I have come across a problem where nvcc (specifically cicc) is hanging when compiling a program. I am using version 5.0.35. Here is one traceback of the hang:
(gdb) bt
#0 0x00002b9cef8a765b in raise () from /lib64/libpthread.so.0
#1 0x000000000128a561 in ?? ()
#2 <signal handler called>
#3 0x0000000000bcbe86 in (anonymous namespace)::MachineConstProp::constProp(llvm::MachineInstr*) ()
#4 0x0000000000bcc36f in (anonymous namespace)::MachineConstProp::runOnMachineFunction(llvm::MachineFunction&) ()
#5 0x0000000000be257d in llvm::MachineFunctionPass::runOnFunction(llvm::Function&) ()
#6 0x0000000001216055 in llvm::FPPassManager::runOnFunction(llvm::Function&)
()
#7 0x00000000012162fd in llvm::FunctionPassManagerImpl::run(llvm::Function&)
()
#8 0x0000000001216526 in llvm::FunctionPassManager::run(llvm::Function&) ()
#9 0x00000000008ef6a1 in module2nvptx(int, char**, llvm::Module*, std::string*, char**) ()
#10 0x00000000008e14ca in nvvmCompileCU ()
#11 0x00000000006b3f23 in libNVVMEntry(char const*, void*, Command&, char**, int*, bool, bool, bool, bool) ()
#12 0x00000000006b8ce6 in ciCompile(char const*, char const*, char const*, char const*, char const*, bool, bool, bool) ()
#13 0x00000000006b2d62 in ciccOffline ()
#14 0x00002b9cf026dc36 in __libc_start_main () from /lib64/libc.so.6
—Type <return> to continue, or q <return> to quit—
#15 0x00000000006b248a in _start ()
Does this problem look familar? I am more than happy to submit a bug with a test case, but am not finding the appropriate NVIDIA website to do this.
Thanks,
Geir Johansen
Cray Inc.