[RISCV] Exclude LPAD in machine outliner

[jaidTw]

After #139993, the RISCVIndirectBranchTracking pass is also ran before the Machine Outliner pass, this yield a possibility that the outliner could also outlined the LPAD instruction that should be placed at the target of an indirect branch. This patch excludes LPAD instruction from the candidate of machine outliner.

[llvmbot]

@llvm/pr-subscribers-backend-risc-v

Author: Jesse Huang (jaidTw)

Changes

After #139993, the RISCVIndirectBranchTracking pass is also ran before the Machine Outliner pass, this yield a possibility that the outliner could also outlined the LPAD instruction that should be placed at the target of an indirect branch. This patch excludes LPAD instruction from the candidate of machine outliner.

---

Full diff: https://github.com/llvm/llvm-project/pull/157220.diff

2 Files Affected:

• (modified) llvm/lib/Target/RISCV/RISCVInstrInfo.cpp (+5)
• (added) llvm/test/CodeGen/RISCV/machine-outliner-lpad.ll (+61)

diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
index ee6d6cdb00096..0d15ca27edb00 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
@@ -3511,6 +3511,11 @@ RISCVInstrInfo::getOutliningTypeImpl(const MachineModuleInfo &MMI,
       return outliner::InstrType::Illegal;
   }
 
+  // LPADs should not be outlined too
+  if (MI.getOpcode() == RISCV::AUIPC &&
+      MI.getOperand(0).getReg() == RISCV::X0)
+    return outliner::InstrType::Illegal;
+
   return outliner::InstrType::Legal;
 }
 
diff --git a/llvm/test/CodeGen/RISCV/machine-outliner-lpad.ll b/llvm/test/CodeGen/RISCV/machine-outliner-lpad.ll
new file mode 100644
index 0000000000000..9d5187c8c5f7c
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/machine-outliner-lpad.ll
@@ -0,0 +1,61 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc -mtriple riscv64 -mattr=+experimental-zicfilp < %s | FileCheck %s --check-prefixes=CHECK,RV64
+
+define i32 @test1() #0 {
+; CHECK-LABEL: test1:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    lpad 0
+; CHECK-NEXT:    li a0, 0
+; CHECK-NEXT:    ret
+entry:
+  ret i32 0
+}
+
+define i32 @test2() #0 {
+; CHECK-LABEL: test2:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    lpad 0
+; CHECK-NEXT:    li a0, 0
+; CHECK-NEXT:    ret
+entry:
+  ret i32 0
+}
+
+define i32 @test3() #0 {
+; CHECK-LABEL: test3:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    lpad 0
+; CHECK-NEXT:    li a0, 0
+; CHECK-NEXT:    ret
+entry:
+  ret i32 0
+}
+
+define i32 @test4() #0 {
+; CHECK-LABEL: test4:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    lpad 0
+; CHECK-NEXT:    li a0, 0
+; CHECK-NEXT:    ret
+entry:
+  ret i32 0
+}
+
+define i32 @main() #0 {
+; CHECK-LABEL: main:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    lpad 0
+; CHECK-NEXT:    li a0, 0
+; CHECK-NEXT:    ret
+entry:
+  ret i32 0
+}
+
+attributes #0 = { minsize }
+
+!llvm.module.flags = !{!0, !1}
+
+!0 = !{i32 8, !"cf-protection-branch", i32 1}
+!1 = !{i32 1, !"cf-branch-label-scheme", !"unlabeled"}
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; RV64: {{.*}}

[github-actions]

✅ With the latest revision this PR passed the C/C++ code formatter.

[mylai-mtk]

Why distinguish these 2?

[jaidTw]

Oh thanks
I originally used i32 and the codegen was different, I forgot to remove them

[mylai-mtk]

Remove --check-prefixes

[mylai-mtk]

I think this comment is a bit redundant (it only says what's done instead of why it's done this way). I suggest to just remove it.

[mylai-mtk]

This comment is also redundant.

[mylai-mtk]

LGTM

[wangpc-pp]

LGTM.