[llvm] r236517 - Revert "Fix IfConverter to handle regmask machine operands."

Pete Cooper peter_cooper at apple.com
Tue May 5 11:49:05 PDT 2015


Author: pete
Date: Tue May  5 13:49:05 2015
New Revision: 236517

URL: http://llvm.org/viewvc/llvm-project?rev=236517&view=rev
Log:
Revert "Fix IfConverter to handle regmask machine operands."

This reverts commit b27413cbfd78d959c18e713bfa271fb69e6b3303 (ie r236515).

This is to get the bots green while i investigate the failures.

Removed:
    llvm/trunk/test/CodeGen/ARM/ifcvt-regmask-noreturn.ll
Modified:
    llvm/trunk/lib/CodeGen/IfConversion.cpp

Modified: llvm/trunk/lib/CodeGen/IfConversion.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/IfConversion.cpp?rev=236517&r1=236516&r2=236517&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/IfConversion.cpp (original)
+++ llvm/trunk/lib/CodeGen/IfConversion.cpp Tue May  5 13:49:05 2015
@@ -985,20 +985,6 @@ static void UpdatePredRedefs(MachineInst
     // take a mutable instruction instead of const.
     MachineInstr *OpMI = const_cast<MachineInstr*>(Op.getParent());
     MachineInstrBuilder MIB(*OpMI->getParent()->getParent(), OpMI);
-
-    if (Op.isRegMask()) {
-      // First handle regmasks.  They clobber any entries in the mask which
-      // means that we need a def for those registers.
-      MIB.addReg(Reg.first, RegState::Implicit | RegState::Undef);
-
-      // We also need to add an implicit def of this register for the later
-      // use to read from.
-      // For the register allocator to have allocated a register clobbered
-      // by the call which is used later, it must be the case that
-      // the call doesn't return.
-      MIB.addReg(Reg.first, RegState::Implicit | RegState::Define);
-      continue;
-    }
     assert(Op.isReg() && "Register operand required");
     MIB.addReg(Reg.first, RegState::Implicit | RegState::Undef);
   }

Removed: llvm/trunk/test/CodeGen/ARM/ifcvt-regmask-noreturn.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/ARM/ifcvt-regmask-noreturn.ll?rev=236516&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/ARM/ifcvt-regmask-noreturn.ll (original)
+++ llvm/trunk/test/CodeGen/ARM/ifcvt-regmask-noreturn.ll (removed)
@@ -1,45 +0,0 @@
-; RUN: llc %s -o - -verify-machineinstrs | FileCheck %s
-
-target datalayout = "e-m:o-p:32:32-f64:32:64-v64:32:64-v128:32:128-a:0:32-n32-S32"
-target triple = "thumbv7s-apple-ios8.0.0"
-
- at debw = external global i8*, align 4
-
-; This test ensures that the stack_chk call correctly puts implicit uses/defs for the regsiters
-; live across it when if converting.  This will be R0 which is passed to the call to free at the end
-; of the function.
-; Prior to this change, the stack_chk call (which does not return) would clobber R0 in its regmask,
-; leading to verifier errors because the later use of R0 in free() is not live.
-
-; CHECK-LABEL: @test
-; CHECK: stack_chk_fail
-
-; Function Attrs: ssp
-define void @test(i32 %argc, i8** nocapture readonly %argv, i32* %ptr, i32 %val) #0 {
-entry:
-  %count.i = alloca [256 x i32], align 4
-  %cmp284.i = icmp eq i32 %val, 0
-  br i1 %cmp284.i, label %for.end31.i, label %for.body21.i
-
-for.body21.i:                                     ; preds = %entry
-  %arrayidx23.i = getelementptr inbounds [256 x i32], [256 x i32]* %count.i, i32 0, i32 1
-  %tmp20 = load i32, i32* %arrayidx23.i, align 4, !tbaa !0
-  store i32 %tmp20, i32* %ptr, align 4, !tbaa !0
-  br label %for.end31.i
-
-for.end31.i:                                      ; preds = %for.body21.i, %entry
-  %tmp21 = load i8*, i8** @debw, align 4, !tbaa !4
-  tail call void @free(i8* %tmp21)
-  ret void
-}
-
-declare void @free(i8* nocapture)
-
-attributes #0 = { ssp "stack-protector-buffer-size"="8" }
-
-!0 = !{!1, !1, i64 0}
-!1 = !{!"int", !2, i64 0}
-!2 = !{!"omnipotent char", !3, i64 0}
-!3 = !{!"Simple C/C++ TBAA"}
-!4 = !{!5, !5, i64 0}
-!5 = !{!"any pointer", !2, i64 0}





More information about the llvm-commits mailing list