[llvm-bugs] [Bug 27135] New: SelectionDAGBuilder::visitSwitch() creates disconnected MBB

[via llvm-bugs]

https://llvm.org/bugs/show_bug.cgi?id=27135

            Bug ID: 27135
           Summary: SelectionDAGBuilder::visitSwitch() creates
                    disconnected MBB
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: normal
          Priority: P
         Component: Common Code Generator Code
          Assignee: unassignedbugs at nondot.org
          Reporter: paulsson at linux.vnet.ibm.com
                CC: congh at google.com, hans at hanshq.net,
                    llvm-bugs at lists.llvm.org
    Classification: Unclassified

Created attachment 16130
  --> https://llvm.org/bugs/attachment.cgi?id=16130&action=edit
Input for compilation

visitSwitch() seems to create an MBB whithout predecessors / successors.

The attached test case is not minimized, but the verifier gives the exact spot.
If you are lucky, you can get there in gdb with:

b MachineFunction:247  // in CreateMachineBasicBlock()

condition [BP] (bb && strcmp(bb->getName().Data,"if.then72")==0)
c // continue once

u // go up the stack a few times, and that should be the SelectionDAGBuilder
handling bb:

if.then72:                                        ; preds = %land.lhs.true71
  %24 = inttoptr i64 %in.addr.sroa.0.0 to %struct.rtx_def*
  %arrayidx74 = getelementptr inbounds %struct.rtx_def, %struct.rtx_def* %24,
i64 0, i32 1, i64 0
  %25 = bitcast %union.rtunion_def* %arrayidx74 to i32**
  %26 = load i32*, i32** %25, align 8, !tbaa !5
  %bf.load75 = load i32, i32* %26, align 8
  %bf.lshr76 = lshr i32 %bf.load75, 16
  switch i32 %bf.lshr76, label %if.end92 [
    i32 99, label %sw.bb
    i32 98, label %sw.bb
    i32 101, label %sw.bb
    i32 97, label %sw.bb84
    i32 96, label %sw.bb84
    i32 100, label %sw.bb84
  ]


(gdb) run -cc1 -triple s390x-ibm-linux -S -disable-free -main-file-name
reload.c -mrelocation-model static -mthread-model posix -fmath-errno
-masm-verbose -mconstructor-aliases -fuse-init-array -target-cpu z13
-momit-leaf-frame-pointer -v -dwarf-column-info -debugger-tuning=gdb  -O3
-ferror-limit 19 -fmessage-length 238 -funroll-loops -fobjc-runtime=gcc
-fdiagnostics-show-option -fcolor-diagnostics -vectorize-loops -vectorize-slp
-o reload.s -x ir ./reload.bc

Add -mllvm -verify-machineinstrs to see the error message:

...

BB#524: derived from LLVM BB %if.then72
    Predecessors according to CFG: BB#17
        %vreg455<def> = LHIMux 1; GR32Bit:%vreg455
        %vreg456<def,tied1> = SLL %vreg455<tied0>, %vreg453, 0;
GR32Bit:%vreg456,%vreg455 ADDR32Bit:%vreg453
        TMLMux %vreg456<kill>, 19, %CC<imp-def>; GR32Bit:%vreg456
        BRC 15, 7, <BB#19>, %CC<imp-use>
        J <BB#18>
    Successors according to CFG: BB#19(0x40000000 / 0x80000000 = 50.00%)
BB#18(0x40000000 / 0x80000000 = 50.00%)

BB#525: derived from LLVM BB %if.then72     <<<<<<<<< Not connected!!

BB#18: derived from LLVM BB %sw.bb
...


*** Bad machine code: PHI operand is not in the CFG ***
- function:    push_reload
- basic block: BB#20 if.end92 (0x85a2f140)
- instruction: %vreg17<def> = PHI
- operand 16:   <BB#525>

*** Bad machine code: PHI operand is not in the CFG ***
- function:    push_reload
- basic block: BB#20 if.end92 (0x85a2f140)
- instruction: %vreg18<def> = PHI
- operand 16:   <BB#525>
fatal error: error in backend: Found 2 machine code errors.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20160330/1b685c42/attachment-0001.html>