JULIA_LLVM_ARGS="--aarch64-sve-vector-bits-min=" LLVM SegFault creating ntuple of VecElement of passed vector length

[freemin7]

When running Julia with JULIA_LLVM_ARGS="--aarch64-sve-vector-bits-min=512" precompiling VectorizationBase causes the error below without the LLVM this error does not appear.

[ Info: Precompiling VectorizationBase [3d5dd08c-fd9d-11e8-17fa-ed2836048c2f]

signal (11): Segmentation fault
in expression starting at /lustre/home/guest19/.julia/packages/VectorizationBase/xHOp9/src/special/exp.jl:200
_ZNK4llvm3EVT16isExtendedVectorEv at /lustre/home/guest19/julia-1.7.0-rc2/bin/../lib/julia/libLLVM-12jl.so (unknown line)
_ZNK4llvm3EVT21isExtended64BitVectorEv at /lustre/home/guest19/julia-1.7.0-rc2/bin/../lib/julia/libLLVM-12jl.so (unknown line)
Allocations: 18783400 (Pool: 18782916; Big: 484); GC: 38
ERROR: Failed to precompile VectorizationBase [3d5dd08c-fd9d-11e8-17fa-ed2836048c2f] to /lustre/home/guest19/.julia/compiled/v1.7/VectorizationBase/jl_k42v0A.
Stacktrace:
 [1] error(s::String)
   @ Base ./error.jl:33
 [2] compilecache(pkg::Base.PkgId, path::String, internal_stderr::IO, internal_stdout::IO, ignore_loaded_modules::Bool)
   @ Base ./loading.jl:1466
 [3] compilecache(pkg::Base.PkgId, path::String)
   @ Base ./loading.jl:1410
 [4] _require(pkg::Base.PkgId)
   @ Base ./loading.jl:1120
 [5] require(uuidkey::Base.PkgId)
   @ Base ./loading.jl:1013
 [6] require(into::Module, mod::Symbol)
   @ Base ./loading.jl:997

This weekend A64FX instances are easily accessible via the OOKAMI which provides access via the ARM HPC User group. See page 14 and 15 of these slides for login data. So anyone can reproduce it.

julia> versioninfo()
Julia Version 1.7.0-rc2
Commit f23fc0d27a (2021-10-20 12:45 UTC)
Platform Info:
  OS: Linux (aarch64-unknown-linux-gnu)
  CPU: unknown
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-12.0.1 (ORCJIT, a64fx)
Environment:
  JULIA_LLVM_ARGS = -aarch64-sve-vector-bits-min=512

The MWE is ntuple(j -> Core.VecElement(j), n) where 9>n>2. For example 3,4 or 8.

Other observations:
JULIA_LLVM_ARGS="-aarch64-sve-vector-bits-min=512" ./julia-1.7.0-rc2/bin/julia -e 'ntuple(j -> Core.VecElement(j), 8)' doesn't trigger it.
When changing the -aarch64-sve-vector-bits-min= 256 causes crashes 2<n<5 and 1024 causes crashes with 2<n<17.

[chriselrod]

For reference, the offending line within VectorizationBase is:

const TABLE_EXP_64_0= Vec(ntuple(j -> Core.VecElement(Float64(2.0^(big(j-1)/16))), Val(8)))

Which doesn't look like it should be problematic.

I wonder if this is similar to?
#32414

Maybe some alignment assumptions differ on this platform?
For example, the A64fx has 256 byte cachelines.

It's easy enough to work around within VectorizationBase, because that code is specifically for AVX512, where you can use shuffle instructions for a small-but-fast lookup table. This approach will never work or be optimal on SVE or any variable width architecture.

If you want to try out VectorizationBase, use the main and master branches of HostCPUFeatures and VectorizationBase.

Using LoopVectorization still segfaults, though.
Not related to this issue, but I got an LLVM ERROR saying

LLVM ERROR: Cannot select: 0x174bcf0: v2i64 = AArch64ISD::DUPLANE64 0xe71ab0, Constant:i64<1>, reduce.jl:411

It seems like this error is being thrown by a call to maximum([1, 8]) during compilation.

[freemin7]

I find it unlikely that this crash's is related to hardware features as the area of the parameters with which that crash occurs is related the SVE size passed via environmental variable.

And I assume for n=1,2 SVE code is not touched and NEON is used instead.

I also got a stacktrace now:

#0  0x00004000017b8fb4 in llvm::EVT::isExtendedVector() const ()
  from /lustre/home/guest19/julia-1.7.0-rc2/bin/../lib/julia/libLLVM-12jl.so
[Current thread is 1 (Thread 0x400000044090 (LWP 1606838))]
Missing separate debuginfos, use: yum debuginfo-install glibc-2.28-151.el8.aarch64
(gdb) bt
#0  0x00004000017b8fb4 in llvm::EVT::isExtendedVector() const ()
  from /lustre/home/guest19/julia-1.7.0-rc2/bin/../lib/julia/libLLVM-12jl.so
#1  0x00004000017b92c4 in llvm::EVT::isExtended64BitVector() const ()
  from /lustre/home/guest19/julia-1.7.0-rc2/bin/../lib/julia/libLLVM-12jl.so
#2  0x0000400003035030 in llvm::AArch64TargetLowering::LowerFormalArguments(llvm::SDValue, unsigned int, bool, llvm::SmallVectorImpl<llvm::ISD::InputArg> const&, llvm::SDLoc const&, llvm::SelectionDAG&, llvm::SmallVectorImpl<llvm::SDValue>&) const ()
  from /lustre/home/guest19/julia-1.7.0-rc2/bin/../lib/julia/libLLVM-12jl.so
#3  0x00004000019604b0 in llvm::SelectionDAGISel::LowerArguments(llvm::Function const&) ()
  from /lustre/home/guest19/julia-1.7.0-rc2/bin/../lib/julia/libLLVM-12jl.so
#4  0x00004000019ab514 in llvm::SelectionDAGISel::SelectAllBasicBlocks(llvm::Function const&) ()
  from /lustre/home/guest19/julia-1.7.0-rc2/bin/../lib/julia/libLLVM-12jl.so
#5  0x00004000019ac414 in llvm::SelectionDAGISel::runOnMachineFunction(llvm::MachineFunction&) [clone .part.869] () from /lustre/home/guest19/julia-1.7.0-rc2/bin/../lib/julia/libLLVM-12jl.so
#6  0x0000400001597ad4 in llvm::MachineFunctionPass::runOnFunction(llvm::Function&) ()
  from /lustre/home/guest19/julia-1.7.0-rc2/bin/../lib/julia/libLLVM-12jl.so
#7  0x00004000013505c4 in llvm::FPPassManager::runOnFunction(llvm::Function&) ()
  from /lustre/home/guest19/julia-1.7.0-rc2/bin/../lib/julia/libLLVM-12jl.so
#8  0x0000400001350d08 in llvm::FPPassManager::runOnModule(llvm::Module&) ()
  from /lustre/home/guest19/julia-1.7.0-rc2/bin/../lib/julia/libLLVM-12jl.so
#9  0x000040000134fa54 in llvm::legacy::PassManagerImpl::run(llvm::Module&) ()
  from /lustre/home/guest19/julia-1.7.0-rc2/bin/../lib/julia/libLLVM-12jl.so
#10 0x0000400000474100 in JuliaOJIT::CompilerT::operator() (this=0x0, M=...)
   at /buildworker/worker/package_linuxaarch64/build/src/jitlayers.cpp:612
#11 0x0000400002aef394 in llvm::orc::IRCompileLayer::emit(std::unique_ptr<llvm::orc::MaterializationResponsibility, std::default_delete<llvm::orc::MaterializationResponsibility> >, llvm::orc::ThreadSafeModule) () from /lustre/home/guest19/julia-1.7.0-rc2/bin/../lib/julia/libLLVM-12jl.so
#12 0x0000400002af97fc in llvm::orc::BasicIRLayerMaterializationUnit::materialize(std::unique_ptr<llvm::orc::MaterializationResponsibility, std::default_delete<llvm::orc::MaterializationResponsibility> >) () from /lustre/home/guest19/julia-1.7.0-rc2/bin/../lib/julia/libLLVM-12jl.so
#13 0x0000400002ad0fb0 in llvm::orc::ExecutionSession::materializeOnCurrentThread(std::unique_ptr<llvm::orc::MaterializationUnit, std::default_delete<llvm::orc::MaterializationUnit> >, std::unique_ptr<llvm::orc::MaterializationResponsibility, std::default_delete<llvm::orc::MaterializationResponsibility> >) () from /lustre/home/guest19/julia-1.7.0-rc2/bin/../lib/julia/libLLVM-12jl.so
#14 0x0000400002acf2e4 in std::_Function_handler<void (std::unique_ptr<llvm::orc::MaterializationUnit, std::default_delete<llvm::orc::MaterializationUnit> >, std::unique_ptr<llvm::orc::MaterializationResponsibility, std::default_delete<llvm::orc::MaterializationResponsibility> >), void (*)(std::unique_ptr<llvm::orc::MaterializationUnit, std::default_delete<llvm::orc::MaterializationUnit> >, std::unique_ptr<llvm::orc::MaterializationResponsibility, std::default_delete<llvm::orc::MaterializationResponsibility> >)>::_M_invoke(std::_Any_data const&, std::unique_ptr<llvm::orc::MaterializationUnit, std::default_delete<llvm::orc::MaterializationUnit> >&&, std::unique_ptr<llvm::orc::MaterializationResponsibility, std::default_delete<llvm::orc::MaterializationResponsibility> >&&) ()
  from /lustre/home/guest19/julia-1.7.0-rc2/bin/../lib/julia/libLLVM-12jl.so
#15 0x0000400002acfa98 in llvm::orc::ExecutionSession::dispatchOutstandingMUs() ()
  from /lustre/home/guest19/julia-1.7.0-rc2/bin/../lib/julia/libLLVM-12jl.so
#16 0x0000400002ad75cc in llvm::orc::ExecutionSession::OL_completeLookup(std::unique_ptr<llvm::orc::InProgressLookupState, std::default_delete<llvm::orc::InProgressLookupState> >, std::shared_ptr<llvm::orc::AsynchronousSymbolQuery>, std::function<void (llvm::DenseMap<llvm::orc::JITDylib*, llvm::DenseSet<llvm::orc::SymbolStringPtr, llvm::DenseMapInfo<llvm::orc::SymbolStringPtr> >, llvm::DenseMapInfo<llvm::orc::JITDylib*>, llvm::detail::DenseMapPair<llvm::orc::JITDylib*, llvm::DenseSet<llvm::orc::SymbolStringPtr, llvm::DenseMapInfo<llvm::orc::SymbolStringPtr> > > > const&)>) ()
  from /lustre/home/guest19/julia-1.7.0-rc2/bin/../lib/julia/libLLVM-12jl.so
--Type <RET> for more, q to quit, c to continue without paging--
#17 0x0000400002ad7ac4 in llvm::orc::InProgressFullLookupState::complete(std::unique_ptr<llvm::orc::InProgressLookupState, std::default_delete<llvm::orc::InProgressLookupState> >) ()
  from /lustre/home/guest19/julia-1.7.0-rc2/bin/../lib/julia/libLLVM-12jl.so
#18 0x0000400002ac8f7c in llvm::orc::ExecutionSession::OL_applyQueryPhase1(std::unique_ptr<llvm::orc::InProgressLookupState, std::default_delete<llvm::orc::InProgressLookupState> >, llvm::Error) ()
  from /lustre/home/guest19/julia-1.7.0-rc2/bin/../lib/julia/libLLVM-12jl.so
#19 0x0000400002acfd64 in llvm::orc::ExecutionSession::lookup(llvm::orc::LookupKind, std::vector<std::pair<llvm::orc::JITDylib*, llvm::orc::JITDylibLookupFlags>, std::allocator<std::pair<llvm::orc::JITDylib*, llvm::orc::JITDylibLookupFlags> > > const&, llvm::orc::SymbolLookupSet, llvm::orc::SymbolState, llvm::unique_function<void (llvm::Expected<llvm::DenseMap<llvm::orc::SymbolStringPtr, llvm::JITEvaluatedSymbol, llvm::DenseMapInfo<llvm::orc::SymbolStringPtr>, llvm::detail::DenseMapPair<llvm::orc::SymbolStringPtr, llvm::JITEvaluatedSymbol> > >)>, std::function<void (llvm::DenseMap<llvm::orc::JITDylib*, llvm::DenseSet<llvm::orc::SymbolStringPtr, llvm::DenseMapInfo<llvm::orc::SymbolStringPtr> >, llvm::DenseMapInfo<llvm::orc::JITDylib*>, llvm::detail::DenseMapPair<llvm::orc::JITDylib*, llvm::DenseSet<llvm::orc::SymbolStringPtr, llvm::DenseMapInfo<llvm::orc::SymbolStringPtr> > > > const&)>) () from /lustre/home/guest19/julia-1.7.0-rc2/bin/../lib/julia/libLLVM-12jl.so
#20 0x0000400002ad03f0 in llvm::orc::ExecutionSession::lookup(std::vector<std::pair<llvm::orc::JITDylib*, llvm::orc::JITDylibLookupFlags>, std::allocator<std::pair<llvm::orc::JITDylib*, llvm::orc::JITDylibLookupFlags> > > const&, llvm::orc::SymbolLookupSet const&, llvm::orc::LookupKind, llvm::orc::SymbolState, std::function<void (llvm::DenseMap<llvm::orc::JITDylib*, llvm::DenseSet<llvm::orc::SymbolStringPtr, llvm::DenseMapInfo<llvm::orc::SymbolStringPtr> >, llvm::DenseMapInfo<llvm::orc::JITDylib*>, llvm::detail::DenseMapPair<llvm::orc::JITDylib*, llvm::DenseSet<llvm::orc::SymbolStringPtr, llvm::DenseMapInfo<llvm::orc::SymbolStringPtr> > > > const&)>) ()
  from /lustre/home/guest19/julia-1.7.0-rc2/bin/../lib/julia/libLLVM-12jl.so
#21 0x0000400002ad074c in llvm::orc::ExecutionSession::lookup(std::vector<std::pair<llvm::orc::JITDylib*, llvm::orc::JITDylibLookupFlags>, std::allocator<std::pair<llvm::orc::JITDylib*, llvm::orc::JITDylibLookupFlags> > > const&, llvm::orc::SymbolStringPtr, llvm::orc::SymbolState) ()
  from /lustre/home/guest19/julia-1.7.0-rc2/bin/../lib/julia/libLLVM-12jl.so
#22 0x0000400002ad0b40 in llvm::orc::ExecutionSession::lookup(llvm::ArrayRef<llvm::orc::JITDylib*>, llvm::orc::SymbolStringPtr, llvm::orc::SymbolState) ()
  from /lustre/home/guest19/julia-1.7.0-rc2/bin/../lib/julia/libLLVM-12jl.so
#23 0x0000400002ad0d74 in llvm::orc::ExecutionSession::lookup(llvm::ArrayRef<llvm::orc::JITDylib*>, llvm::StringRef, llvm::orc::SymbolState) ()
  from /lustre/home/guest19/julia-1.7.0-rc2/bin/../lib/julia/libLLVM-12jl.so
#24 0x00004000004767b8 in JuliaOJIT::addModule (this=0x0, M=...)
   at /buildworker/worker/package_linuxaarch64/build/src/jitlayers.cpp:779
#25 0x0000400000476b38 in jl_add_to_ee (m=...)
   at /buildworker/worker/package_linuxaarch64/build/src/jitlayers.cpp:1058
#26 0x00004000004772a4 in jl_add_to_ee (M=..., NewExports=..., Queued=..., ToMerge=..., 
   depth=depth@entry=1) at /buildworker/worker/package_linuxaarch64/build/src/jitlayers.cpp:1102
#27 0x0000400000477e4c in jl_add_to_ee (NewExports=..., M=...)
   at /buildworker/worker/package_linuxaarch64/build/src/jitlayers.cpp:1124
#28 _jl_compile_codeinst (codeinst=codeinst@entry=0x400005eeca30, src=<optimized out>, 
   world=world@entry=31307)
   at /buildworker/worker/package_linuxaarch64/build/src/jitlayers.cpp:154
#29 0x0000400000478c04 in jl_generate_fptr (mi=mi@entry=0x400005389be0, world=world@entry=31307)
   at /buildworker/worker/package_linuxaarch64/build/src/jitlayers.cpp:350
#30 0x00004000003ecab8 in jl_compile_method_internal (mi=mi@entry=0x400005389be0, 
   world=world@entry=31307) at /buildworker/worker/package_linuxaarch64/build/src/gf.c:1980
#31 0x00004000003ef1f4 in jl_compile_method_internal (world=31307, mi=0x400005389be0)
   at /buildworker/worker/package_linuxaarch64/build/src/gf.c:2247
#32 _jl_invoke (world=31307, mfunc=<optimized out>, nargs=3, args=0x400029cfe870, 
--Type <RET> for more, q to quit, c to continue without paging--
   F=0x40000b84d2c0 <jl_system_image_data+13377856>)
   at /buildworker/worker/package_linuxaarch64/build/src/gf.c:2239
#33 jl_apply_generic (F=0x40000b84d2c0 <jl_system_image_data+13377856>, args=0x400029cfe870, 
   nargs=3) at /buildworker/worker/package_linuxaarch64/build/src/gf.c:2429
#34 0x0000400009a64020 in julia_#43_46692 ()
   at /buildworker/worker/package_linuxaarch64/build/usr/share/julia/stdlib/v1.7/REPL/src/REPL.jl:264
#35 0x0000400009a640dc in jfptr_YY.43_46693 () at dict.jl:481
#36 0x00004000003ef148 in _jl_invoke (world=31307, mfunc=<optimized out>, nargs=1, 
   args=0x400029cfe9d0, F=0x40000538fa70)
   at /buildworker/worker/package_linuxaarch64/build/src/gf.c:2247
#37 jl_apply_generic (F=0x40000538fa70, args=0x400029cfe9d0, nargs=1)
   at /buildworker/worker/package_linuxaarch64/build/src/gf.c:2429
#38 0x0000400009885a14 in japi1_with_repl_linfo_47053 ()
   at /buildworker/worker/package_linuxaarch64/build/usr/share/julia/stdlib/v1.7/REPL/src/REPL.jl:508
#39 0x00004000003ef148 in _jl_invoke (world=31307, mfunc=<optimized out>, nargs=2, 
   args=0x400029cfeb00, F=0x40000bea4e80 <jl_system_image_data+20028672>)
   at /buildworker/worker/package_linuxaarch64/build/src/gf.c:2247
#40 jl_apply_generic (F=0x40000bea4e80 <jl_system_image_data+20028672>, args=0x400029cfeb00, 
   nargs=2) at /buildworker/worker/package_linuxaarch64/build/src/gf.c:2429
#41 0x00004000098f24c0 in japi1_display_28922 ()
   at /buildworker/worker/package_linuxaarch64/build/usr/share/julia/stdlib/v1.7/REPL/src/REPL.jl:257
#42 0x00004000098f252c in japi1_display_28920 ()
   at /buildworker/worker/package_linuxaarch64/build/usr/share/julia/stdlib/v1.7/REPL/src/REPL.jl:269
#43 0x00004000003ef148 in _jl_invoke (world=31307, mfunc=<optimized out>, nargs=2, 
   args=0x400029cfee08, F=0x40000bea3c20 <jl_system_image_data+20023968>)
   at /buildworker/worker/package_linuxaarch64/build/src/gf.c:2247
#44 jl_apply_generic (F=0x40000bea3c20 <jl_system_image_data+20023968>, args=0x400029cfee08, 
   nargs=2) at /buildworker/worker/package_linuxaarch64/build/src/gf.c:2429
#45 0x0000400009ef0024 in japi1_display_28915 () at multimedia.jl:328
#46 0x00004000003ef148 in _jl_invoke (world=31307, mfunc=<optimized out>, nargs=1, 
   args=0x400029cff350, F=0x40000bea3c20 <jl_system_image_data+20023968>)
   at /buildworker/worker/package_linuxaarch64/build/src/gf.c:2247
#47 jl_apply_generic (F=0x40000bea3c20 <jl_system_image_data+20023968>, args=0x400029cff350, 
   nargs=1) at /buildworker/worker/package_linuxaarch64/build/src/gf.c:2429
#48 0x00004000003ff76c in jl_apply (nargs=2, args=0x400029cff348)
   at /buildworker/worker/package_linuxaarch64/build/src/julia.h:1790
#49 jl_f__call_latest (F=<optimized out>, args=0x400029cff348, nargs=2)
   at /buildworker/worker/package_linuxaarch64/build/src/builtins.c:757
#50 0x0000400009832a08 in julia_print_response_46648 ()
   at /buildworker/worker/package_linuxaarch64/build/usr/share/julia/stdlib/v1.7/REPL/src/REPL.jl:315
#51 0x0000400009a66e70 in julia_#45_46750 ()
   at /buildworker/worker/package_linuxaarch64/build/usr/share/julia/stdlib/v1.7/REPL/src/REPL.jl:275
#52 0x0000400009a66eb0 in jfptr_YY.45_46751 () at dict.jl:481
#53 0x00004000003ef148 in _jl_invoke (world=31306, mfunc=<optimized out>, nargs=1, 
   args=0x400029cff540, F=0x40000539b790)
   at /buildworker/worker/package_linuxaarch64/build/src/gf.c:2247
--Type <RET> for more, q to quit, c to continue without paging--
#54 jl_apply_generic (F=0x40000539b790, args=0x400029cff540, nargs=1)
   at /buildworker/worker/package_linuxaarch64/build/src/gf.c:2429
#55 0x0000400009885a14 in japi1_with_repl_linfo_47053 ()
   at /buildworker/worker/package_linuxaarch64/build/usr/share/julia/stdlib/v1.7/REPL/src/REPL.jl:508
#56 0x00004000003ef148 in _jl_invoke (world=31306, mfunc=<optimized out>, nargs=2, 
   args=0x400029cff678, F=0x40000bea4e80 <jl_system_image_data+20028672>)
   at /buildworker/worker/package_linuxaarch64/build/src/gf.c:2247
#57 jl_apply_generic (F=0x40000bea4e80 <jl_system_image_data+20028672>, args=0x400029cff678, 
   nargs=2) at /buildworker/worker/package_linuxaarch64/build/src/gf.c:2429
#58 0x0000400009839ba0 in julia_print_response_46653 ()
   at /buildworker/worker/package_linuxaarch64/build/usr/share/julia/stdlib/v1.7/REPL/src/REPL.jl:273
#59 0x000040000981f908 in julia_do_respond_47126 ()
   at /buildworker/worker/package_linuxaarch64/build/usr/share/julia/stdlib/v1.7/REPL/src/REPL.jl:844
#60 0x000040000981fa8c in jfptr_do_respond_47127 ()
  from /lustre/home/guest19/julia-1.7.0-rc2/lib/julia/sys.so
#61 0x00004000003ef148 in _jl_invoke (world=31306, mfunc=<optimized out>, nargs=3, 
   args=0x400029cffb58, F=0x400004aedf00)
   at /buildworker/worker/package_linuxaarch64/build/src/gf.c:2247
#62 jl_apply_generic (F=0x400004aedf00, args=0x400029cffb58, nargs=3)
   at /buildworker/worker/package_linuxaarch64/build/src/gf.c:2429
#63 0x00004000003ff76c in jl_apply (nargs=4, args=0x400029cffb50)
   at /buildworker/worker/package_linuxaarch64/build/src/julia.h:1790
#64 jl_f__call_latest (F=<optimized out>, args=0x400029cffb50, nargs=4)
   at /buildworker/worker/package_linuxaarch64/build/src/builtins.c:757
#65 0x00004000097f549c in #invokelatest#2 () at essentials.jl:716
#66 invokelatest () at essentials.jl:714
#67 julia_run_interface_47732 ()
   at /buildworker/worker/package_linuxaarch64/build/usr/share/julia/stdlib/v1.7/REPL/src/LineEdit.jl:2493
#68 0x00004000097f585c in jfptr_run_interface_47733 ()
   at /buildworker/worker/package_linuxaarch64/build/usr/share/julia/stdlib/v1.7/REPL/src/LineEdit.jl:2493
#69 0x00004000003ef148 in _jl_invoke (world=31306, mfunc=<optimized out>, nargs=3, 
   args=0x400029cffd08, F=0x40000be5ff10 <jl_system_image_data+19746192>)
   at /buildworker/worker/package_linuxaarch64/build/src/gf.c:2247
#70 jl_apply_generic (F=0x40000be5ff10 <jl_system_image_data+19746192>, args=0x400029cffd08, 
   nargs=3) at /buildworker/worker/package_linuxaarch64/build/src/gf.c:2429
#71 0x0000400009803204 in julia_run_frontend_47192 ()
   at /buildworker/worker/package_linuxaarch64/build/usr/share/julia/stdlib/v1.7/REPL/src/REPL.jl:1230
#72 0x00004000098036a0 in julia_#49_47024 () at task.jl:411
#73 0x00004000098036b0 in jfptr_YY.49_47025 ()
   at /buildworker/worker/package_linuxaarch64/build/usr/share/julia/stdlib/v1.7/REPL/src/LineEdit.jl:2493
#74 0x00004000003ef148 in _jl_invoke (world=31306, mfunc=<optimized out>, nargs=0, 
   args=0x4000049809f8, F=0x400004bc0cb0)
   at /buildworker/worker/package_linuxaarch64/build/src/gf.c:2247
#75 jl_apply_generic (
../../gdb/inline-frame.c:160: internal-error: void inline_frame_this_id(frame_info*, void**, frame_id*): Assertion `!frame_id_eq (*this_id, outer_frame_id)' failed.

the build-in rr doesn't seem to work.

[vchuravy]

If you can get the LLVM IR from the Module, that would help.

[freemin7]

I only got core dumps now. I had access to the machine the weekend I created the issue. Would those be helpful and how could I get them to you?

[giordano]

Precompilation of VectorizationBase is now successful (tested with Julia v1.7.2), but now I get a runtime crash (signal 6) when running functions with @turbo.

[giordano]

Actually, nightly build of Julia works also at runtime:

$ JULIA_LLVM_ARGS="-aarch64-sve-vector-bits-min=512" ./julia-7889b2a6a2/bin/julia -q
julia> versioninfo()
Julia Version 1.8.0-DEV.1572
Commit 7889b2a6a2 (2022-02-16 21:17 UTC)
Platform Info:
  OS: Linux (aarch64-unknown-linux-gnu)
  CPU: 48 × unknown
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-13.0.1 (ORCJIT, a64fx)
  Threads: 1 on 48 virtual cores

julia> using LoopVectorization, BenchmarkTools

julia> function mydotavx(a, b)
          s = 0.0
          @turbo for i ∈ eachindex(a,b)
              s += a[i]*b[i]
          end
          s
       end
mydotavx (generic function with 1 method)

julia> @btime mydotavx(a, b) setup=(a=fill(2.0, 256); b=fill(3.14, 256))
  209.311 ns (0 allocations: 0 bytes)
1607.68

Closing the issue.