }
diff --git a/llvm/test/CodeGen/PowerPC/NoCRFieldRedefWhenSpillingCRBIT.mir b/llvm/test/CodeGen/PowerPC/NoCRFieldRedefWhenSpillingCRBIT.mir
index 41e21248a3f0e..2796cdb3ae87d 100644
--- a/llvm/test/CodeGen/PowerPC/NoCRFieldRedefWhenSpillingCRBIT.mir
+++ b/llvm/test/CodeGen/PowerPC/NoCRFieldRedefWhenSpillingCRBIT.mir
@@ -1,6 +1,12 @@
 # RUN: llc -mcpu=pwr8 -mtriple=powerpc64le-unknown-linux-gnu -start-after \
 # RUN:   virtregrewriter -ppc-asm-full-reg-names -verify-machineinstrs %s \
 # RUN:   -o - | FileCheck %s
+# RUN: llc -mcpu=pwr9 -mtriple=powerpc64le-unknown-linux-gnu -start-after \
+# RUN:   virtregrewriter -ppc-asm-full-reg-names -verify-machineinstrs %s \
+# RUN:   -o - | FileCheck %s
+# RUN: llc -mcpu=pwr10 -mtriple=powerpc64le-unknown-linux-gnu -start-after \
+# RUN:   virtregrewriter -ppc-asm-full-reg-names -verify-machineinstrs %s \
+# RUN:   -o - | FileCheck %s
 
 --- |
   ; ModuleID = 'a.ll'
@@ -30,7 +36,7 @@
   ; Function Attrs: nounwind
   declare void @llvm.stackprotector(ptr, ptr) #1
   
-  attributes #0 = { nounwind "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "frame-pointer"="none" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="ppc64le" "target-features"="+altivec,+bpermd,+crypto,+direct-move,+extdiv,+htm,+power8-vector,+vsx,-power9-vector" "unsafe-fp-math"="false" "use-soft-float"="false" }
+  attributes #0 = { nounwind "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "frame-pointer"="none" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
   attributes #1 = { nounwind }
   
   !llvm.ident = !{!0}

[nikic]

For reference, the fix for the generic case was 3c7ac64.

[lei137]

Thank-you for the fix.
To prevent the same mistake from happening again when new CRbit spilling is added in the future, please move the "general" statement for this behaviour documented on lines 1137-1141 to the top of this "default:" block so new code added will take this into account.

[pmur]

Ping

[nikic]

Ping :)

[amy-kwan]

Thank you for the fix. I think this LGTM.

[nikic]

/cherry-pick 5f86456

[github-actions]

@pmur 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!

[llvmbot]

/pull-request #152654