[Bug]: backend cannot select in neon code

[georgthegreat]

Description

Hi.

I am trying to update xnnpack to 2023-04-18 / 71f70de09ca17f125d3c24ccceeb12045a674d0c in order to update tensorflow-lite to 2.12.0.

Unfortunately clang for android segfaults when compiling src/bf16-gemm/gen/bf16-gemm-1x4c8-minmax-neonbf16-bfdot.c for android-armv7a.

I am able to reproduce this:

${NDK}/llvm-toolchain/bin/clang --target=armv7a-linux-androideabi21 -isystem ${NDK}/sources/cxx-stl/llvm-libc++abi/include -fsigned-char -march=armv7-a -mfloat-abi=softfp -c -o ${SRC}/contrib/restricted/google/xnnpack/src/bf16-gemm/gen/bf16-gemm-6x8c2-minmax-neonbf16-bfdot-lane-ld128.c.o -I${SRC} -I${SRC}/contrib/libs/fxdiv -I${SRC}/contrib/libs/pthreadpool/include -I${SRC}/contrib/restricted/google/xnnpack/include -I${SRC}/contrib/libs/fxdiv/include -I${SRC}/contrib/restricted/cpuinfo/include -I${SRC}/contrib/restricted/google/xnnpack/src -I${SRC}/contrib/libs/cxxsupp/libcxx/include -I${SRC}/contrib/libs/fp16/include -I${SRC}/contrib/libs/zlib/include -I${SRC}/contrib/libs/double-conversion -I${SRC}/contrib/libs/libc_compat/include/ifaddrs -fdebug-prefix-map=${SRC}=/-B -Xclang -fdebug-compilation-dir -Xclang /tmp -pipe -O3 -g -fexceptions -fno-common -fcolor-diagnostics -faligned-allocation -fdebug-default-version=4 -ffunction-sections -fdata-sections -w  -DXNN_ENABLE_ARM_BF16=1 -DXNN_ENABLE_ARM_DOTPROD=1 -DXNN_ENABLE_ARM_FP16_SCALAR=1 -DXNN_ENABLE_ARM_FP16_VECTOR=1 -DXNN_ENABLE_ASSEMBLY=1 -DXNN_ENABLE_DWCONV_MULTIPASS=0 -DXNN_ENABLE_GEMM_M_SPECIALIZATION=1 -DXNN_ENABLE_JIT=0 -DXNN_ENABLE_MEMOPT=1 -DXNN_ENABLE_RISCV_VECTOR=1 -DXNN_ENABLE_SPARSE=1 -DXNN_LOG_LEVEL=0 -Werror-implicit-function-declaration -nostdinc++ -fPIE -fPIE -march=armv8.2-a+bf16 -mfpu=neon-fp-armv8 ${SRC}/contrib/restricted/google/xnnpack/src/bf16-gemm/gen/bf16-gemm-6x8c2-minmax-neonbf16-bfdot-lane-ld128.c

I have tested NDK's r25b and r25c, but the segfault still persists and there is no way to fix this.

fatal error: error in backend: Cannot select: 0x90a21a8: v4bf16 = extract_subvector 0x909e178, Constant:i32<0>, /home/thegeorg/arcadia/contrib/restricted/google/xnnpack/src/bf16-gemm/gen/bf
16-gemm-6x8c2-minmax-neonbf16-bfdot-lane-ld128.c:173:33
  0x909e178: v8bf16,i32,ch = ARMISD::VLD1_UPD<(load (s128) from %ir.362, align 2)> 0x8f06068, 0x8cafc48, 0x8cb05a0, Constant:i32<1>, /home/thegeorg/arcadia/contrib/restricted/google/xnnpack
/src/bf16-gemm/gen/bf16-gemm-6x8c2-minmax-neonbf16-bfdot-lane-ld128.c:163:32
    0x8cafc48: i32,ch = CopyFromReg 0x8f06068, Register:i32 %84, /home/thegeorg/arcadia/contrib/restricted/google/xnnpack/src/bf16-gemm/gen/bf16-gemm-6x8c2-minmax-neonbf16-bfdot-lane-ld128.
c:163:32
      0x8cb0260: i32 = Register %84
    0x8cb05a0: i32,ch = CopyFromReg 0x8f06068, Register:i32 %70, /home/thegeorg/arcadia/contrib/restricted/google/xnnpack/src/bf16-gemm/gen/bf16-gemm-6x8c2-minmax-neonbf16-bfdot-lane-ld128.
c:158:89
      0x8cafe50: i32 = Register %70
    0x909e380: i32 = Constant<1>
  0x90b3a08: i32 = Constant<0>
In function: xnn_bf16_gemm_minmax_ukernel_6x8c2__neonbf16_bfdot_lane_ld128
PLEASE submit a bug report to https://github.com/android-ndk/ndk/issues and include the crash backtrace, preprocessed source, and associated run script.

I have tested ndk r25b and r25c, but the issue persists.
Classic clang 14.0.x does not seems to be affected by the problem.

Upstream bug

No response

Commit to cherry-pick

No response

Affected versions

r25

Canary version

No response

Host OS

Linux

Host OS version

Ubuntu 20.04

Affected ABIs

armeabi-v7a

[georgthegreat]

Сorresponding xnnpack can be found here:
google/XNNPACK#4775

[pirama-arumuga-nainar]

PLEASE submit a bug report to https://github.com/android-ndk/ndk/issues and include the crash backtrace, preprocessed source, and associated run script.

Can you attach this? It'd let me reproduce the issue without checking-out the xnnpack repo.

[leleliu008]

I'm facing this too.

FAILED: CMakeFiles/microkernels-all.dir/src/bf16-gemm/gen/bf16-gemm-1x4c8-minmax-neonbf16-bfdot.c.o 
/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang --target=armv7-none-linux-androideabi21 --sysroot=/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/sysroot -DXNN_ENABLE_ARM_BF16=1 -DXNN_ENABLE_ARM_DOTPROD=1 -DXNN_ENABLE_ARM_FP16_SCALAR=1 -DXNN_ENABLE_ARM_FP16_VECTOR=1 -DXNN_ENABLE_ASSEMBLY=1 -DXNN_ENABLE_DWCONV_MULTIPASS=0 -DXNN_ENABLE_GEMM_M_SPECIALIZATION=1 -DXNN_ENABLE_JIT=0 -DXNN_ENABLE_MEMOPT=1 -DXNN_ENABLE_RISCV_VECTOR=1 -DXNN_ENABLE_SPARSE=1 -I/home/leleliu008/.ndk-pkg/installing/android-21/armeabi-v7a/xnnpack/src/src --sysroot /home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/sysroot -Qunused-arguments -fPIC -fno-common -Os -I/home/leleliu008/.ndk-pkg/installing/android-21/armeabi-v7a/xnnpack/include -I/home/leleliu008/.ndk-pkg/installed/android-21/armeabi-v7a/pthreadpool/include -I/home/leleliu008/.ndk-pkg/installed/android-21/armeabi-v7a/fxdiv/include -I/home/leleliu008/.ndk-pkg/installed/android-21/armeabi-v7a/fp16/include -I/home/leleliu008/.ndk-pkg/installed/android-21/armeabi-v7a/cpuinfo/include --sysroot /home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/sysroot -Qunused-arguments -include /home/leleliu008/.ndk-pkg/installing/android-21/armeabi-v7a/xnnpack/include/ndk-pkg-common.h -DNDEBUG -O3 -DNDEBUG -fPIC -O2 -std=c11  -fno-math-errno  -marm  -march=armv8.2-a+bf16 -mfpu=neon-fp-armv8 -MD -MT CMakeFiles/microkernels-all.dir/src/bf16-gemm/gen/bf16-gemm-1x4c8-minmax-neonbf16-bfdot.c.o -MF CMakeFiles/microkernels-all.dir/src/bf16-gemm/gen/bf16-gemm-1x4c8-minmax-neonbf16-bfdot.c.o.d -o CMakeFiles/microkernels-all.dir/src/bf16-gemm/gen/bf16-gemm-1x4c8-minmax-neonbf16-bfdot.c.o -c /home/leleliu008/.ndk-pkg/installing/android-21/armeabi-v7a/xnnpack/src/src/bf16-gemm/gen/bf16-gemm-1x4c8-minmax-neonbf16-bfdot.c
fatal error: error in backend: Cannot select: 0x98f7608: v4bf16 = ARMISD::VEXT 0x986fde8, 0x986fde8, Constant:i32<2>
  0x986fde8: v4bf16,ch = CopyFromReg 0x96c7878, Register:v4bf16 %54
    0x9aa05e0: v4bf16 = Register %54
  0x986fde8: v4bf16,ch = CopyFromReg 0x96c7878, Register:v4bf16 %54
    0x9aa05e0: v4bf16 = Register %54
  0x9870260: i32 = Constant<2>
In function: xnn_bf16_gemm_minmax_ukernel_1x4c8__neonbf16_bfdot
PLEASE submit a bug report to https://github.com/android-ndk/ndk/issues and include the crash backtrace, preprocessed source, and associated run script.
Stack dump:
0.	Program arguments: /home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang --target=armv7-none-linux-androideabi21 --sysroot=/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/sysroot -DXNN_ENABLE_ARM_BF16=1 -DXNN_ENABLE_ARM_DOTPROD=1 -DXNN_ENABLE_ARM_FP16_SCALAR=1 -DXNN_ENABLE_ARM_FP16_VECTOR=1 -DXNN_ENABLE_ASSEMBLY=1 -DXNN_ENABLE_DWCONV_MULTIPASS=0 -DXNN_ENABLE_GEMM_M_SPECIALIZATION=1 -DXNN_ENABLE_JIT=0 -DXNN_ENABLE_MEMOPT=1 -DXNN_ENABLE_RISCV_VECTOR=1 -DXNN_ENABLE_SPARSE=1 -I/home/leleliu008/.ndk-pkg/installing/android-21/armeabi-v7a/xnnpack/src/src --sysroot /home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/sysroot -Qunused-arguments -fPIC -fno-common -Os -I/home/leleliu008/.ndk-pkg/installing/android-21/armeabi-v7a/xnnpack/include -I/home/leleliu008/.ndk-pkg/installed/android-21/armeabi-v7a/pthreadpool/include -I/home/leleliu008/.ndk-pkg/installed/android-21/armeabi-v7a/fxdiv/include -I/home/leleliu008/.ndk-pkg/installed/android-21/armeabi-v7a/fp16/include -I/home/leleliu008/.ndk-pkg/installed/android-21/armeabi-v7a/cpuinfo/include --sysroot /home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/sysroot -Qunused-arguments -include /home/leleliu008/.ndk-pkg/installing/android-21/armeabi-v7a/xnnpack/include/ndk-pkg-common.h -DNDEBUG -O3 -DNDEBUG -fPIC -O2 -std=c11 -fno-math-errno -marm -march=armv8.2-a+bf16 -mfpu=neon-fp-armv8 -MD -MT CMakeFiles/microkernels-all.dir/src/bf16-gemm/gen/bf16-gemm-1x4c8-minmax-neonbf16-bfdot.c.o -MF CMakeFiles/microkernels-all.dir/src/bf16-gemm/gen/bf16-gemm-1x4c8-minmax-neonbf16-bfdot.c.o.d -o CMakeFiles/microkernels-all.dir/src/bf16-gemm/gen/bf16-gemm-1x4c8-minmax-neonbf16-bfdot.c.o -c /home/leleliu008/.ndk-pkg/installing/android-21/armeabi-v7a/xnnpack/src/src/bf16-gemm/gen/bf16-gemm-1x4c8-minmax-neonbf16-bfdot.c
1.	<eof> parser at end of file
2.	Code generation
3.	Running pass 'Function Pass Manager' on module '/home/leleliu008/.ndk-pkg/installing/android-21/armeabi-v7a/xnnpack/src/src/bf16-gemm/gen/bf16-gemm-1x4c8-minmax-neonbf16-bfdot.c'.
4.	Running pass 'ARM Instruction Selection' on function '@xnn_bf16_gemm_minmax_ukernel_1x4c8__neonbf16_bfdot'
 #0 0x00000000047d91d8 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x47d91d8)
 #1 0x00000000047d8340 llvm::sys::RunSignalHandlers() (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x47d8340)
 #2 0x00000000047a3dc3 (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x47a3dc3)
 #3 0x00000000047a3d7b (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x47a3d7b)
 #4 0x00000000047d7a87 llvm::sys::Process::Exit(int, bool) (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x47d7a87)
 #5 0x00000000040dc70a (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x40dc70a)
 #6 0x0000000003083072 llvm::report_fatal_error(llvm::Twine const&, bool) (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x3083072)
 #7 0x000000000282b5f5 llvm::SelectionDAGISel::CannotYetSelect(llvm::SDNode*) (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x282b5f5)
 #8 0x0000000006cf4e77 (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x6cf4e77)
 #9 0x000000000641f425 (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x641f425)
#10 0x0000000005e86b63 llvm::SelectionDAGISel::DoInstructionSelection() (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x5e86b63)
#11 0x0000000005e8710a llvm::SelectionDAGISel::CodeGenAndEmitDAG() (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x5e8710a)
#12 0x0000000006417d3c llvm::SelectionDAGISel::SelectAllBasicBlocks(llvm::Function const&) (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x6417d3c)
#13 0x0000000006457ad3 llvm::SelectionDAGISel::runOnMachineFunction(llvm::MachineFunction&) (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x6457ad3)
#14 0x00000000064572df (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x64572df)
#15 0x0000000005d9faea llvm::MachineFunctionPass::runOnFunction(llvm::Function&) (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x5d9faea)
#16 0x0000000005da0113 llvm::FPPassManager::runOnFunction(llvm::Function&) (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x5da0113)
#17 0x0000000005d9fc6f llvm::FPPassManager::runOnModule(llvm::Module&) (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x5d9fc6f)
#18 0x00000000063aa794 llvm::legacy::PassManagerImpl::run(llvm::Module&) (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x63aa794)
#19 0x00000000065d6968 clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, std::__1::unique_ptr<llvm::raw_pwrite_stream, std::__1::default_delete<llvm::raw_pwrite_stream> >) (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x65d6968)
#20 0x00000000060524d5 (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x60524d5)
#21 0x0000000005ea25a9 clang::ParseAST(clang::Sema&, bool, bool) (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x5ea25a9)
#22 0x00000000063c128d clang::FrontendAction::Execute() (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x63c128d)
#23 0x00000000063c112d clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x63c112d)
#24 0x00000000063c1541 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x63c1541)
#25 0x00000000066a9f54 cc1_main(llvm::ArrayRef<char const*>, char const*, void*) (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x66a9f54)
#26 0x00000000066a6de3 (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x66a6de3)
#27 0x00000000066a6c92 (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x66a6c92)
#28 0x00000000066a6c61 llvm::CrashRecoveryContext::RunSafely(llvm::function_ref<void ()>) (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x66a6c61)
#29 0x00000000066a69f4 clang::driver::CC1Command::Execute(llvm::ArrayRef<llvm::Optional<llvm::StringRef> >, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> >*, bool*) const (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x66a69f4)
#30 0x00000000066a685f clang::driver::Compilation::ExecuteCommand(clang::driver::Command const&, clang::driver::Command const*&) const (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x66a685f)
#31 0x00000000066a66f2 clang::driver::Driver::ExecuteCompilation(clang::driver::Compilation&, llvm::SmallVectorImpl<std::__1::pair<int, clang::driver::Command const*> >&) (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x66a66f2)
#32 0x00000000066752ee main (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x66752ee)
#33 0x00007f6a26c23a90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:74:3
#34 0x00007f6a26c23b49 call_init ./csu/../csu/libc-start.c:128:20
#35 0x00007f6a26c23b49 __libc_start_main ./csu/../csu/libc-start.c:347:5
#36 0x00000000064cce69 _start (/home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin/clang+0x64cce69)
clang-14: error: clang frontend command failed with exit code 70 (use -v to see invocation)
Android (9352603, based on r450784d1) clang version 14.0.7 (https://android.googlesource.com/toolchain/llvm-project 4c603efb0cca074e9238af8b4106c30add4418f6)
Target: armv7-none-linux-android21
Thread model: posix
InstalledDir: /home/leleliu008/Downloads/android-ndk-r25c/toolchains/llvm/prebuilt/linux-x86_64/bin
clang-14: note: diagnostic msg: 
********************

PLEASE ATTACH THE FOLLOWING FILES TO THE BUG REPORT:
Preprocessed source(s) and associated run script(s) are located at:
clang-14: note: diagnostic msg: /tmp/bf16-gemm-1x4c8-minmax-neonbf16-bfdot-0b16c9.c
clang-14: note: diagnostic msg: /tmp/bf16-gemm-1x4c8-minmax-neonbf16-bfdot-0b16c9.sh
clang-14: note: diagnostic msg: 

********************
ninja: build stopped: subcommand failed.

bf16-gemm-1x4c8-minmax-neonbf16-bfdot-0b16c9.tar.gz

[georgthegreat]

Can you attach this? It'd let me reproduce the issue without checking-out the xnnpack repo.

I am struggling to reproduce this as we are using custom build system.
Will attach the files later if this will still be mandated.

[georgthegreat]

This is my reproducer:

clang.tar.gz

[pirama-arumuga-nainar]

Does it reproduce with the toolchain in r26 canary? I can reproduce with r25 but r26 and ToT clang give several errors that look like below:

/home/thegeorg/.ya/tools/v4/4444634769/toolchains/llvm/prebuilt/linux-x86_64/lib64/clang/14.0.7/include/arm_neon.h:39533:11: error: invalid conversion between vector type 'float32x4_t' (vector of 4 'float32_t' 
values) and integer type 'int' of different size                                                         
  __ret = (float32x4_t) __builtin_neon_vbfdotq_v((int8x16_t)__p0, (int8x16_t)__p1, (int8x16_t)__p2, 41);                                                                                                                    ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

This maybe a forward compatibility issue with the pre-processed headers.

[appujee]

It might be easier to work with a preprocessed file(.I).

[georgthegreat]

Does it reproduce with the toolchain in r26 canary?

As discussed in XNNPACK issue, it looks like the issue does not reproduce with clang16.