[llvm-dev] Debugging UNREACHABLE "Couldn't join subrange" in RegisterCoalescer (out-of-tree backend)

Matthias Braun via llvm-dev llvm-dev at lists.llvm.org
Mon Apr 24 09:58:18 PDT 2017


This looks like a bug related to the subregister liveness updating in the RegisterCoalescer.
As this is always hard to debug without having the actual target at hand:
- Make sure the testcase (or rather all testcases) passes `llc -verify-machineinstrs`
- It would be good to see the full `llc -debug-only=regalloc` dump (looking below I would at least see which slotindexes belong to which instructions). Best to file a PR with the information.
- There is one open PR about coalescing subregister liveness right now:  http://llvm.org/PR32688 <http://llvm.org/PR32688> (but it's hard to say without investigation whether it is the same).

- Matthias

> On Apr 24, 2017, at 9:10 AM, Johnson, Nicholas Paul via llvm-dev <llvm-dev at lists.llvm.org> wrote:
> 
> Hello,
> 
> I have a minimal testcase which crashes RegisterCoalescer in my out-of-tree target.  It only crashes in Debug builds of llc---not in Release builds.  Also, interesting to note that the x86 backend lowers this same testcase successfully.  I did a quick search of bugs.llvm.org and found no matches.
> 
> This implies that the problem is in my backend and/or how my backend interacts with RegisterCoalescer.   I'm not too familiar with RegisterCoalescer.   Can anyone give me a hint on where I should start looking for the problem in my backend?  
> 
> Thank you,
> Nick Johnson
> D. E. Shaw Research
> 
> 
> Here's my minimal testcase:
> 
> $ llvm-dis -o - bugpoint.reduced.simplified.bc
> ; ModuleID = 'bugpoint.reduced.simplified.bc'
> source_filename = "bugpoint-output-7eb0c3b.bc"
> target datalayout = "e-m:e-p:32:32-i8:8:32-i16:16:32-i32:32:32-i64:64:64-v64:64:64-v128:128:128-n32-S128"
> target triple = "desgcv3-unknown-elf"
> 
> ; Function Attrs: noreturn nounwind
> define void @wobble() local_unnamed_addr #0 {
>  %1 = extractelement <4 x i32> <i32 0, i32 2, i32 3, i32 4>, i32 undef
>  %2 = sub nsw i32 0, %1
>  %3 = icmp sgt i32 %2, 0
>  br i1 %3, label %4, label %5
> 
> ; <label>:4:                                      ; preds = %5, %0
>  unreachable
> 
> ; <label>:5:                                      ; preds = %0
>  %6 = extractelement <4 x i32> <i32 0, i32 2, i32 3, i32 4>, i32 undef
>  %7 = sub nsw i32 0, %6
>  %8 = icmp sgt i32 %7, 0
>  br i1 %8, label %9, label %4
> 
> ; <label>:9:                                      ; preds = %5
>  unreachable
> }
> 
> attributes #0 = { noreturn nounwind "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="generic" "unsafe-fp-math"="false" "use-soft-float"="false" }
> 
> 
> Here's the tail-end of the log, with debugging turned on:
> 
> $llc bugpoint.reduced.simplified.bc -debug
> ...
> 208B    %vreg13:sub_64_1<def> = COPY %vreg34:sub_64_1; VecRegs:%vreg13,%vreg34
>        Considering merging to VecRegs with %vreg34 in %vreg13
>                RHS = %vreg34 [160r,240r:0)[240r,384B:1)[400B,480r:1)[480r,496r:2)[496r,672r:3)  0 at 160r 1 at 240r 2 at 480r 3 at 496r L00000020 [240r,384B:1)[400B,672r:1)  0 at x 1 at 240r L00000010 [160r,384B:1)[400B,672r:1)  0 at x 1 at 160r L00000002 [480r,480d:1)[496r,672r:0)  0 at 496r 1 at 480r L00000001 [480r,672r:0)  0 at 480r
>                LHS = %vreg13 [96r,112r:0)[112r,208r:1)[208r,288r:2)  0 at 96r 1 at 112r 2 at 208r L00000002 [96r,96d:1)[112r,288r:0)  0 at 112r 1 at 96r L00000030 [208r,288r:1)  0 at x 1 at 208r L00000001 [96r,288r:0)  0 at 96r
>                merge %vreg13:2 at 208r into %vreg34:0 at 160r --> @160r
>                LHST = %vreg13 %vreg13 [96r,112r:0)[112r,208r:1)[208r,288r:2)  0 at 96r 1 at 112r 2 at 208r L00000002 [96r,96d:1)[112r,288r:0)  0 at 112r 1 at 96r L00000030 [208r,288r:1)  0 at x 1 at 208r L00000001 [96r,288r:0)  0 at 96r
>                Copy+Merge 00000030 into 00000020
>                Reduce Lane to 00000010
>                conflict at %vreg34:1 at 240r
> *** Couldn't join subrange!
> 
> UNREACHABLE executed at /d/en/johnsoni-0/gctools/llvm/lib/CodeGen/RegisterCoalescer.cpp:2693!
> #0 0x00000000027395af llvm::sys::PrintStackTrace(llvm::raw_ostream&) /d/en/johnsoni-0/gctools/llvm/lib/Support/Unix/Signals.inc:402:0
> #1 0x0000000002739971 PrintStackTraceSignalHandler(void*) /d/en/johnsoni-0/gctools/llvm/lib/Support/Unix/Signals.inc:466:0
> #2 0x0000000002737bce llvm::sys::RunSignalHandlers() /d/en/johnsoni-0/gctools/llvm/lib/Support/Signals.cpp:44:0
> #3 0x0000000002738f41 SignalHandler(int) /d/en/johnsoni-0/gctools/llvm/lib/Support/Unix/Signals.inc:256:0
> #4 0x000000396140f7e0 __restore_rt (/lib64/libpthread.so.0+0x396140f7e0)
> #5 0x0000003960c325e5 __GI_raise (/lib64/libc.so.6+0x3960c325e5)
> #6 0x0000003960c33dc5 __GI_abort (/lib64/libc.so.6+0x3960c33dc5)
> #7 0x00000000026d4c42 bindingsErrorHandler(void*, std::string const&, bool) /d/en/johnsoni-0/gctools/llvm/lib/Support/ErrorHandling.cpp:127:0
> #8 0x0000000001fdbc85 (anonymous namespace)::RegisterCoalescer::joinSubRegRanges(llvm::LiveRange&, llvm::LiveRange&, llvm::LaneBitmask, llvm::CoalescerPair const&) /d/en/johnsoni-0/gctools/llvm/lib/CodeGen/RegisterCoalescer.cpp:2695:0
> #9 0x0000000001fdc3df (anonymous namespace)::RegisterCoalescer::mergeSubRangeInto(llvm::LiveInterval&, llvm::LiveRange const&, llvm::LaneBitmask, llvm::CoalescerPair&) /d/en/johnsoni-0/gctools/llvm/lib/CodeGen/RegisterCoalescer.cpp:2769:0
> #10 0x0000000001fdcbd5 (anonymous namespace)::RegisterCoalescer::joinVirtRegs(llvm::CoalescerPair&) /d/en/johnsoni-0/gctools/llvm/lib/CodeGen/RegisterCoalescer.cpp:2832:0
> #11 0x0000000001fdd174 (anonymous namespace)::RegisterCoalescer::joinIntervals(llvm::CoalescerPair&) /d/en/johnsoni-0/gctools/llvm/lib/CodeGen/RegisterCoalescer.cpp:2892:0
> #12 0x0000000001fd5fc3 (anonymous namespace)::RegisterCoalescer::joinCopy(llvm::MachineInstr*, bool&) /d/en/johnsoni-0/gctools/llvm/lib/CodeGen/RegisterCoalescer.cpp:1468:0
> #13 0x0000000001fdd4b6 (anonymous namespace)::RegisterCoalescer::copyCoalesceWorkList(llvm::MutableArrayRef<llvm::MachineInstr*>) /d/en/johnsoni-0/gctools/llvm/lib/CodeGen/RegisterCoalescer.cpp:2962:0
> #14 0x0000000001fddf32 (anonymous namespace)::RegisterCoalescer::coalesceLocals() /d/en/johnsoni-0/gctools/llvm/lib/CodeGen/RegisterCoalescer.cpp:3087:0
> #15 0x0000000001fde3a1 (anonymous namespace)::RegisterCoalescer::joinAllIntervals() /d/en/johnsoni-0/gctools/llvm/lib/CodeGen/RegisterCoalescer.cpp:3121:0
> #16 0x0000000001fde6e2 (anonymous namespace)::RegisterCoalescer::runOnMachineFunction(llvm::MachineFunction&) /d/en/johnsoni-0/gctools/llvm/lib/CodeGen/RegisterCoalescer.cpp:3167:0
> #17 0x0000000001d28d87 llvm::MachineFunctionPass::runOnFunction(llvm::Function&) /d/en/johnsoni-0/gctools/llvm/lib/CodeGen/MachineFunctionPass.cpp:62:0
> ...
> 
> 
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20170424/6fad9309/attachment.html>


More information about the llvm-dev mailing list