Ubuntu
pytorch package

pytorch 2.1.2+dfsg-4 FTBFS with llvm 18

Bug #2067720 reported by Benjamin Drung on 2024-05-31

This bug affects 1 person

Affects		Status	Importance	Assigned to	Milestone
	pytorch (Ubuntu)	Fix Released	Undecided	Sergio Durigan Junior

Bug Description

pytorch 2.1.2+dfsg-4 fails to build from source with llvm 18:

```
[957/1892] /usr/bin/c++ -DAT_PER_OPERATOR_HEADERS -DCAFFE2_BUILD_MAIN_LIB -DFMT_HEADER_ONLY=1 -DHAVE_MALLOC_USABLE_SIZE=1 -DHAVE_MMAP=1 -DHAVE_SHM_OPEN=1 -DHAVE_SHM_UNLINK=1 -DMINIZ_DISABLE_ZIP_READER_CRC32_CHECKS -DONNXIFI_ENABLE_EXT=1 -DONNX_ML=1 -DONNX_NAMESPACE=onnx -DTORCH_ENABLE_LLVM -DUSE_C10D_GLOO -DUSE_DISTRIBUTED -DUSE_EXTERNAL_MZCRC -DUSE_RPC -DUSE_TENSORPIPE -D_FILE_OFFSET_BITS=64 -Dtorch_cpu_EXPORTS -I/<<PKGBUILDDIR>>/build/aten/src -I/<<PKGBUILDDIR>>/aten/src -I/<<PKGBUILDDIR>>/build -I/<<PKGBUILDDIR>> -I/<<PKGBUILDDIR>>/cmake/../third_party/benchmark/include -I/usr/lib/llvm-18/include -I/<<PKGBUILDDIR>>/debian/foxi -I/<<PKGBUILDDIR>>/build/debian/foxi -I/<<PKGBUILDDIR>>/torch/csrc/api -I/<<PKGBUILDDIR>>/torch/csrc/api/include -I/<<PKGBUILDDIR>>/caffe2/aten/src/TH -I/<<PKGBUILDDIR>>/build/caffe2/aten/src/TH -I/<<PKGBUILDDIR>>/build/caffe2/aten/src -I/<<PKGBUILDDIR>>/build/caffe2/../aten/src -I/<<PKGBUILDDIR>>/torch/csrc -I/<<PKGBUILDDIR>>/third_party/miniz-2.1.0 -I/<<PKGBUILDDIR>>/debian/kineto/libkineto/include -I/<<PKGBUILDDIR>>/debian/kineto/libkineto/src -I/<<PKGBUILDDIR>>/aten/src/ATen/.. -I/<<PKGBUILDDIR>>/c10/.. -I/<<PKGBUILDDIR>>/aten/src/ATen/native/quantized/cpu/qnnpack/include -I/<<PKGBUILDDIR>>/aten/src/ATen/native/quantized/cpu/qnnpack/src -I/<<PKGBUILDDIR>>/aten/src/ATen/native/quantized/cpu/qnnpack/deps/clog/include -I/<<PKGBUILDDIR>>/third_party/flatbuffers/include -isystem /<<PKGBUILDDIR>>/build/third_party/gloo -isystem /<<PKGBUILDDIR>>/cmake/../third_party/gloo -isystem /<<PKGBUILDDIR>>/cmake/../third_party/googletest/googlemock/include -isystem /<<PKGBUILDDIR>>/cmake/../third_party/googletest/googletest/include -isystem /usr/include/opencv4 -isystem /usr/include/eigen3 -isystem /<<PKGBUILDDIR>>/caffe2 -Wdate-time -D_FORTIFY_SOURCE=3 -g -O2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer -ffile-prefix-map=/<<PKGBUILDDIR>>=. -fstack-protector-strong -fstack-clash-protection -Wformat -Werror=format-security -fcf-protection -fdebug-prefix-map=/<<PKGBUILDDIR>>=/usr/src/pytorch-2.1.2+dfsg-4 -gsplit-dwarf -Wno-dangling-reference -I/usr -D_GLIBCXX_USE_CXX11_ABI=1 -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -DUSE_KINETO -DLIBKINETO_NOCUPTI -DLIBKINETO_NOROCTRACER -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DSYMBOLICATE_MOBILE_DEBUG_HANDLE -O2 -fPIC -Wall -Wextra -Werror=return-type -Werror=non-virtual-dtor -Werror=range-loop-construct -Werror=bool-operation -Wnarrowing -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-unused-parameter -Wno-unused-function -Wno-unused-result -Wno-strict-overflow -Wno-strict-aliasing -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=old-style-cast -Wno-invalid-partial-specialization -Wno-unused-private-field -Wno-aligned-allocation-unavailable -Wno-missing-braces -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Werror=cast-function-type -Wno-stringop-overflow -DHAVE_AVX512_CPU_DEFINITION -DHAVE_AVX2_CPU_DEFINITION -O2 -g -DNDEBUG -std=gnu++17 -fPIC -DCAFFE2_USE_GLOO -DTH_HAVE_THREAD -Wall -Wextra -Wno-unused-parameter -Wno-unused-function -Wno-unused-result -Wno-missing-field-initializers -Wno-unknown-pragmas -Wno-type-limits -Wno-array-bounds -Wno-strict-overflow -Wno-strict-aliasing -Wno-missing-braces -Wno-maybe-uninitialized -fvisibility=hidden -O2 -fopenmp -Wno-init-list-lifetime -MD -MT caffe2/CMakeFiles/torch_cpu.dir/__/torch/csrc/jit/tensorexpr/llvm_codegen.cpp.o -MF caffe2/CMakeFiles/torch_cpu.dir/__/torch/csrc/jit/tensorexpr/llvm_codegen.cpp.o.d -o caffe2/CMakeFiles/torch_cpu.dir/__/torch/csrc/jit/tensorexpr/llvm_codegen.cpp.o -c /<<PKGBUILDDIR>>/torch/csrc/jit/tensorexpr/llvm_codegen.cpp
FAILED: caffe2/CMakeFiles/torch_cpu.dir/__/torch/csrc/jit/tensorexpr/llvm_codegen.cpp.o
/usr/bin/c++ -DAT_PER_OPERATOR_HEADERS -DCAFFE2_BUILD_MAIN_LIB -DFMT_HEADER_ONLY=1 -DHAVE_MALLOC_USABLE_SIZE=1 -DHAVE_MMAP=1 -DHAVE_SHM_OPEN=1 -DHAVE_SHM_UNLINK=1 -DMINIZ_DISABLE_ZIP_READER_CRC32_CHECKS -DONNXIFI_ENABLE_EXT=1 -DONNX_ML=1 -DONNX_NAMESPACE=onnx -DTORCH_ENABLE_LLVM -DUSE_C10D_GLOO -DUSE_DISTRIBUTED -DUSE_EXTERNAL_MZCRC -DUSE_RPC -DUSE_TENSORPIPE -D_FILE_OFFSET_BITS=64 -Dtorch_cpu_EXPORTS -I/<<PKGBUILDDIR>>/build/aten/src -I/<<PKGBUILDDIR>>/aten/src -I/<<PKGBUILDDIR>>/build -I/<<PKGBUILDDIR>> -I/<<PKGBUILDDIR>>/cmake/../third_party/benchmark/include -I/usr/lib/llvm-18/include -I/<<PKGBUILDDIR>>/debian/foxi -I/<<PKGBUILDDIR>>/build/debian/foxi -I/<<PKGBUILDDIR>>/torch/csrc/api -I/<<PKGBUILDDIR>>/torch/csrc/api/include -I/<<PKGBUILDDIR>>/caffe2/aten/src/TH -I/<<PKGBUILDDIR>>/build/caffe2/aten/src/TH -I/<<PKGBUILDDIR>>/build/caffe2/aten/src -I/<<PKGBUILDDIR>>/build/caffe2/../aten/src -I/<<PKGBUILDDIR>>/torch/csrc -I/<<PKGBUILDDIR>>/third_party/miniz-2.1.0 -I/<<PKGBUILDDIR>>/debian/kineto/libkineto/include -I/<<PKGBUILDDIR>>/debian/kineto/libkineto/src -I/<<PKGBUILDDIR>>/aten/src/ATen/.. -I/<<PKGBUILDDIR>>/c10/.. -I/<<PKGBUILDDIR>>/aten/src/ATen/native/quantized/cpu/qnnpack/include -I/<<PKGBUILDDIR>>/aten/src/ATen/native/quantized/cpu/qnnpack/src -I/<<PKGBUILDDIR>>/aten/src/ATen/native/quantized/cpu/qnnpack/deps/clog/include -I/<<PKGBUILDDIR>>/third_party/flatbuffers/include -isystem /<<PKGBUILDDIR>>/build/third_party/gloo -isystem /<<PKGBUILDDIR>>/cmake/../third_party/gloo -isystem /<<PKGBUILDDIR>>/cmake/../third_party/googletest/googlemock/include -isystem /<<PKGBUILDDIR>>/cmake/../third_party/googletest/googletest/include -isystem /usr/include/opencv4 -isystem /usr/include/eigen3 -isystem /<<PKGBUILDDIR>>/caffe2 -Wdate-time -D_FORTIFY_SOURCE=3 -g -O2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer -ffile-prefix-map=/<<PKGBUILDDIR>>=. -fstack-protector-strong -fstack-clash-protection -Wformat -Werror=format-security -fcf-protection -fdebug-prefix-map=/<<PKGBUILDDIR>>=/usr/src/pytorch-2.1.2+dfsg-4 -gsplit-dwarf -Wno-dangling-reference -I/usr -D_GLIBCXX_USE_CXX11_ABI=1 -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -DUSE_KINETO -DLIBKINETO_NOCUPTI -DLIBKINETO_NOROCTRACER -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DSYMBOLICATE_MOBILE_DEBUG_HANDLE -O2 -fPIC -Wall -Wextra -Werror=return-type -Werror=non-virtual-dtor -Werror=range-loop-construct -Werror=bool-operation -Wnarrowing -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-unused-parameter -Wno-unused-function -Wno-unused-result -Wno-strict-overflow -Wno-strict-aliasing -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=old-style-cast -Wno-invalid-partial-specialization -Wno-unused-private-field -Wno-aligned-allocation-unavailable -Wno-missing-braces -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Werror=cast-function-type -Wno-stringop-overflow -DHAVE_AVX512_CPU_DEFINITION -DHAVE_AVX2_CPU_DEFINITION -O2 -g -DNDEBUG -std=gnu++17 -fPIC -DCAFFE2_USE_GLOO -DTH_HAVE_THREAD -Wall -Wextra -Wno-unused-parameter -Wno-unused-function -Wno-unused-result -Wno-missing-field-initializers -Wno-unknown-pragmas -Wno-type-limits -Wno-array-bounds -Wno-strict-overflow -Wno-strict-aliasing -Wno-missing-braces -Wno-maybe-uninitialized -fvisibility=hidden -O2 -fopenmp -Wno-init-list-lifetime -MD -MT caffe2/CMakeFiles/torch_cpu.dir/__/torch/csrc/jit/tensorexpr/llvm_codegen.cpp.o -MF caffe2/CMakeFiles/torch_cpu.dir/__/torch/csrc/jit/tensorexpr/llvm_codegen.cpp.o.d -o caffe2/CMakeFiles/torch_cpu.dir/__/torch/csrc/jit/tensorexpr/llvm_codegen.cpp.o -c /<<PKGBUILDDIR>>/torch/csrc/jit/tensorexpr/llvm_codegen.cpp
/<<PKGBUILDDIR>>/torch/csrc/jit/tensorexpr/llvm_codegen.cpp:37:10: fatal error: llvm/Support/Host.h: No such file or directory
37 | #include <llvm/Support/Host.h>
| ^~~~~~~~~~~~~~~~~~~~~
compilation terminated.
```

Full log: https://launchpadlibrarian.net/732495262/buildlog_ubuntu-oracular-amd64.pytorch_2.1.2+dfsg-4_BUILDING.txt.gz

llvm-dev pulls in llvm 18 and /usr/include/llvm-$version/llvm/Support/Host.h is gone in that llvm version.

Tags:

Revision history for this message

Benjamin Drung (bdrung) wrote on 2024-05-31:

Looking at the upstream git repository, I found a fix for llvm >= 18: https://github.com/pytorch/pytorch/pull/115652

Revision history for this message

Benjamin Drung (bdrung) wrote on 2024-05-31:

Download full text (11.3 KiB)

That commit is not enough. The build fails then with:

```
/usr/bin/c++ -DAT_PER_OPERATOR_HEADERS -DCAFFE2_BUILD_MAIN_LIB -DFMT_HEADER_ONLY=1 -DHAVE_MALLOC_USABLE_SIZE=1 -DHAVE_MMAP=1 -DHAVE_SHM_OPEN=1 -DHAVE_SHM_UNLINK=1 -DMINIZ_DISABLE_ZIP_READER_CRC32_CHECKS -DONNXIFI_ENABLE_EXT=1 -DONNX_ML=1 -DONNX_NAMESPACE=onnx -DTORCH_ENABLE_LLVM -DUSE_C10D_GLOO -DUSE_DISTRIBUTED -DUSE_EXTERNAL_MZCRC -DUSE_RPC -DUSE_TENSORPIPE -D_FILE_OFFSET_BITS=64 -Dtorch_cpu_EXPORTS -I/<<PKGBUILDDIR>>/build/aten/src -I/<<PKGBUILDDIR>>/aten/src -I/<<PKGBUILDDIR>>/build -I/<<PKGBUILDDIR>> -I/<<PKGBUILDDIR>>/cmake/../third_party/benchmark/include -I/usr/lib/llvm-18/include -I/<<PKGBUILDDIR>>/debian/foxi -I/<<PKGBUILDDIR>>/build/debian/foxi -I/<<PKGBUILDDIR>>/torch/csrc/api -I/<<PKGBUILDDIR>>/torch/csrc/api/include -I/<<PKGBUILDDIR>>/caffe2/aten/src/TH -I/<<PKGBUILDDIR>>/build/caffe2/aten/src/TH -I/<<PKGBUILDDIR>>/build/caffe2/aten/src -I/<<PKGBUILDDIR>>/build/caffe2/../aten/src -I/<<PKGBUILDDIR>>/torch/csrc -I/<<PKGBUILDDIR>>/third_party/miniz-2.1.0 -I/<<PKGBUILDDIR>>/debian/kineto/libkineto/include -I/<<PKGBUILDDIR>>/debian/kineto/libkineto/src -I/<<PKGBUILDDIR>>/aten/src/ATen/.. -I/<<PKGBUILDDIR>>/c10/.. -I/<<PKGBUILDDIR>>/aten/src/ATen/native/quantized/cpu/qnnpack/include -I/<<PKGBUILDDIR>>/aten/src/ATen/native/quantized/cpu/qnnpack/src -I/<<PKGBUILDDIR>>/aten/src/ATen/native/quantized/cpu/qnnpack/deps/clog/include -I/<<PKGBUILDDIR>>/third_party/flatbuffers/include -isystem /<<PKGBUILDDIR>>/build/third_party/gloo -isystem /<<PKGBUILDDIR>>/cmake/../third_party/gloo -isystem /<<PKGBUILDDIR>>/cmake/../third_party/googletest/googlemock/include -isystem /<<PKGBUILDDIR>>/cmake/../third_party/googletest/googletest/include -isystem /usr/include/opencv4 -isystem /usr/include/eigen3 -isystem /<<PKGBUILDDIR>>/caffe2 -Wdate-time -D_FORTIFY_SOURCE=3 -g -O2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer -ffile-prefix-map=/<<PKGBUILDDIR>>=. -fstack-protector-strong -fstack-clash-protection -Wformat -Werror=format-security -fcf-protection -fdebug-prefix-map=/<<PKGBUILDDIR>>=/usr/src/pytorch-2.1.2+dfsg-4ubuntu1 -gsplit-dwarf -Wno-dangling-reference  -I/usr -D_GLIBCXX_USE_CXX11_ABI=1 -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -DUSE_KINETO -DLIBKINETO_NOCUPTI -DLIBKINETO_NOROCTRACER -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DSYMBOLICATE_MOBILE_DEBUG_HANDLE -O2 -fPIC -Wall -Wextra -Werror=return-type -Werror=non-virtual-dtor -Werror=range-loop-construct -Werror=bool-operation -Wnarrowing -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-unused-parameter -Wno-unused-function -Wno-unused-result -Wno-strict-overflow -Wno-strict-aliasing -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=old-style-cast -Wno-invalid-partial-specialization -Wno-unused-private-field -Wno-aligned-allocation-unavailable -Wno-missing-braces -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Werror=cast-function-type -Wno-stringop-overflow -DHAVE_AVX512_CPU_DEFINITION -DHAVE_AVX2_CPU_DEFINITION -O2 -g -DNDEBUG -std=gnu++17 -fPIC -DCAFFE2_USE_GLOO -DTH_HAVE_THREAD -Wall -Wextra -Wno-unused-parameter -Wno-unused-function -Wno-unused-result -Wno-missing-field-initializers -Wno-unknown-pragmas -Wno-type-limits -Wno-array-bounds -Wno-strict-overflow -Wno-strict-aliasing -Wno-missing-braces -Wno-maybe-uninitialized -fvisibility=hidden -O2 -fopenmp -Wno-redundant-move -Wno-noexcept-type -MD -MT caffe2/CMakeFiles/torch_cpu.dir/__/torch/csrc/jit/tensorexpr/llvm_jit.cpp.o -MF caffe2/CMakeFiles/torch_cpu.dir/__/torch/csrc/jit/tensorexpr/llvm_jit.cpp.o.d -o caffe2/CMakeFiles/torch_cpu.dir/__/torch/csrc/jit/tensorexpr/llvm_jit.cpp.o -c /<<PKGBUILDDIR>>/torch/csrc/jit/tensorexpr/llvm_jit.cpp
/<<PKGBUILDDIR>>/torch/csrc/jit/tensorexpr/llvm_jit.cpp: In function ‘llvm::orc::JITTargetMachineBuilder makeTargetMachineBuilder(c10::optional<std::__cxx11::basic_string<char> >, c10::optional<std::__cxx11::basic_string<char> >, c10::optional<std::__cxx11::basic_string<char> >)’:
/<<PKGBUILDDIR>>/torch/csrc/jit/tensorexpr/llvm_jit.cpp:108:45: error: ‘Default’ is not a member of ‘llvm::CodeGenOpt’
  108 |   JTMB.setCodeGenOptLevel(llvm::CodeGenOpt::Default);
      |                                             ^~~~~~~
/<<PKGBUILDDIR>>/torch/csrc/jit/tensorexpr/llvm_jit.cpp:108:45: note: suggested alternatives:
In file included from /<<PKGBUILDDIR>>/aten/src/ATen/core/ivalue_inl.h:12,
                 from /<<PKGBUILDDIR>>/aten/src/ATen/core/ivalue.h:1499,
                 from /<<PKGBUILDDIR>>/aten/src/ATen/core/List_inl.h:4,
                 from /<<PKGBUILDDIR>>/aten/src/ATen/core/List.h:490,
                 from /<<PKGBUILDDIR>>/aten/src/ATen/core/IListRef_inl.h:3,
                 from /<<PKGBUILDDIR>>/aten/src/ATen/core/IListRef.h:632,
                 from /<<PKGBUILDDIR>>/aten/src/ATen/DeviceGuard.h:3,
                 from /<<PKGBUILDDIR>>/build/aten/src/ATen/Functions.h:60,
                 from /<<PKGBUILDDIR>>/torch/csrc/jit/tensorexpr/external_functions.h:4,
                 from /<<PKGBUILDDIR>>/torch/csrc/jit/tensorexpr/llvm_jit.cpp:5:
/<<PKGBUILDDIR>>/aten/src/ATen/core/jit_type.h:1704:3: note:   ‘c10::TypeVerbosity::Default’
 1704 |   Default = Full,
      |   ^~~~~~~
In file included from /<<PKGBUILDDIR>>/aten/src/ATen/Context.h:4,
                 from /<<PKGBUILDDIR>>/build/aten/src/ATen/Functions.h:59:
/<<PKGBUILDDIR>>/aten/src/ATen/LinalgBackend.h:10:37: note:   ‘at::LinalgBackend::Default’
   10 | enum class LinalgBackend : int8_t { Default, Cusolver, Magma };
      |                                     ^~~~~~~
In file included from /usr/lib/llvm-18/include/llvm/Support/Mutex.h:16,
                 from /usr/lib/llvm-18/include/llvm/ExecutionEngine/ExecutionEngine.h:29,
                 from /<<PKGBUILDDIR>>/torch/csrc/jit/tensorexpr/llvm_jit.cpp:10:
/usr/lib/llvm-18/include/llvm/Support/Threading.h:253:5: note:   ‘llvm::ThreadPriority::Default’
  253 |     Default = 2,
      |     ^~~~~~~
In file included from /usr/lib/llvm-18/include/llvm/Target/TargetMachine.h:23,
                 from /<<PKGBUILDDIR>>/torch/csrc/jit/tensorexpr/llvm_jit.h:14,
                 from /<<PKGBUILDDIR>>/torch/csrc/jit/tensorexpr/llvm_jit.cpp:7:
/usr/lib/llvm-18/include/llvm/Target/CGPassBuilderOption.h:23:27: note:   ‘llvm::RegAllocType::Default’
   23 | enum class RegAllocType { Default, Basic, Fast, Greedy, PBQP };
      |                           ^~~~~~~
In file included from /usr/lib/llvm-18/include/llvm/Target/CGPassBuilderOption.h:17:
/usr/lib/llvm-18/include/llvm/Target/TargetOptions.h:98:5: note:   ‘llvm::DebuggerKind::Default’
   98 |     Default, ///< No specific tuning requested.
      |     ^~~~~~~
/usr/lib/llvm-18/include/llvm/Target/TargetOptions.h:78:5: note:   ‘llvm::EABI::Default’
   78 |     Default, // Default means not specified
      |     ^~~~~~~
In file included from /usr/lib/llvm-18/include/llvm/Target/TargetOptions.h:18:
/usr/lib/llvm-18/include/llvm/MC/MCTargetOptions.h:34:3: note:   ‘llvm::EmitDwarfUnwindType::Default’
   34 |   Default,         // Default behavior is based on the target
      |   ^~~~~~~
In file included from /usr/lib/llvm-18/include/llvm/IR/Attributes.h:24,
                 from /usr/lib/llvm-18/include/llvm/IR/Module.h:22,
                 from /usr/lib/llvm-18/include/llvm/ExecutionEngine/Orc/ThreadSafeModule.h:17,
                 from /<<PKGBUILDDIR>>/torch/csrc/jit/tensorexpr/llvm_jit.h:13:
/usr/lib/llvm-18/include/llvm/Support/CodeGen.h:124:5: note:   ‘llvm::UWTableKind::Default’
  124 |     Default = 2,
      |     ^~~~~~~
/usr/lib/llvm-18/include/llvm/Support/CodeGen.h:57:5: note:   ‘llvm::CodeGenOptLevel::Default’
   57 |     Default = 2,   ///< -O2, -Os
      |     ^~~~~~~
In file included from /usr/lib/llvm-18/include/llvm/ExecutionEngine/Orc/ObjectLinkingLayer.h:19,
                 from /usr/lib/llvm-18/include/llvm/ExecutionEngine/Orc/ExecutionUtils.h:20,
                 from /<<PKGBUILDDIR>>/torch/csrc/jit/tensorexpr/llvm_jit.cpp:15:
/usr/lib/llvm-18/include/llvm/ExecutionEngine/JITLink/JITLink.h:392:3: note:   ‘llvm::jitlink::Scope::Default’
  392 |   Default,
      |   ^~~~~~~
/usr/lib/llvm-18/include/llvm/Support/CodeGen.h:40:18: note:   ‘llvm::PIELevel::Default’
   40 |     enum Level { Default=0, Small=1, Large=2 };
      |                  ^~~~~~~
/usr/lib/llvm-18/include/llvm/Target/TargetOptions.h:29:7: note:   ‘llvm::FloatABI::Default’
   29 |       Default, // Target-specific (either soft or hard depending on triple, etc).
      |       ^~~~~~~
```

and

```
/<<PKGBUILDDIR>>/torch/csrc/jit/tensorexpr/llvm_codegen.cpp: In member function ‘void torch::jit::tensorexpr::LLVMCodeGenImpl::emitKernel(torch::jit::tensorexpr::StmtPtr, const std::vector<llvm::Type*, std::allocator<llvm::Type*> >&)’:
/<<PKGBUILDDIR>>/torch/csrc/jit/tensorexpr/llvm_codegen.cpp:750:30: error: ‘CGFT_AssemblyFile’ is not a member of ‘llvm::CodeGenFileType’; did you mean ‘AssemblyFile’?
  750 |       llvm::CodeGenFileType::CGFT_AssemblyFile);
      |                              ^~~~~~~~~~~~~~~~~
      |                              AssemblyFile
/<<PKGBUILDDIR>>/torch/csrc/jit/tensorexpr/llvm_codegen.cpp: In member function ‘torch::jit::tensorexpr::{anonymous}::TypedPointer torch::jit::tensorexpr::LLVMCodeGenImpl::packFuncArgs(const std::vector<llvm::Value*, std::allocator<llvm::Value*> >&)’:
/<<PKGBUILDDIR>>/torch/csrc/jit/tensorexpr/llvm_codegen.cpp:1485:50: error: ‘getInt8PtrTy’ is not a member of ‘llvm::Type’
 1485 |     llvm::PointerType* VoidPtrType = llvm::Type::getInt8PtrTy(getContext());
      |                                                  ^~~~~~~~~~~~
/<<PKGBUILDDIR>>/torch/csrc/jit/tensorexpr/llvm_codegen.cpp: In member function ‘virtual void torch::jit::tensorexpr::LLVMCodeGenImpl::visit(torch::jit::tensorexpr::AllocatePtr)’:
/<<PKGBUILDDIR>>/torch/csrc/jit/tensorexpr/llvm_codegen.cpp:2580:42: error: ‘CreateMalloc’ is not a member of ‘llvm::CallInst’
 2580 |   llvm::Instruction* I = llvm::CallInst::CreateMalloc(
      |                                          ^~~~~~~~~~~~
/<<PKGBUILDDIR>>/torch/csrc/jit/tensorexpr/llvm_codegen.cpp: In member function ‘virtual void torch::jit::tensorexpr::LLVMCodeGenImpl::visit(torch::jit::tensorexpr::FreePtr)’:
/<<PKGBUILDDIR>>/torch/csrc/jit/tensorexpr/llvm_codegen.cpp:2614:33: error: ‘CreateFree’ is not a member of ‘llvm::CallInst’
 2614 |     irb_.Insert(llvm::CallInst::CreateFree(ptr, irb_.GetInsertBlock()));
      |                                 ^~~~~~~~~~
/<<PKGBUILDDIR>>/torch/csrc/jit/tensorexpr/llvm_codegen.cpp: In member function ‘void torch::jit::tensorexpr::LLVMCodeGenImpl::optimize(llvm::Module&)’:
/<<PKGBUILDDIR>>/torch/csrc/jit/tensorexpr/llvm_codegen.cpp:2749:34: error: no matching function for call to ‘llvm::TargetMachine::registerPassBuilderCallbacks(llvm::PassBuilder&)’
 2749 |   TM.registerPassBuilderCallbacks(PB);
      |   ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~
In file included from /<<PKGBUILDDIR>>/torch/csrc/jit/tensorexpr/llvm_jit.h:14:
/usr/lib/llvm-18/include/llvm/Target/TargetMachine.h:369:16: note: candidate: ‘virtual void llvm::TargetMachine::registerPassBuilderCallbacks(llvm::PassBuilder&, bool)’
  369 |   virtual void registerPassBuilderCallbacks(PassBuilder &,
      |                ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
/usr/lib/llvm-18/include/llvm/Target/TargetMachine.h:369:16: note:   candidate expects 2 arguments, 1 provided
```

Revision history for this message

Athos Ribeiro (athos-ribeiro) wrote on 2024-06-05:

There are 2 other patches upstream related to LLVM-18. The whole set so far is:

https://github.com/pytorch/pytorch/commit/20f394f10a389bcf13485929be8862f98ad4b322
https://github.com/pytorch/pytorch/commit/23bff71de4a6438e968a684f6fe8259180a6ab03
https://github.com/pytorch/pytorch/commit/f0d71de4aca4eb97180def7db4e5f8fbbc302772

The builds are still failing with those (with different errors though). Here is a PPA with the patches above applied:

https://launchpad.net/~athos-ribeiro/+archive/ubuntu/pytorch-llvm18/+packages

Revision history for this message

Sergio Durigan Junior (sergiodj) wrote on 2024-06-10:

I may have found the right set of patches to backport. I'm doing a PPA build now to confirm, and will upload if everything goes OK.

Changed in pytorch (Ubuntu):
assignee:	nobody → Sergio Durigan Junior (sergiodj)
status:	New → Triaged

Revision history for this message

Sergio Durigan Junior (sergiodj) wrote on 2024-06-11:

Package uploaded.

Changed in pytorch (Ubuntu):
status:	Triaged → In Progress

Revision history for this message

Launchpad Janitor (janitor) wrote on 2024-06-11:

This bug was fixed in the package pytorch - 2.1.2+dfsg-4ubuntu1

---------------
pytorch (2.1.2+dfsg-4ubuntu1) oracular; urgency=medium

  * Fix FTBFS with LLVM 18. (LP: #2067720)
    - d/p/0011-Update-caffe2-with-LLVM-18-API-change-109408.patch:
    - d/p/0012-Fix-build-for-llvm-18-115652.patch:
    - d/p/0013-Update-for-an-API-change-in-LLVM-18.-117086.patch:
    - d/p/0014-Replaced-getInt8PtrTy-with-getUnqual-113455.patch:
    - d/p/0015-Update-apis-incompatible-with-llvm-versions-in-codeg.patch:
    - d/p/0016-Revert-upstream-commit-32643bc9266271ebf3b935573189e.patch:
      Backport patches from upstream to fix the issue.

-- Sergio Durigan Junior <email address hidden> Mon, 10 Jun 2024 14:38:33 -0400

Changed in pytorch (Ubuntu):
status:	In Progress → Fix Released

Report a bug

This report contains Public information

Everyone can see this information.

You are

Subscribing...

Edit bug mail

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.

Ubuntupytorch package

pytorch 2.1.2+dfsg-4 FTBFS with llvm 18

Bug Description

Other bug subscribers

Remote bug watches

Ubuntu
pytorch package