[Flang][OpenMP] Compilation error of using a reduction clause (min) with a variable (kind=16) #87955

ohno-fj · 2024-04-08T01:49:26Z

Version of flang-new : 19.0.0(cbcdf126ccc774c063b5d5140c1393ff5305dded)/AArch64

A compilation error occurs when using a reduction clause (min) with a variable (kind=16).
The same error occurs when max intrinsic function is used.

Compilation terminates normally in the following cases:

change kind=16 to kind=8
remove reduction clause (min)
compile without -fopenmp option

The following are the test program, Flang-new, Gfortran and ifort compilation/execution result.

snfmm235_3.f90:

program main
  real(kind=16) r3/1/
!$omp parallel reduction(min:r3)
  r3=min(r3,1.0_16)
  write(6,*) "r3 = ", r3
  if (r3 < 1.0_16) then
     r3 = 1.0_16
  end if
!$omp end parallel
  write(6,*) "pass"
end program main

$ flang-new -fopenmp snfmm235_3.f90
LLVM ERROR: Cannot select: t7: f128 = fminimum contract t4, t6
  t4: f128,ch = load<(dereferenceable load (s128) from @_QFEr3)> t0, t41, undef:i64
    t41: i64 = AArch64ISD::ADDlow t40, TargetGlobalAddress:i64<ptr @_QFEr3> 0 [TF=34]
      t40: i64 = AArch64ISD::ADRP TargetGlobalAddress:i64<ptr @_QFEr3> 0 [TF=1]
        t38: i64 = TargetGlobalAddress<ptr @_QFEr3> 0 [TF=1]
      t39: i64 = TargetGlobalAddress<ptr @_QFEr3> 0 [TF=34]
    t3: i64 = undef
  t6: f128,ch = load<(dereferenceable load (s128) from %ir.1)> t0, FrameIndex:i64<1>, undef:i64
    t5: i64 = FrameIndex<1>
    t3: i64 = undef
In function: _QQmain..omp_par
PLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.
Stack dump:
0.      Program arguments: /lustre/home/ohno/LLVM_20240321/release/bin/flang-new -fc1 -triple aarch64-unknown-linux-gnu -emit-obj -fopenmp -mrelocation-model pic -pic-level 2 -pic-is-pie -target-cpu generic -target-feature +outline-atomics -target-feature +v8a -target-feature +fp-armv8 -target-feature +neon -mframe-pointer=non-leaf -o /tmp/snfmm235_3-4e13f7.o -x f95-cpp-input snfmm235_3.f90
1.      Running pass 'Function Pass Manager' on module 'FIRModule'.
2.      Running pass 'AArch64 Instruction Selection' on function '@_QQmain..omp_par'
 #0 0x00000000041f9f00 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x41f9f00)
 #1 0x00000000041f7e08 llvm::sys::RunSignalHandlers() (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x41f7e08)
 #2 0x00000000041f7f8c SignalHandler(int) Signals.cpp:0:0
 #3 0x000040001c8a07a0 (linux-vdso.so.1+0x7a0)
 #4 0x000040001cd46274 raise (/lib64/libc.so.6+0x36274)
 #5 0x000040001cd30a2c abort (/lib64/libc.so.6+0x20a2c)
 #6 0x0000000004182d58 llvm::report_fatal_error(llvm::StringRef, bool) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x4182d58)
 #7 0x0000000004676b10 llvm::SelectionDAGISel::CannotYetSelect(llvm::SDNode*) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x4676b10)
 #8 0x000000000467a91c llvm::SelectionDAGISel::SelectCodeCommon(llvm::SDNode*, unsigned char const*, unsigned int) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x467a91c)
 #9 0x0000000003f9be2c (anonymous namespace)::AArch64DAGToDAGISel::Select(llvm::SDNode*) AArch64ISelDAGToDAG.cpp:0:0
#10 0x0000000004674eac llvm::SelectionDAGISel::DoInstructionSelection() (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x4674eac)
#11 0x0000000004681830 llvm::SelectionDAGISel::CodeGenAndEmitDAG() (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x4681830)
#12 0x000000000468452c llvm::SelectionDAGISel::SelectAllBasicBlocks(llvm::Function const&) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x468452c)
#13 0x00000000046859b0 llvm::SelectionDAGISel::runOnMachineFunction(llvm::MachineFunction&) (.part.1268) SelectionDAGISel.cpp:0:0
#14 0x00000000068cbda0 llvm::MachineFunctionPass::runOnFunction(llvm::Function&) (.part.78) MachineFunctionPass.cpp:0:0
#15 0x0000000007f91e90 llvm::FPPassManager::runOnFunction(llvm::Function&) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x7f91e90)
#16 0x0000000007f920c8 llvm::FPPassManager::runOnModule(llvm::Module&) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x7f920c8)
#17 0x0000000007f92bb4 llvm::legacy::PassManagerImpl::run(llvm::Module&) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x7f92bb4)
#18 0x000000000485663c Fortran::frontend::CodeGenAction::executeAction() (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x485663c)
#19 0x000000000422f164 Fortran::frontend::FrontendAction::execute() (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x422f164)
#20 0x00000000042238dc Fortran::frontend::CompilerInstance::executeAction(Fortran::frontend::FrontendAction&) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x42238dc)
#21 0x0000000004233de0 Fortran::frontend::executeCompilerInvocation(Fortran::frontend::CompilerInstance*) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x4233de0)
#22 0x0000000003e45678 fc1_main(llvm::ArrayRef<char const*>, char const*) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x3e45678)
#23 0x0000000003dccd10 main (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x3dccd10)
#24 0x000040001cd34384 __libc_start_main (/lib64/libc.so.6+0x24384)
#25 0x0000000003e444fc _start (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x3e444fc)
flang-new: error: unable to execute command: Aborted (core dumped)
flang-new: error: flang frontend command failed due to signal (use -v to see invocation)
flang-new version 19.0.0git (https://github.com/llvm/llvm-project.git cbcdf126ccc774c063b5d5140c1393ff5305dded)
Target: aarch64-unknown-linux-gnu
Thread model: posix
InstalledDir: /lustre/home/ohno/LLVM_20240321/release/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/snfmm235_3-027b31
flang-new: note: diagnostic msg: /tmp/snfmm235_3-027b31.sh
flang-new: note: diagnostic msg:

********************
$

$ cat /tmp/snfmm235_3-027b31
#line "./snfmm235_3.f90" 1
      program main
      real(kind=16) r3/1/
!$omp parallel reduction(min:r3)
      r3=min(r3,1.0_16)
      write(6,*) "r3 = ", r3
      if(r3 < 1.0_16) then
      r3 = 1.0_16
      end if
!$omp end parallel
      write(6,*) "pass"
      end program main
$

$ cat /tmp/snfmm235_3-027b31.sh
# Crash reproducer for clang version 19.0.0git (https://github.com/llvm/llvm-project.git cbcdf126ccc774c063b5d5140c1393ff5305dded)
# Driver args: "-fopenmp" "snfmm235_3.f90"
# Original command:  "/lustre/home/ohno/LLVM_20240321/release/bin/flang-new" "-fc1" "-triple" "aarch64-unknown-linux-gnu" "-emit-obj" "-fopenmp" "-mrelocation-model" "pic" "-pic-level" "2" "-pic-is-pie" "-target-cpu" "generic" "-target-feature" "+outline-atomics" "-target-feature" "+v8a" "-target-feature" "+fp-armv8" "-target-feature" "+neon" "-mframe-pointer=non-leaf" "-o" "/tmp/snfmm235_3-4e13f7.o" "-x" "f95-cpp-input" "snfmm235_3.f90"
 "/lustre/home/ohno/LLVM_20240321/release/bin/flang-new" "-fc1" "-triple" "aarch64-unknown-linux-gnu" "-emit-obj" "-fopenmp" "-mrelocation-model" "pic" "-pic-level" "2" "-pic-is-pie" "-target-cpu" "generic" "-target-feature" "+outline-atomics" "-target-feature" "+v8a" "-target-feature" "+fp-armv8" "-target-feature" "+neon" "-mframe-pointer=non-leaf" "-x" "f95-cpp-input" "snfmm235_3-027b31"
$

$ export OMP_NUM_THREADS=2; gfortran -fopenmp snfmm235_3.f90; ./a.out
 r3 =    1.00000000000000000000000000000000000
 r3 =    1.00000000000000000000000000000000000
 pass
$

$ export OMP_NUM_THREADS=2; ifort -qopenmp -diag-disable=10448 snfmm235_3.f90; ./a.out
 r3 =    1.00000000000000000000000000000000000
 r3 =    1.00000000000000000000000000000000000
 pass
$

The text was updated successfully, but these errors were encountered:

llvmbot · 2024-04-08T01:49:44Z

@llvm/issue-subscribers-openmp

Author: None (ohno-fj)

``` Version of flang-new : 19.0.0(cbcdf12)/AArch64 ``` A compilation error occurs when using a `reduction` clause (`min`) with a variable (`kind=16`). The same error occurs when `max` intrinsic function is used.

Compilation terminates normally in the following cases:

change kind=16 to kind=8
remove reduction clause (min)
compile without -fopenmp option

The following are the test program, Flang-new, Gfortran and ifort compilation/execution result.

snfmm235_3.f90:

program main
  real(kind=16) r3/1/
!$omp parallel reduction(min:r3)
  r3=min(r3,1.0_16)
  write(6,*) "r3 = ", r3
  if (r3 &lt; 1.0_16) then
     r3 = 1.0_16
  end if
!$omp end parallel
  write(6,*) "pass"
end program main

$ flang-new -fopenmp snfmm235_3.f90
LLVM ERROR: Cannot select: t7: f128 = fminimum contract t4, t6
  t4: f128,ch = load&lt;(dereferenceable load (s128) from @<!-- -->_QFEr3)&gt; t0, t41, undef:i64
    t41: i64 = AArch64ISD::ADDlow t40, TargetGlobalAddress:i64&lt;ptr @<!-- -->_QFEr3&gt; 0 [TF=34]
      t40: i64 = AArch64ISD::ADRP TargetGlobalAddress:i64&lt;ptr @<!-- -->_QFEr3&gt; 0 [TF=1]
        t38: i64 = TargetGlobalAddress&lt;ptr @<!-- -->_QFEr3&gt; 0 [TF=1]
      t39: i64 = TargetGlobalAddress&lt;ptr @<!-- -->_QFEr3&gt; 0 [TF=34]
    t3: i64 = undef
  t6: f128,ch = load&lt;(dereferenceable load (s128) from %ir.1)&gt; t0, FrameIndex:i64&lt;1&gt;, undef:i64
    t5: i64 = FrameIndex&lt;1&gt;
    t3: i64 = undef
In function: _QQmain..omp_par
PLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.
Stack dump:
0.      Program arguments: /lustre/home/ohno/LLVM_20240321/release/bin/flang-new -fc1 -triple aarch64-unknown-linux-gnu -emit-obj -fopenmp -mrelocation-model pic -pic-level 2 -pic-is-pie -target-cpu generic -target-feature +outline-atomics -target-feature +v8a -target-feature +fp-armv8 -target-feature +neon -mframe-pointer=non-leaf -o /tmp/snfmm235_3-4e13f7.o -x f95-cpp-input snfmm235_3.f90
1.      Running pass 'Function Pass Manager' on module 'FIRModule'.
2.      Running pass 'AArch64 Instruction Selection' on function '@<!-- -->_QQmain..omp_par'
 #<!-- -->0 0x00000000041f9f00 llvm::sys::PrintStackTrace(llvm::raw_ostream&amp;, int) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x41f9f00)
 #<!-- -->1 0x00000000041f7e08 llvm::sys::RunSignalHandlers() (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x41f7e08)
 #<!-- -->2 0x00000000041f7f8c SignalHandler(int) Signals.cpp:0:0
 #<!-- -->3 0x000040001c8a07a0 (linux-vdso.so.1+0x7a0)
 #<!-- -->4 0x000040001cd46274 raise (/lib64/libc.so.6+0x36274)
 #<!-- -->5 0x000040001cd30a2c abort (/lib64/libc.so.6+0x20a2c)
 #<!-- -->6 0x0000000004182d58 llvm::report_fatal_error(llvm::StringRef, bool) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x4182d58)
 #<!-- -->7 0x0000000004676b10 llvm::SelectionDAGISel::CannotYetSelect(llvm::SDNode*) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x4676b10)
 #<!-- -->8 0x000000000467a91c llvm::SelectionDAGISel::SelectCodeCommon(llvm::SDNode*, unsigned char const*, unsigned int) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x467a91c)
 #<!-- -->9 0x0000000003f9be2c (anonymous namespace)::AArch64DAGToDAGISel::Select(llvm::SDNode*) AArch64ISelDAGToDAG.cpp:0:0
#<!-- -->10 0x0000000004674eac llvm::SelectionDAGISel::DoInstructionSelection() (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x4674eac)
#<!-- -->11 0x0000000004681830 llvm::SelectionDAGISel::CodeGenAndEmitDAG() (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x4681830)
#<!-- -->12 0x000000000468452c llvm::SelectionDAGISel::SelectAllBasicBlocks(llvm::Function const&amp;) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x468452c)
#<!-- -->13 0x00000000046859b0 llvm::SelectionDAGISel::runOnMachineFunction(llvm::MachineFunction&amp;) (.part.1268) SelectionDAGISel.cpp:0:0
#<!-- -->14 0x00000000068cbda0 llvm::MachineFunctionPass::runOnFunction(llvm::Function&amp;) (.part.78) MachineFunctionPass.cpp:0:0
#<!-- -->15 0x0000000007f91e90 llvm::FPPassManager::runOnFunction(llvm::Function&amp;) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x7f91e90)
#<!-- -->16 0x0000000007f920c8 llvm::FPPassManager::runOnModule(llvm::Module&amp;) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x7f920c8)
#<!-- -->17 0x0000000007f92bb4 llvm::legacy::PassManagerImpl::run(llvm::Module&amp;) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x7f92bb4)
#<!-- -->18 0x000000000485663c Fortran::frontend::CodeGenAction::executeAction() (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x485663c)
#<!-- -->19 0x000000000422f164 Fortran::frontend::FrontendAction::execute() (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x422f164)
#<!-- -->20 0x00000000042238dc Fortran::frontend::CompilerInstance::executeAction(Fortran::frontend::FrontendAction&amp;) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x42238dc)
#<!-- -->21 0x0000000004233de0 Fortran::frontend::executeCompilerInvocation(Fortran::frontend::CompilerInstance*) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x4233de0)
#<!-- -->22 0x0000000003e45678 fc1_main(llvm::ArrayRef&lt;char const*&gt;, char const*) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x3e45678)
#<!-- -->23 0x0000000003dccd10 main (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x3dccd10)
#<!-- -->24 0x000040001cd34384 __libc_start_main (/lib64/libc.so.6+0x24384)
#<!-- -->25 0x0000000003e444fc _start (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x3e444fc)
flang-new: error: unable to execute command: Aborted (core dumped)
flang-new: error: flang frontend command failed due to signal (use -v to see invocation)
flang-new version 19.0.0git (https://github.com/llvm/llvm-project.git cbcdf126ccc774c063b5d5140c1393ff5305dded)
Target: aarch64-unknown-linux-gnu
Thread model: posix
InstalledDir: /lustre/home/ohno/LLVM_20240321/release/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/snfmm235_3-027b31
flang-new: note: diagnostic msg: /tmp/snfmm235_3-027b31.sh
flang-new: note: diagnostic msg:

********************
$

$ cat /tmp/snfmm235_3-027b31
#line "./snfmm235_3.f90" 1
      program main
      real(kind=16) r3/1/
!$omp parallel reduction(min:r3)
      r3=min(r3,1.0_16)
      write(6,*) "r3 = ", r3
      if(r3 &lt; 1.0_16) then
      r3 = 1.0_16
      end if
!$omp end parallel
      write(6,*) "pass"
      end program main
$

$ cat /tmp/snfmm235_3-027b31.sh
# Crash reproducer for clang version 19.0.0git (https://github.com/llvm/llvm-project.git cbcdf126ccc774c063b5d5140c1393ff5305dded)
# Driver args: "-fopenmp" "snfmm235_3.f90"
# Original command:  "/lustre/home/ohno/LLVM_20240321/release/bin/flang-new" "-fc1" "-triple" "aarch64-unknown-linux-gnu" "-emit-obj" "-fopenmp" "-mrelocation-model" "pic" "-pic-level" "2" "-pic-is-pie" "-target-cpu" "generic" "-target-feature" "+outline-atomics" "-target-feature" "+v8a" "-target-feature" "+fp-armv8" "-target-feature" "+neon" "-mframe-pointer=non-leaf" "-o" "/tmp/snfmm235_3-4e13f7.o" "-x" "f95-cpp-input" "snfmm235_3.f90"
 "/lustre/home/ohno/LLVM_20240321/release/bin/flang-new" "-fc1" "-triple" "aarch64-unknown-linux-gnu" "-emit-obj" "-fopenmp" "-mrelocation-model" "pic" "-pic-level" "2" "-pic-is-pie" "-target-cpu" "generic" "-target-feature" "+outline-atomics" "-target-feature" "+v8a" "-target-feature" "+fp-armv8" "-target-feature" "+neon" "-mframe-pointer=non-leaf" "-x" "f95-cpp-input" "snfmm235_3-027b31"
$

$ export OMP_NUM_THREADS=2; gfortran -fopenmp snfmm235_3.f90; ./a.out
 r3 =    1.00000000000000000000000000000000000
 r3 =    1.00000000000000000000000000000000000
 pass
$

$ export OMP_NUM_THREADS=2; ifort -qopenmp -diag-disable=10448 snfmm235_3.f90; ./a.out
 r3 =    1.00000000000000000000000000000000000
 r3 =    1.00000000000000000000000000000000000
 pass
$

llvmbot · 2024-04-08T03:30:33Z

@llvm/issue-subscribers-flang-ir

Author: None (ohno-fj)

``` Version of flang-new : 19.0.0(cbcdf12)/AArch64 ``` A compilation error occurs when using a `reduction` clause (`min`) with a variable (`kind=16`). The same error occurs when `max` intrinsic function is used.

Compilation terminates normally in the following cases:

change kind=16 to kind=8
remove reduction clause (min)
compile without -fopenmp option

The following are the test program, Flang-new, Gfortran and ifort compilation/execution result.

snfmm235_3.f90:

program main
  real(kind=16) r3/1/
!$omp parallel reduction(min:r3)
  r3=min(r3,1.0_16)
  write(6,*) "r3 = ", r3
  if (r3 &lt; 1.0_16) then
     r3 = 1.0_16
  end if
!$omp end parallel
  write(6,*) "pass"
end program main

$ flang-new -fopenmp snfmm235_3.f90
LLVM ERROR: Cannot select: t7: f128 = fminimum contract t4, t6
  t4: f128,ch = load&lt;(dereferenceable load (s128) from @<!-- -->_QFEr3)&gt; t0, t41, undef:i64
    t41: i64 = AArch64ISD::ADDlow t40, TargetGlobalAddress:i64&lt;ptr @<!-- -->_QFEr3&gt; 0 [TF=34]
      t40: i64 = AArch64ISD::ADRP TargetGlobalAddress:i64&lt;ptr @<!-- -->_QFEr3&gt; 0 [TF=1]
        t38: i64 = TargetGlobalAddress&lt;ptr @<!-- -->_QFEr3&gt; 0 [TF=1]
      t39: i64 = TargetGlobalAddress&lt;ptr @<!-- -->_QFEr3&gt; 0 [TF=34]
    t3: i64 = undef
  t6: f128,ch = load&lt;(dereferenceable load (s128) from %ir.1)&gt; t0, FrameIndex:i64&lt;1&gt;, undef:i64
    t5: i64 = FrameIndex&lt;1&gt;
    t3: i64 = undef
In function: _QQmain..omp_par
PLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.
Stack dump:
0.      Program arguments: /lustre/home/ohno/LLVM_20240321/release/bin/flang-new -fc1 -triple aarch64-unknown-linux-gnu -emit-obj -fopenmp -mrelocation-model pic -pic-level 2 -pic-is-pie -target-cpu generic -target-feature +outline-atomics -target-feature +v8a -target-feature +fp-armv8 -target-feature +neon -mframe-pointer=non-leaf -o /tmp/snfmm235_3-4e13f7.o -x f95-cpp-input snfmm235_3.f90
1.      Running pass 'Function Pass Manager' on module 'FIRModule'.
2.      Running pass 'AArch64 Instruction Selection' on function '@<!-- -->_QQmain..omp_par'
 #<!-- -->0 0x00000000041f9f00 llvm::sys::PrintStackTrace(llvm::raw_ostream&amp;, int) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x41f9f00)
 #<!-- -->1 0x00000000041f7e08 llvm::sys::RunSignalHandlers() (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x41f7e08)
 #<!-- -->2 0x00000000041f7f8c SignalHandler(int) Signals.cpp:0:0
 #<!-- -->3 0x000040001c8a07a0 (linux-vdso.so.1+0x7a0)
 #<!-- -->4 0x000040001cd46274 raise (/lib64/libc.so.6+0x36274)
 #<!-- -->5 0x000040001cd30a2c abort (/lib64/libc.so.6+0x20a2c)
 #<!-- -->6 0x0000000004182d58 llvm::report_fatal_error(llvm::StringRef, bool) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x4182d58)
 #<!-- -->7 0x0000000004676b10 llvm::SelectionDAGISel::CannotYetSelect(llvm::SDNode*) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x4676b10)
 #<!-- -->8 0x000000000467a91c llvm::SelectionDAGISel::SelectCodeCommon(llvm::SDNode*, unsigned char const*, unsigned int) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x467a91c)
 #<!-- -->9 0x0000000003f9be2c (anonymous namespace)::AArch64DAGToDAGISel::Select(llvm::SDNode*) AArch64ISelDAGToDAG.cpp:0:0
#<!-- -->10 0x0000000004674eac llvm::SelectionDAGISel::DoInstructionSelection() (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x4674eac)
#<!-- -->11 0x0000000004681830 llvm::SelectionDAGISel::CodeGenAndEmitDAG() (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x4681830)
#<!-- -->12 0x000000000468452c llvm::SelectionDAGISel::SelectAllBasicBlocks(llvm::Function const&amp;) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x468452c)
#<!-- -->13 0x00000000046859b0 llvm::SelectionDAGISel::runOnMachineFunction(llvm::MachineFunction&amp;) (.part.1268) SelectionDAGISel.cpp:0:0
#<!-- -->14 0x00000000068cbda0 llvm::MachineFunctionPass::runOnFunction(llvm::Function&amp;) (.part.78) MachineFunctionPass.cpp:0:0
#<!-- -->15 0x0000000007f91e90 llvm::FPPassManager::runOnFunction(llvm::Function&amp;) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x7f91e90)
#<!-- -->16 0x0000000007f920c8 llvm::FPPassManager::runOnModule(llvm::Module&amp;) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x7f920c8)
#<!-- -->17 0x0000000007f92bb4 llvm::legacy::PassManagerImpl::run(llvm::Module&amp;) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x7f92bb4)
#<!-- -->18 0x000000000485663c Fortran::frontend::CodeGenAction::executeAction() (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x485663c)
#<!-- -->19 0x000000000422f164 Fortran::frontend::FrontendAction::execute() (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x422f164)
#<!-- -->20 0x00000000042238dc Fortran::frontend::CompilerInstance::executeAction(Fortran::frontend::FrontendAction&amp;) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x42238dc)
#<!-- -->21 0x0000000004233de0 Fortran::frontend::executeCompilerInvocation(Fortran::frontend::CompilerInstance*) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x4233de0)
#<!-- -->22 0x0000000003e45678 fc1_main(llvm::ArrayRef&lt;char const*&gt;, char const*) (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x3e45678)
#<!-- -->23 0x0000000003dccd10 main (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x3dccd10)
#<!-- -->24 0x000040001cd34384 __libc_start_main (/lib64/libc.so.6+0x24384)
#<!-- -->25 0x0000000003e444fc _start (/lustre/home/ohno/LLVM_20240321/release/bin/flang-new+0x3e444fc)
flang-new: error: unable to execute command: Aborted (core dumped)
flang-new: error: flang frontend command failed due to signal (use -v to see invocation)
flang-new version 19.0.0git (https://github.com/llvm/llvm-project.git cbcdf126ccc774c063b5d5140c1393ff5305dded)
Target: aarch64-unknown-linux-gnu
Thread model: posix
InstalledDir: /lustre/home/ohno/LLVM_20240321/release/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/snfmm235_3-027b31
flang-new: note: diagnostic msg: /tmp/snfmm235_3-027b31.sh
flang-new: note: diagnostic msg:

********************
$

$ cat /tmp/snfmm235_3-027b31
#line "./snfmm235_3.f90" 1
      program main
      real(kind=16) r3/1/
!$omp parallel reduction(min:r3)
      r3=min(r3,1.0_16)
      write(6,*) "r3 = ", r3
      if(r3 &lt; 1.0_16) then
      r3 = 1.0_16
      end if
!$omp end parallel
      write(6,*) "pass"
      end program main
$

$ cat /tmp/snfmm235_3-027b31.sh
# Crash reproducer for clang version 19.0.0git (https://github.com/llvm/llvm-project.git cbcdf126ccc774c063b5d5140c1393ff5305dded)
# Driver args: "-fopenmp" "snfmm235_3.f90"
# Original command:  "/lustre/home/ohno/LLVM_20240321/release/bin/flang-new" "-fc1" "-triple" "aarch64-unknown-linux-gnu" "-emit-obj" "-fopenmp" "-mrelocation-model" "pic" "-pic-level" "2" "-pic-is-pie" "-target-cpu" "generic" "-target-feature" "+outline-atomics" "-target-feature" "+v8a" "-target-feature" "+fp-armv8" "-target-feature" "+neon" "-mframe-pointer=non-leaf" "-o" "/tmp/snfmm235_3-4e13f7.o" "-x" "f95-cpp-input" "snfmm235_3.f90"
 "/lustre/home/ohno/LLVM_20240321/release/bin/flang-new" "-fc1" "-triple" "aarch64-unknown-linux-gnu" "-emit-obj" "-fopenmp" "-mrelocation-model" "pic" "-pic-level" "2" "-pic-is-pie" "-target-cpu" "generic" "-target-feature" "+outline-atomics" "-target-feature" "+v8a" "-target-feature" "+fp-armv8" "-target-feature" "+neon" "-mframe-pointer=non-leaf" "-x" "f95-cpp-input" "snfmm235_3-027b31"
$

$ export OMP_NUM_THREADS=2; gfortran -fopenmp snfmm235_3.f90; ./a.out
 r3 =    1.00000000000000000000000000000000000
 r3 =    1.00000000000000000000000000000000000
 pass
$

$ export OMP_NUM_THREADS=2; ifort -qopenmp -diag-disable=10448 snfmm235_3.f90; ./a.out
 r3 =    1.00000000000000000000000000000000000
 r3 =    1.00000000000000000000000000000000000
 pass
$

…perators (#89258) This patch changes the lowering of max and min to be lowered to arith::MaxNumFop and arith::MinNumFOp instead of using arith::MaximumFOp and arith::MinimumFOp. The arith::MaximumFOp and arith::MinimumFOp map to the corresponding intrinsics llvm.maximum.* and llvm.minimum.* intrinsics which conform to the semantics specified in the draft of IEEE 754-2019, which is not supported by all hardware. Instead using arith::MaximumFOp and arith::MinimumFOp will allow code generation for more targets and match the code generated by clang OpenMP. fixes #87955

…perators (llvm#89258) This patch changes the lowering of max and min to be lowered to arith::MaxNumFop and arith::MinNumFOp instead of using arith::MaximumFOp and arith::MinimumFOp. The arith::MaximumFOp and arith::MinimumFOp map to the corresponding intrinsics llvm.maximum.* and llvm.minimum.* intrinsics which conform to the semantics specified in the draft of IEEE 754-2019, which is not supported by all hardware. Instead using arith::MaximumFOp and arith::MinimumFOp will allow code generation for more targets and match the code generated by clang OpenMP. fixes llvm#87955

ohno-fj added openmp flang Flang issues not falling into any other category labels Apr 8, 2024

EugeneZelenko added flang:ir flang:openmp and removed openmp flang Flang issues not falling into any other category labels Apr 8, 2024

kiranchandramohan mentioned this issue Apr 18, 2024

[flang][OpenMP] Use maxnum/minnum for lowering of max/min reduction operators #89258

Merged

kparzysz mentioned this issue Apr 18, 2024

Codegen failure for llvm.{min,max}imum(fp128, fp128) #89278

Open

jsjodin closed this as completed in #89258 Apr 19, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[Flang][OpenMP] Compilation error of using a reduction clause (min) with a variable (kind=16) #87955

[Flang][OpenMP] Compilation error of using a reduction clause (min) with a variable (kind=16) #87955

ohno-fj commented Apr 8, 2024

llvmbot commented Apr 8, 2024

llvmbot commented Apr 8, 2024

[Flang][OpenMP] Compilation error of using a reduction clause (min) with a variable (kind=16) #87955

[Flang][OpenMP] Compilation error of using a reduction clause (min) with a variable (kind=16) #87955

Comments

ohno-fj commented Apr 8, 2024

llvmbot commented Apr 8, 2024

llvmbot commented Apr 8, 2024