[Flang] Compilation of a large array of a derived type with an initialized member needs a lot of memory #77877
Description
This is an issue from Fujitsu testsuite.
Flang-new terminates abnormally when compiling the following program.
When I checked with the top command, %MEM gradually increased to 100%.
It seems that a lot of memory was used.
There is no problem with gfortran compiling on the same machine.
The following are the test program, Flang-new and gfortran compilation result, and the stack trace of Flang-new.
! test.f90
type t1
complex :: c = (0.0, 0.0)
end type t1
type(t1) val(1000000)
print *, val(500000)
end$ flang-new -v test.f90
flang-new version 18.0.0git (https://github.com/llvm/llvm-project.git 2af0695b875911144ed1f7b4e5679be500b86a77)
Target: aarch64-unknown-linux-gnu
Thread model: posix
InstalledDir: /path/to/install/bin
Found candidate GCC installation: /opt/gcc/11.2.0/lib/gcc/aarch64-unknown-linux-gnu/11.2.0
Selected GCC installation: /opt/gcc/11.2.0/lib/gcc/aarch64-unknown-linux-gnu/11.2.0
Candidate multilib: .;@m64
Selected multilib: .;@m64
"/path/to/install/bin/flang-new" -fc1 -triple aarch64-unknown-linux-gnu -emit-obj -fcolor-diagnostics -mrelocation-model pic -pic-level 2 -pic-is-pie -target-cpu generic -target-feature +neon -target-feature +v8a -mframe-pointer=non-leaf -o /tmp/test-f371a2.o -x f95-cpp-input test.f90
flang-new: error: unable to execute command: Killed
flang-new: error: flang frontend command failed due to signal (use -v to see invocation)
flang-new version 18.0.0git (https://github.com/llvm/llvm-project.git 2af0695b875911144ed1f7b4e5679be500b86a77)
Target: aarch64-unknown-linux-gnu
Thread model: posix
InstalledDir: /path/to/install/bin
flang-new: note: diagnostic msg:
********************
PLEASE ATTACH THE FOLLOWING FILES TO THE BUG REPORT:
Preprocessed source(s) and associated run script(s) are located at:
flang-new: note: diagnostic msg: /tmp/test-e5be39
flang-new: note: diagnostic msg: /tmp/test-e5be39.sh
flang-new: note: diagnostic msg:
********************$ gfortran -v test.f90
Driving: gfortran -v test.f90 -l gfortran -l m -shared-libgcc
Using built-in specs.
COLLECT_GCC=gfortran
COLLECT_LTO_WRAPPER=/usr/libexec/gcc/aarch64-redhat-linux/8/lto-wrapper
Target: aarch64-redhat-linux
Configured with: ../configure --enable-bootstrap --enable-languages=c,c++,fortran,lto --prefix=/usr --mandir=/usr/share/man --infodir=/usr/share/info --with-bugurl=http://bugzilla.redhat.com/bugzilla --enable-shared --enable-threads=posix --enable-checking=release --enable-multilib --with-system-zlib --enable-__cxa_atexit --disable-libunwind-exceptions --enable-gnu-unique-object --enable-linker-build-id --with-gcc-major-version-only --with-linker-hash-style=gnu --enable-plugin --enable-initfini-array --with-isl --disable-libmpx --enable-gnu-indirect-function --build=aarch64-redhat-linux
Thread model: posix
gcc version 8.3.1 20190507 (Red Hat 8.3.1-4) (GCC)
:
$#0 0x00007ffff090ff7b in llvm::Constant::getAggregateElement (this=0x7ffc5236d810, Elt=117955)
at /path/to/llvm-project/llvm/lib/IR/Constants.cpp:423
#1 0x00007ffff08f021c in llvm::ConstantFoldInsertValueInstruction (Agg=0x7ffc5236d810, Val=0x555555967600, Idxs=...)
at /path/to/llvm-project/llvm/lib/IR/ConstantFold.cpp:605
#2 0x00007ffff0ad9596 in llvm::ConstantFolder::FoldInsertValue (this=0x7fffffffb9b8, Agg=0x7ffc5236d810, Val=0x555555967600, IdxList=...)
at /path/to/llvm-project/llvm/include/llvm/IR/ConstantFolder.h:145
#3 0x00007fffdfa3d743 in llvm::IRBuilderBase::CreateInsertValue (this=0x7fffffffb940, Agg=0x7ffc5236d810, Val=0x555555967600, Idxs=..., Name=...)
at /path/to/llvm-project/llvm/include/llvm/IR/IRBuilder.h:2498
#4 0x00007fffdfa177e4 in convertOperationImpl (opInst=..., builder=..., moduleTranslation=...)
at /path/to/build/tools/mlir/include/mlir/Dialect/LLVMIR/LLVMConversions.inc:211
#5 0x00007fffdfa2bccb in (anonymous namespace)::LLVMDialectLLVMIRTranslationInterface::convertOperation (this=0x5555555ed730, op=0x5555558fbb90,
builder=..., moduleTranslation=...) at /path/to/llvm-project/mlir/lib/Target/LLVMIR/Dialect/LLVMIR/LLVMToLLVMIRTranslation.cpp:415
#6 0x00007fffe487ab27 in mlir::LLVM::ModuleTranslation::convertOperation (this=0x7fffffffbc00, op=..., builder=..., recordInsertions=false)
at /path/to/llvm-project/mlir/lib/Target/LLVMIR/ModuleTranslation.cpp:753
#7 0x00007fffe487ba28 in mlir::LLVM::ModuleTranslation::convertGlobals (this=0x7fffffffbc00)
at /path/to/llvm-project/mlir/lib/Target/LLVMIR/ModuleTranslation.cpp:925
#8 0x00007fffe4880201 in mlir::translateModuleToLLVMIR (module=0x5555556752c0, llvmContext=..., name=...)
at /path/to/llvm-project/mlir/lib/Target/LLVMIR/ModuleTranslation.cpp:1566
#9 0x00007ffff6c17c69 in Fortran::frontend::CodeGenAction::generateLLVMIR (this=0x5555555998b0)
at /path/to/llvm-project/flang/lib/Frontend/FrontendActions.cpp:777
#10 0x00007ffff6c192f7 in Fortran::frontend::CodeGenAction::executeAction (this=0x5555555998b0)
at /path/to/llvm-project/flang/lib/Frontend/FrontendActions.cpp:1148
#11 0x00007ffff6c10650 in Fortran::frontend::FrontendAction::execute (this=0x5555555998b0)
at /path/to/llvm-project/flang/lib/Frontend/FrontendAction.cpp:112
#12 0x00007ffff6be0c11 in Fortran::frontend::CompilerInstance::executeAction (this=0x555555594970, act=...)
at /path/to/llvm-project/flang/lib/Frontend/CompilerInstance.cpp:173
#13 0x00007ffff71952b7 in Fortran::frontend::executeCompilerInvocation (flang=0x555555594970)
at /path/to/llvm-project/flang/lib/FrontendTool/ExecuteCompilerInvocation.cpp:221
#14 0x000055555555e2c5 in fc1_main (argv=..., argv0=0x7fffffffe4ca "/path/to/build/bin/flang-new")
at /path/to/llvm-project/flang/tools/flang-driver/fc1_main.cpp:67
#15 0x0000555555557509 in executeFC1Tool (argV=...) at /path/to/llvm-project/flang/tools/flang-driver/driver.cpp:66
#16 0x00005555555579de in main (argc=19, argv=0x7fffffffe1d8) at /path/to/llvm-project/flang/tools/flang-driver/driver.cpp:109
The following is a similar program but an array size is smaller.
! small_test.f90
type t1
complex :: c = (0.0, 0.0)
end type t1
type(t1) val(100)
print *, val(50)
endThat can be compiled normally, but Flang generates large MLIR with LLVM dialect.
I suspect this is the cause of the abnormal termination.
$ flang-new -Xflang -save-temps small_test.f90
$ cat small_test-llvmir.mlir
module attributes {dlti.dl_spec = #dlti.dl_spec<#dlti.dl_entry<i1, dense<8> : vector<2xi64>>, #dlti.dl_entry<!llvm.ptr, dense<64> : vector<4xi64>>, #dlti.dl_entry<f16, dense<16> : vector<2xi64>>, #dlti.dl_entry<i32, dense<32> : vector<2xi64>>, #dlti.dl_entry<i16, dense<[16, 32]> : vector<2xi64>>, #dlti.dl_entry<i8, dense<[8, 32]> : vector<2xi64>>, #dlti.dl_entry<f64, dense<64> : vector<2xi64>>, #dlti.dl_entry<i128, dense<128> : vector<2xi64>>, #dlti.dl_entry<f128, dense<128> : vector<2xi64>>, #dlti.dl_entry<i64, dense<64> : vector<2xi64>>, #dlti.dl_entry<"dlti.stack_alignment", 128 : i64>, #dlti.dl_entry<"dlti.endianness", "little">>, fir.defaultkind = "a1c4d8i4l4r4", fir.kindmap = "", llvm.data_layout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128", llvm.target_triple = "aarch64-unknown-linux-gnu"} {
llvm.func @_QQmain() attributes {frame_pointer = #llvm.framePointerKind<"non-leaf">} {
%0 = llvm.mlir.constant(1 : i32) : i32
:
llvm.return
}
:
llvm.mlir.global internal @_QFEval() {addr_space = 0 : i32} : !llvm.array<100 x struct<"_QFTt1", (struct<(f32, f32)>)>> {
%0 = llvm.mlir.constant(0.000000e+00 : f32) : f32
%1 = llvm.mlir.undef : !llvm.struct<"_QFTt1", (struct<(f32, f32)>)>
%2 = llvm.mlir.undef : !llvm.struct<(f32, f32)>
%3 = llvm.insertvalue %0, %2[0] : !llvm.struct<(f32, f32)>
%4 = llvm.insertvalue %0, %3[1] : !llvm.struct<(f32, f32)>
%5 = llvm.insertvalue %4, %1[0] : !llvm.struct<"_QFTt1", (struct<(f32, f32)>)>
%6 = llvm.mlir.undef : !llvm.array<100 x struct<"_QFTt1", (struct<(f32, f32)>)>>
%7 = llvm.insertvalue %5, %6[0] : !llvm.array<100 x struct<"_QFTt1", (struct<(f32, f32)>)>>
%8 = llvm.insertvalue %5, %7[1] : !llvm.array<100 x struct<"_QFTt1", (struct<(f32, f32)>)>>
%9 = llvm.insertvalue %5, %8[2] : !llvm.array<100 x struct<"_QFTt1", (struct<(f32, f32)>)>>
%10 = llvm.insertvalue %5, %9[3] : !llvm.array<100 x struct<"_QFTt1", (struct<(f32, f32)>)>>
:
%106 = llvm.insertvalue %5, %105[99] : !llvm.array<100 x struct<"_QFTt1", (struct<(f32, f32)>)>>
llvm.return %106 : !llvm.array<100 x struct<"_QFTt1", (struct<(f32, f32)>)>>
}
llvm.mlir.global linkonce_odr constant @_QFE.c.t1() comdat(@__llvm_comdat::@_QFE.c.t1) {addr_space = 0 : i32} : !llvm.array<1 x struct<"_QM__fortran_type_infoTcomponent", (struct<(ptr, i64, i32, i8, i8, i8, i8)>, i8, i8, i8, i8, array<4 x i8>, i64, struct<"_QM__fortran_type_infoTvalue", (i8, array<7 x i8>, i64)>, struct<(ptr, i64, i32, i8, i8, i8, i8, ptr, array<1 x i64>)>, struct<(ptr, i64, i32, i8, i8, i8, i8, array<1 x array<3 x i64>>, ptr, array<1 x i64>)>, struct<(ptr, i64, i32, i8, i8, i8, i8, array<2 x array<3 x i64>>, ptr, array<1 x i64>)>, struct<"_QM__fortran_builtinsT__builtin_c_ptr", (i64)>)>> {
%0 = llvm.mlir.constant(0 : index) : i64
%1 = llvm.mlir.constant(0 : i64) : i64
%2 = llvm.mlir.constant(0 : i8) : i8
%3 = llvm.mlir.constant(4 : i8) : i8
%4 = llvm.mlir.constant(2 : i8) : i8
%5 = llvm.mlir.constant(1 : i8) : i8
%6 = llvm.mlir.undef : !llvm.array<1 x struct<"_QM__fortran_type_infoTcomponent", (struct<(ptr, i64, i32, i8, i8, i8, i8)>, i8, i8, i8, i8, array<4 x i8>, i64, struct<"_QM__fortran_type_infoTvalue", (i8, array<7 x i8>, i64)>, struct<(ptr, i64, i32, i8, i8, i8, i8, ptr, array<1 x i64>)>, struct<(ptr, i64, i32, i8, i8, i8, i8, array<1 x array<3 x i64>>, ptr, array<1 x i64>)>, struct<(ptr, i64, i32, i8, i8, i8, i8, array<2 x array<3 x i64>>, ptr, array<1 x i64>)>, struct<"_QM__fortran_builtinsT__builtin_c_ptr", (i64)>)>>
%7 = llvm.mlir.undef : !llvm.struct<"_QM__fortran_type_infoTcomponent", (struct<(ptr, i64, i32, i8, i8, i8, i8)>, i8, i8, i8, i8, array<4 x i8>, i64, struct<"_QM__fortran_type_infoTvalue", (i8, array<7 x i8>, i64)>, struct<(ptr, i64, i32, i8, i8, i8, i8, ptr, array<1 x i64>)>, struct<(ptr, i64, i32, i8, i8, i8, i8, array<1 x array<3 x i64>>, ptr, array<1 x i64>)>, struct<(ptr, i64, i32, i8, i8, i8, i8, array<2 x array<3 x i64>>, ptr, array<1 x i64>)>, struct<"_QM__fortran_builtinsT__builtin_c_ptr", (i64)>)>
%8 = llvm.mlir.addressof @_QFE.n.c : !llvm.ptr
%9 = llvm.mlir.constant(40 : i32) : i32
%10 = llvm.mlir.zero : !llvm.ptr
%11 = llvm.getelementptr %10[1] : (!llvm.ptr) -> !llvm.ptr, !llvm.array<1 x i8>
%12 = llvm.ptrtoint %11 : !llvm.ptr to i64
%13 = llvm.mlir.undef : !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)>
%14 = llvm.insertvalue %12, %13[1] : !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)>
%15 = llvm.mlir.constant(20180515 : i32) : i32
%16 = llvm.insertvalue %15, %14[2] : !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)>
%17 = llvm.mlir.constant(0 : i32) : i32
%18 = llvm.trunc %17 : i32 to i8
%19 = llvm.insertvalue %18, %16[3] : !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)>
%20 = llvm.trunc %9 : i32 to i8
%21 = llvm.insertvalue %20, %19[4] : !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)>
%22 = llvm.mlir.constant(0 : i32) : i32
%23 = llvm.trunc %22 : i32 to i8
%24 = llvm.insertvalue %23, %21[5] : !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)>
%25 = llvm.mlir.constant(0 : i32) : i32
%26 = llvm.trunc %25 : i32 to i8
%27 = llvm.insertvalue %26, %24[6] : !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)>
%28 = llvm.insertvalue %8, %27[0] : !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)>
%29 = llvm.mlir.constant(1 : i64) : i64
%30 = llvm.mlir.constant(40 : i32) : i32
%31 = llvm.mlir.zero : !llvm.ptr
%32 = llvm.getelementptr %31[1] : (!llvm.ptr) -> !llvm.ptr, i8
%33 = llvm.ptrtoint %32 : !llvm.ptr to i64
%34 = llvm.mul %33, %29 : i64
%35 = llvm.mlir.undef : !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)>
%36 = llvm.insertvalue %34, %35[1] : !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)>
%37 = llvm.mlir.constant(20180515 : i32) : i32
%38 = llvm.insertvalue %37, %36[2] : !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)>
%39 = llvm.mlir.constant(0 : i32) : i32
%40 = llvm.trunc %39 : i32 to i8
%41 = llvm.insertvalue %40, %38[3] : !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)>
%42 = llvm.trunc %30 : i32 to i8
%43 = llvm.insertvalue %42, %41[4] : !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)>
%44 = llvm.mlir.constant(1 : i32) : i32
%45 = llvm.trunc %44 : i32 to i8
%46 = llvm.insertvalue %45, %43[5] : !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)>
%47 = llvm.mlir.constant(0 : i32) : i32
%48 = llvm.trunc %47 : i32 to i8
%49 = llvm.insertvalue %48, %46[6] : !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)>
%50 = llvm.extractvalue %28[0] : !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)>
%51 = llvm.mlir.constant(0 : i64) : i64
%52 = llvm.mlir.constant(1 : i64) : i64
%53 = llvm.insertvalue %50, %49[0] : !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)>
%54 = llvm.insertvalue %53, %7[0] : !llvm.struct<"_QM__fortran_type_infoTcomponent", (struct<(ptr, i64, i32, i8, i8, i8, i8)>, i8, i8, i8, i8, array<4 x i8>, i64, struct<"_QM__fortran_type_infoTvalue", (i8, array<7 x i8>, i64)>, struct<(ptr, i64, i32, i8, i8, i8, i8, ptr, array<1 x i64>)>, struct<(ptr, i64, i32, i8, i8, i8, i8, array<1 x array<3 x i64>>, ptr, array<1 x i64>)>, struct<(ptr, i64, i32, i8, i8, i8, i8, array<2 x array<3 x i64>>, ptr, array<1 x i64>)>, struct<"_QM__fortran_builtinsT__builtin_c_ptr", (i64)>)>
%55 = llvm.insertvalue %5, %54[1] : !llvm.struct<"_QM__fortran_type_infoTcomponent", (struct<(ptr, i64, i32, i8, i8, i8, i8)>, i8, i8, i8, i8, array<4 x i8>, i64, struct<"_QM__fortran_type_infoTvalue", (i8, array<7 x i8>, i64)>, struct<(ptr, i64, i32, i8, i8, i8, i8, ptr, array<1 x i64>)>, struct<(ptr, i64, i32, i8, i8, i8, i8, array<1 x array<3 x i64>>, ptr, array<1 x i64>)>, struct<(ptr, i64, i32, i8, i8, i8, i8, array<2 x array<3 x i64>>, ptr, array<1 x i64>)>, struct<"_QM__fortran_builtinsT__builtin_c_ptr", (i64)>)>
: