[PATCH] D122052: [RISCV] Add test case for miscompile caused by treating ANY_EXTEND of constants as SIGN_EXTEND.

Craig Topper via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Fri Mar 18 18:24:36 PDT 2022


craig.topper created this revision.
craig.topper added reviewers: jrtc27, asb, frasercrmck, luismarques.
Herald added subscribers: s, VincentWu, luke957, StephenFan, vkmr, evandro, apazos, sameer.abuasal, s.egerton, Jim, benna, psnobl, jocewei, PkmX, the_o, brucehoult, MartinMosbeck, rogfer01, edward-jones, zzheng, kito-cheng, niosHD, sabuasal, simoncook, johnrusso, rbar, arichardson.
Herald added a project: All.
craig.topper requested review of this revision.
Herald added subscribers: pcwang-thead, eopXD, MaskRay.
Herald added a project: LLVM.

The code that inserts AssertZExt based on predecessor information assumes
constants are zero extended for phi incoming values this allows
AssertZExt to be created in blocks consuming a Phi.
SelectionDAG::getNode treats any_extend of i32 constants as sext for RISCV.
The code that creates phi incoming values in the predecessors creates an
any_extend for the constants which then gets treated as a sext by getNode.
This makes the AssertZExt incorrect and can cause zexts to be
incorrectly removed.

This bug was introduced by D105918 <https://reviews.llvm.org/D105918>


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D122052

Files:
  llvm/test/CodeGen/RISCV/aext-to-sext.ll


Index: llvm/test/CodeGen/RISCV/aext-to-sext.ll
===================================================================
--- llvm/test/CodeGen/RISCV/aext-to-sext.ll
+++ llvm/test/CodeGen/RISCV/aext-to-sext.ll
@@ -75,3 +75,36 @@
 bar:
   ret i32 %b
 }
+
+; The code that inserts AssertZExt based on predecessor information assumes
+; constants are zero extended for phi incoming values so an AssertZExt is
+; created in 'merge' allowing the zext to be removed.
+; SelectionDAG::getNode treats any_extend of i32 constants as sext for RISCV.
+; The code that creates phi incoming values in the predecessors creates an
+; any_extend for the constants which then gets treated as a sext by getNode.
+; This means the zext was not safe to remove.
+define i64 @miscompile(i32 %c) {
+; RV64I-LABEL: miscompile:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    sext.w a1, a0
+; RV64I-NEXT:    li a0, -1
+; RV64I-NEXT:    beqz a1, .LBB2_2
+; RV64I-NEXT:  # %bb.1: # %merge
+; RV64I-NEXT:    ret
+; RV64I-NEXT:  .LBB2_2: # %iffalse
+; RV64I-NEXT:    li a0, -2
+; RV64I-NEXT:    ret
+  %a = icmp ne i32 %c, 0
+  br i1 %a, label %iftrue, label %iffalse
+
+iftrue:
+  br label %merge
+
+iffalse:
+  br label %merge
+
+merge:
+  %b = phi i32 [-1, %iftrue], [-2, %iffalse]
+  %d = zext i32 %b to i64
+  ret i64 %d
+}


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D122052.416648.patch
Type: text/x-patch
Size: 1288 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20220319/6763ab9c/attachment.bin>


More information about the llvm-commits mailing list