[PATCH] D83365: [PowerPC] start and end parameters for fixupIsDeadOrKill may exist in different block before RA

ChenZheng via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Wed Jul 22 03:27:35 PDT 2020


This revision was automatically updated to reflect the committed changes.
Closed by commit rG36f9fe2d3493: [PowerPC] fixupIsDeadOrKill start and end in different block fixing (authored by shchenz).

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D83365/new/

https://reviews.llvm.org/D83365

Files:
  llvm/lib/Target/PowerPC/PPCInstrInfo.cpp
  llvm/lib/Target/PowerPC/PPCInstrInfo.h
  llvm/test/CodeGen/PowerPC/fixup-kill-dead-flag-crash.mir


Index: llvm/test/CodeGen/PowerPC/fixup-kill-dead-flag-crash.mir
===================================================================
--- /dev/null
+++ llvm/test/CodeGen/PowerPC/fixup-kill-dead-flag-crash.mir
@@ -0,0 +1,21 @@
+# RUN: llc -mtriple=powerpc64le-unknown-linux-gnu -verify-machineinstrs -start-before ppc-mi-peepholes \
+# RUN:   -stop-after ppc-mi-peepholes %s -o - | FileCheck %s
+
+---
+name: test
+#CHECK : name : test
+tracksRegLiveness: true
+body: |
+  bb.0.entry:
+    liveins: $x3
+    %0:g8rc = COPY $x3
+    %1:gprc = COPY %0.sub_32:g8rc
+    %2:g8rc = LI8 63
+
+  bb.1:
+    %3:gprc = COPY %2.sub_32:g8rc
+    ; CHECK: %4:gprc = LI 0
+    %4:gprc = XORI killed %3:gprc, 63
+    STW killed %4:gprc, %4:gprc, 100
+    BLR8 implicit $lr8, implicit $rm
+...
Index: llvm/lib/Target/PowerPC/PPCInstrInfo.h
===================================================================
--- llvm/lib/Target/PowerPC/PPCInstrInfo.h
+++ llvm/lib/Target/PowerPC/PPCInstrInfo.h
@@ -565,14 +565,18 @@
                              int64_t OffsetImm) const;
 
   /// Fixup killed/dead flag for register \p RegNo between instructions [\p
-  /// StartMI, \p EndMI]. Some PostRA transformations may violate register
-  /// killed/dead flags semantics, this function can be called to fix up. Before
-  /// calling this function,
+  /// StartMI, \p EndMI]. Some pre-RA or post-RA transformations may violate
+  /// register killed/dead flags semantics, this function can be called to fix
+  /// up. Before calling this function,
   /// 1. Ensure that \p RegNo liveness is killed after instruction \p EndMI.
   /// 2. Ensure that there is no new definition between (\p StartMI, \p EndMI)
   ///    and possible definition for \p RegNo is \p StartMI or \p EndMI.
-  /// 3. Ensure that all instructions between [\p StartMI, \p EndMI] are in same
-  ///    basic block.
+  /// 3. We can do accurate fixup for the case when all instructions between
+  ///    [\p StartMI, \p EndMI] are in same basic block.
+  /// 4. For the case when \p StartMI and \p EndMI are not in same basic block,
+  ///    we conservatively clear kill flag for all uses of \p RegNo for pre-RA
+  ///    and for post-RA, we give an assertion as without reaching definition
+  ///    analysis post-RA, \p StartMI and \p EndMI are hard to keep right.
   void fixupIsDeadOrKill(MachineInstr &StartMI, MachineInstr &EndMI,
                          unsigned RegNo) const;
   void replaceInstrWithLI(MachineInstr &MI, const LoadImmediateInfo &LII) const;
Index: llvm/lib/Target/PowerPC/PPCInstrInfo.cpp
===================================================================
--- llvm/lib/Target/PowerPC/PPCInstrInfo.cpp
+++ llvm/lib/Target/PowerPC/PPCInstrInfo.cpp
@@ -2668,6 +2668,15 @@
 
 void PPCInstrInfo::fixupIsDeadOrKill(MachineInstr &StartMI, MachineInstr &EndMI,
                                      unsigned RegNo) const {
+  // Conservatively clear kill flag for the register if the instructions are in
+  // different basic blocks and in SSA form, because the kill flag may no longer
+  // be right. There is no need to bother with dead flags since defs with no
+  // uses will be handled by DCE.
+  MachineRegisterInfo &MRI = StartMI.getParent()->getParent()->getRegInfo();
+  if (MRI.isSSA() && (StartMI.getParent() != EndMI.getParent())) {
+    MRI.clearKillFlags(RegNo);
+    return;
+  }
 
   // Instructions between [StartMI, EndMI] should be in same basic block.
   assert((StartMI.getParent() == EndMI.getParent()) &&


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D83365.279752.patch
Type: text/x-patch
Size: 3490 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20200722/bae66f35/attachment.bin>


More information about the llvm-commits mailing list