LLVM/NVVM kernel crashes LibNVVM

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi everyone,

I have no luck compiling the following NVVM IR kernel with libNVVM (Version 1.0, as available for download in devzone). The compilation step fails with:

Stack dump:
0. Running pass ‘Function Pass Manager’ on module ‘’.

  1. Running pass ‘Simplify the CFG’ on function ‘@function’
    Segmentation fault (core dumped)

The API calls go along the simple example.

Please find the kernel source code here: [url]https://gist.github.com/fwinter/5525072[/url]

2

Is somebody looking into this?

To reproduce the issue just load the kernel with simple.c (instead of simple-gpu64.ll).

It would be very helpful to know if it’s my fault (malformed kernel) or there is an issue with libNVVM.

Maybe this is already solved and doesn’t happen any more with the Compiler SDK in CUDA 5.5. In which case it would also be good to know.

Best wishes,
Frank