Bug 99539 - Make LAMMPS OpenCL GPU support work on Clover and RadeonSI
Summary: Make LAMMPS OpenCL GPU support work on Clover and RadeonSI
Status: RESOLVED FIXED
Alias: None
Product: Mesa
Classification: Unclassified
Component: Drivers/Gallium/radeonsi (show other bugs)
Version: git
Hardware: Other All
: medium normal
Assignee: Vedran Miletić
QA Contact: Vedran Miletić
URL: http://lammps.sandia.gov/
Whiteboard:
Keywords:
Depends on:
Blocks: 99553
  Show dependency treegraph
 
Reported: 2017-01-25 19:32 UTC by Vedran Miletić
Modified: 2017-10-28 14:48 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Vedran Miletić 2017-01-25 19:32:48 UTC
LAMMPS is a popular molecular dynamics software that provides GPU acceleration using OpenCL: http://lammps.sandia.gov/doc/accelerate_gpu.html

Right now running fails with:

--------------------------------------------------------------------------
- Using acceleration for lj/cut:
-  with 1 proc(s) per device.
-  with OpenCL Parameters for: AMD Cypress
--------------------------------------------------------------------------
Device 0: AMD HAWAII (DRM 2.48.0 / 4.9.5-200.fc25.x86_64, LLVM 5.0.0), 44 CUs, 16 GB, 0.93 GHZ (Mixed Precision)
--------------------------------------------------------------------------

Initializing Device and compiling on process 0...Done.
Initializing Device 0 on core 0...Done.

Setting up Verlet run ...
  Unit style    : lj
  Current step  : 0
  Time step     : 0.0025
OpenCL error in file 'geryon/ocl_kernel.h' in line 218 : -48.
--------------------------------------------------------------------------
MPI_ABORT was invoked on rank 0 in communicator MPI_COMM_WORLD
with errorcode -1.

NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes.
You may or may not see output from other processes, depending on
exactly when Open MPI kills them.
--------------------------------------------------------------------------
Comment 1 Vedran Miletić 2017-02-24 19:50:24 UTC
After adding native_recip() like in bug 99764:

$ cd examples/accelerate
$ ../../src/lmp_serial -i in.lj -suffix gpu

$ cd examples/balance
$ ../../src/lmp_serial -i in.balance -suffix gpu

OpenCL kernel compiles fine and both examples run.
Comment 2 Jan Vesely 2017-10-28 14:48:19 UTC
native_recip was added to libclc in r313107


Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.