SplitKit: Fix subreg copy related problems

Fix two problems related to r298025:
- SplitKit would create duplicate VNIs in some cases leading to crashs
  when hoisting copies.
- VirtRegMap could fail expanding copies at the beginning of a basic
  block.

This fixes http://llvm.org/PR32353

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@298448 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/CodeGen/SplitKit.cpp b/lib/CodeGen/SplitKit.cpp
index 4e4298726c42c8ccd66ce295c10317c2652d82df..3a50aaa69985d3509e27ef7d848521d0b32bb647 100644 (file)
--- a/lib/CodeGen/SplitKit.cpp
+++ b/lib/CodeGen/SplitKit.cpp
@@ -502,7 +502,6 @@ SlotIndex SplitEditor::buildSingleSubRegCopy(unsigned FromReg, unsigned ToReg,
   if (FirstCopy) {
     SlotIndexes &Indexes = *LIS.getSlotIndexes();
     Def = Indexes.insertMachineInstrInMaps(*CopyMI, Late).getRegSlot();
-    DestLI.createDeadDef(Def, Allocator);
   } else {
     CopyMI->bundleWithPred();
   }

diff --git a/lib/CodeGen/VirtRegMap.cpp b/lib/CodeGen/VirtRegMap.cpp
index 06a97fd6d4d6192a858f33de8de06861354e6997..c8946010e9d15fb426f493bb3527608e6cd39bc0 100644 (file)
--- a/lib/CodeGen/VirtRegMap.cpp
+++ b/lib/CodeGen/VirtRegMap.cpp
@@ -383,8 +383,10 @@ void VirtRegRewriter::expandCopyBundle(MachineInstr &MI) const {
 
   if (MI.isBundledWithPred() && !MI.isBundledWithSucc()) {
     // Only do this when the complete bundle is made out of COPYs.
+    MachineBasicBlock &MBB = *MI.getParent();
     for (MachineBasicBlock::reverse_instr_iterator I =
-         std::next(MI.getReverseIterator()); I->isBundledWithSucc(); ++I) {
+         std::next(MI.getReverseIterator()), E = MBB.instr_rend();
+         I != E && I->isBundledWithSucc(); ++I) {
       if (!I->isCopy())
         return;
     }

diff --git a/test/CodeGen/AMDGPU/splitkit.mir b/test/CodeGen/AMDGPU/splitkit.mir
index ee58138b1af5fe93367153891f691be8ec1a0937..41782af40e3cb76a0022688b6addbf101aa16ed8 100644 (file)
--- a/test/CodeGen/AMDGPU/splitkit.mir
+++ b/test/CodeGen/AMDGPU/splitkit.mir
@@ -2,6 +2,7 @@
 --- |
   define amdgpu_kernel void @func0() #0 { ret void }
   define amdgpu_kernel void @func1() #0 { ret void }
+  define amdgpu_kernel void @splitHoist() #0 { ret void }
 
   attributes #0 = { "amdgpu-num-sgpr"="12" }
 ...
@@ -62,3 +63,43 @@ body: |
     S_NOP 0, implicit %0.sub0
     S_NOP 0, implicit %0.sub2
 ...
+---
+# Check that copy hoisting out of loops works. This mainly should not crash the
+# compiler when it hoists a subreg copy sequence.
+# CHECK-LABEL: name: splitHoist
+# CHECK: S_NOP 0, implicit-def %sgpr0
+# CHECK: S_NOP 0, implicit-def %sgpr3
+# CHECK-NEXT: SI_SPILL_S128_SAVE
+name: splitHoist
+tracksRegLiveness: true
+body: |
+  bb.0:
+    successors: %bb.1, %bb.2
+    S_NOP 0, implicit-def undef %0.sub0 : sreg_128
+    S_NOP 0, implicit-def %0.sub3 : sreg_128
+
+    S_CBRANCH_VCCNZ %bb.1, implicit undef %vcc
+    S_BRANCH %bb.2
+
+  bb.1:
+    successors: %bb.1, %bb.3
+    S_NOP 0, implicit %0.sub0
+
+    ; Clobber registers
+    S_NOP 0, implicit-def dead %sgpr0, implicit-def dead %sgpr1, implicit-def dead %sgpr2, implicit-def dead %sgpr3, implicit-def dead %sgpr4, implicit-def dead %sgpr5, implicit-def dead %sgpr6, implicit-def dead %sgpr7, implicit-def dead %sgpr8, implicit-def dead %sgpr9, implicit-def dead %sgpr10, implicit-def dead %sgpr11
+
+    S_CBRANCH_VCCNZ %bb.1, implicit undef %vcc
+    S_BRANCH %bb.3
+
+  bb.2:
+    successors: %bb.3
+    ; Clobber registers
+    S_NOP 0, implicit-def dead %sgpr0, implicit-def dead %sgpr1, implicit-def dead %sgpr2, implicit-def dead %sgpr3, implicit-def dead %sgpr4, implicit-def dead %sgpr5, implicit-def dead %sgpr6, implicit-def dead %sgpr7, implicit-def dead %sgpr8, implicit-def dead %sgpr9, implicit-def dead %sgpr10, implicit-def dead %sgpr11
+    S_BRANCH %bb.3
+
+  bb.3:
+    S_NOP 0, implicit %0.sub0
+    S_NOP 0, implicit %0.sub3
+    S_NOP 0, implicit %0.sub0
+    S_NOP 0, implicit %0.sub3
+...