MachineVerifier: Print name of failing subregister index

[arsenm]

I'm not sure of a good example to test the "does not fully support"
case.

[arsenm]

• MachineVerifier: Print name of failing subregister index #129491 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[llvmbot]

@llvm/pr-subscribers-backend-amdgpu

Author: Matt Arsenault (arsenm)

Changes

I'm not sure of a good example to test the "does not fully support"
case.

---

Full diff: https://github.com/llvm/llvm-project/pull/129491.diff

2 Files Affected:

• (modified) llvm/lib/CodeGen/MachineVerifier.cpp (+4-2)
• (added) llvm/test/MachineVerifier/AMDGPU/unsupported-subreg-index.mir (+15)

diff --git a/llvm/lib/CodeGen/MachineVerifier.cpp b/llvm/lib/CodeGen/MachineVerifier.cpp
index 4ab6fc2ddc28c..87d3033038414 100644
--- a/llvm/lib/CodeGen/MachineVerifier.cpp
+++ b/llvm/lib/CodeGen/MachineVerifier.cpp
@@ -2749,13 +2749,15 @@ MachineVerifier::visitMachineOperand(const MachineOperand *MO, unsigned MONum) {
         if (!SRC) {
           report("Invalid subregister index for virtual register", MO, MONum);
           OS << "Register class " << TRI->getRegClassName(RC)
-             << " does not support subreg index " << SubIdx << '\n';
+             << " does not support subreg index "
+             << TRI->getSubRegIndexName(SubIdx) << '\n';
           return;
         }
         if (RC != SRC) {
           report("Invalid register class for subregister index", MO, MONum);
           OS << "Register class " << TRI->getRegClassName(RC)
-             << " does not fully support subreg index " << SubIdx << '\n';
+             << " does not fully support subreg index "
+             << TRI->getSubRegIndexName(SubIdx) << '\n';
           return;
         }
       }
diff --git a/llvm/test/MachineVerifier/AMDGPU/unsupported-subreg-index.mir b/llvm/test/MachineVerifier/AMDGPU/unsupported-subreg-index.mir
new file mode 100644
index 0000000000000..80631fdcdda3e
--- /dev/null
+++ b/llvm/test/MachineVerifier/AMDGPU/unsupported-subreg-index.mir
@@ -0,0 +1,15 @@
+# RUN: not --crash llc -mtriple=amdgcn -mcpu=gfx900 -run-pass=none -o /dev/null %s 2>&1 | FileCheck %s
+
+---
+name: unsupported_subreg_index
+body:             |
+  bb.0:
+    liveins: $vgpr0_vgpr1, $sgpr0_sgpr1, $agpr0_agpr1_agpr2_agpr3
+
+    ; CHECK: *** Bad machine code: Invalid subregister index for virtual register ***
+    undef %0.sub3:vreg_96 = COPY $vgpr0
+
+    ; CHECK: *** Bad machine code: Invalid subregister index for virtual register ***
+    $vgpr0 = COPY %0.sub3
+
+...

[s-barannikov]

I'm not sure of a good example to test the "does not fully support" case.

Maybe ARM's DPR class with ssub_0 or ssub_1 register index? Not sure it is worth testing though.

[s-barannikov]

LGTM