[RegAlloc][AArch64] Add test case for terminal rule. NFC

[hstk30-hw]

Add testcase for terminal rule(#169219)

[llvmbot]

@llvm/pr-subscribers-backend-aarch64

Author: None (hstk30-hw)

Changes

Add testcase for terminal rule(#169219)

---

Full diff: https://github.com/llvm/llvm-project/pull/170035.diff

1 Files Affected:

• (added) llvm/test/CodeGen/AArch64/apply-terminal-rule.mir (+98)

diff --git a/llvm/test/CodeGen/AArch64/apply-terminal-rule.mir b/llvm/test/CodeGen/AArch64/apply-terminal-rule.mir
new file mode 100644
index 0000000000000..f59452d8f253d
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/apply-terminal-rule.mir
@@ -0,0 +1,98 @@
+# RUN: llc -mtriple=aarch64 -run-pass=register-coalescer -terminal-rule=1 -o - %s | FileCheck %s
+
+# Apply terminal rule for: %5
+--- |
+  @A = external dso_local global [100 x i32], align 4
+  
+  define i32 @test(i32 %n) {
+  entry:
+    %cmp63 = icmp sgt i32 %n, 0
+    br i1 %cmp63, label %for.body.preheader, label %for.cond.cleanup
+  
+  for.body.preheader:                               ; preds = %entry
+    %0 = load i32, i32* getelementptr inbounds ([100 x i32], [100 x i32]* @A, i64 0, i64 0), align 4
+    br label %for.body
+  
+  for.cond.cleanup:                                 ; preds = %for.body, %entry
+    %sum.0.lcssa = phi i32 [ %n, %entry ], [ %div, %for.body ]
+    ret i32 %sum.0.lcssa
+  
+  for.body:                                         ; preds = %for.body, %for.body.preheader
+    %lsr.iv = phi i32 [ %n, %for.body.preheader ], [ %lsr.iv.next, %for.body ]
+    %sum.065 = phi i32 [ %div, %for.body ], [ %n, %for.body.preheader ]
+    %call = tail call i32 @_Z3usei(i32 %0)
+    %div = sdiv i32 %sum.065, %call
+    %lsr.iv.next = add i32 %lsr.iv, -1
+    %exitcond.not = icmp eq i32 %lsr.iv.next, 0
+    br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
+  }
+  
+  declare i32 @_Z3usei(i32)
+
+...
+---
+name:            test
+alignment:       4
+tracksRegLiveness: true
+liveins:
+  - { reg: '$w0', virtual-reg: '%6' }
+frameInfo:
+  maxAlignment:    1
+  adjustsStack:    true
+  hasCalls:        true
+  maxCallFrameSize: 0
+machineFunctionInfo: {}
+body:             |
+  bb.0.entry:
+    successors: %bb.1(0x50000000), %bb.4(0x30000000)
+    liveins: $w0
+
+    %0:gpr32common = COPY $w0
+    dead $wzr = SUBSWri %0:gpr32common, 1, 0, implicit-def $nzcv
+    Bcc 10, %bb.1, implicit killed $nzcv
+    B %bb.4
+
+  bb.1.for.body.preheader:
+    successors: %bb.3(0x80000000)
+
+    %1:gpr64common = ADRP target-flags(aarch64-page) @A
+    %2:gpr32 = LDRWui %1:gpr64common, target-flags(aarch64-pageoff, aarch64-nc) @A :: (dereferenceable load (s32) from `i32* getelementptr inbounds ([100 x i32], [100 x i32]* @A, i64 0, i64 0)`)
+    %3:gpr32all = COPY %2:gpr32
+    %4:gpr32sp = COPY %0:gpr32common
+    %5:gpr32 = COPY %0:gpr32common
+    B %bb.3
+
+  bb.2.for.cond.cleanup:
+    %6:gpr32all = COPY %7:gpr32all
+    $w0 = COPY %6:gpr32all
+    RET_ReallyLR implicit $w0
+
+  bb.3.for.body:
+    successors: %bb.2(0x04000000), %bb.3(0x7c000000)
+    
+    %8:gpr32 = COPY %5:gpr32
+    %9:gpr32sp = COPY %4:gpr32sp
+    ADJCALLSTACKDOWN 0, 0, implicit-def dead $sp, implicit $sp
+    $w0 = COPY %3:gpr32all
+    BL @_Z3usei, csr_aarch64_aapcs, implicit-def dead $lr, implicit $sp,implicit killed $w0, implicit-def $sp, implicit-def $w0
+    ADJCALLSTACKUP 0, 0, implicit-def dead $sp, implicit $sp
+    %10:gpr32 = COPY $w0
+    %11:gpr32 = SDIVWr %8:gpr32, %10:gpr32
+    %12:gpr32all = COPY %11:gpr32
+    %13:gpr32 = SUBSWri %9:gpr32sp, 1, 0, implicit-def $nzcv
+    %14:gpr32all = COPY %13:gpr32
+    %7:gpr32all = COPY %12:gpr32all
+    %4:gpr32sp = COPY %14:gpr32all
+    %5:gpr32 = COPY %12:gpr32all
+    Bcc 0, %bb.2, implicit killed $nzcv
+    B %bb.3
+
+  bb.4:
+  ; CHECK-LABEL: bb.4:
+  ; CHECK-NOT: %6:gpr32 = COPY %5
+    successors: %bb.2(0x80000000)
+
+    %7:gpr32all = COPY %0:gpr32common
+    B %bb.2
+
+...

[davemgreen]

Can we just update_mir_test_checks the test?

I believe we saw a number of perf regressions after #168661 from loops that contained extra copies that were not present before. IIUC the performance went back down in some cases to where they were before the terminal rule was enabled.