AnsweredAssumed Answered

clEnqueueWriteBuffer() crashes on a OpenCL CPU device

Question asked by thejascr on Jun 22, 2013
Latest reply on Jun 24, 2013 by himanshu.gautam

We are trying to run an OpenCL program simultaneously on a CPU and GPU device on AMD fusion box.

But whenever we try to use clEnqueueWriteBuffer() on the CPU opencl device, the program crashes.

Please find the clinfo for the machine attached. The same program runs fine with when run on GPU opencl device.

Any quick help on any workarounds or fixes would be appreciated.

 

Thanks

 

============================================================================

Program terminated with signal 11, Segmentation fault.

#0  0x00007f2abaedbaa9 in clEnqueueWriteBuffer () from /usr/lib/fglrx/libOpenCL.so.1

(gdb) bt

#0  0x00007f2abaedbaa9 in clEnqueueWriteBuffer () from /usr/lib/fglrx/libOpenCL.so.1

#1  0x0000000000404bd7 in transferrow_point (device=0x1e0aa78, kargs=0x1f58040, srow=831, ncols=<optimized out>, recv_proc=<optimized out>, ptr=<optimized out>, nrows=<optimized out>, wo=0, scol=1)

    at heat2d.cpp:694

#2  0x0000000000409060 in pack_recv_point_1_0 (recv_proc=1, kargs=0x1f58040, device=0x1e0aa78, ts8=<optimized out>, ts6=<optimized out>, ts4=25, ts2=0, ts1=<optimized out>, ts3=<optimized out>,

    ts5=<optimized out>, ts7=<optimized out>, prt=<optimized out>) at heat2d.cpp:898

#3  pack_point_1_0 (nprocs=2, my_rank=0, prt=0x1e0ab08, kargs=0x1f58040, device=0x1e0aa78, ts4=25, ts2=0, ts1=<optimized out>, ts3=<optimized out>) at heat2d.cpp:917

#4  Device::CopyFlowOutData_dsfo_with_tiling_1 (this=0x1e0aa78, args=0x1f58040) at heat2d.cpp:961

#5  0x0000000000409d73 in Device::CopyOut (this=0x1e0aa78, args=0x1f58040) at heat2d.cpp:1231

#6  0x0000000000409f45 in threadFunc (args=0x1f58040) at heat2d.cpp:1537

#7  0x00007f2abb0e6efc in start_thread () from /lib/x86_64-linux-gnu/libpthread.so.0

#8  0x00007f2aba479f4d in clone () from /lib/x86_64-linux-gnu/libc.so.6

#9  0x0000000000000000 in ?? ()

Attachments

Outcomes