Small Time step error

[gihanmudalige]

I have compiled this version of Cloverleaf with PGI compilers (pgfortran 15.1-0) and when running on our K80 GPU (well 1/2 of the K80 as there are two GPUs on one card), I get a small timestep error at step 476.

The clover.in I am using is:

*clover
state 1 density=0.2 energy=1.0
state 2 density=1.0 energy=2.5 geometry=cuboid xmin=0.0 xmax=5.0 ymin=0.0 ymax=2.0 zmin=0.0 zmax=4.0

x_cells=96
y_cells=96
z_cells=96

xmin=0.0
ymin=0.0
zmin=0.0
xmax=10.0
ymax=10.0
zmax=10.0

initial_timestep=0.1
max_timestep=0.1
end_step=2955
end_time=1000.1
!profiler_on=0

use_opencl_kernels
opencl_vendor=nvidia
opencl_type=GPU
opencl_device=0

!profiler_on=1

*endclover

Did any one else get this error ?

[waynegaudin]

As this is OpenCL, could you have a look at it Mike?

A small timestep could be caused by a bug in almost any kernel leading to
incorrect compression, or it could be the reduction isn't working properly
for the timestep.

Do you know if the field summaries match the Fortran reference version
before the bad time step or do they diverge early from on in the
calculation?

On Mon, Jun 29, 2015 at 1:07 PM, Gihan R. Mudalige <[email protected]

wrote:

I have compiled this version of Cloverleaf with PGI compilers (pgfortran
15.1-0) and when running on our K80 GPU (well 1/2 of the K80 as there are
two GPUs on one card), I get a small timestep error at step 476.

The clover.in I am using is:

*clover
state 1 density=0.2 energy=1.0
state 2 density=1.0 energy=2.5 geometry=cuboid xmin=0.0 xmax=5.0 ymin=0.0
ymax=2.0 zmin=0.0 zmax=4.0

x_cells=96
y_cells=96
z_cells=96

xmin=0.0
ymin=0.0
zmin=0.0
xmax=10.0
ymax=10.0
zmax=10.0

initial_timestep=0.1
max_timestep=0.1
end_step=2955
end_time=1000.1
!profiler_on=0

use_opencl_kernels
opencl_vendor=nvidia
opencl_type=GPU
opencl_device=0

!profiler_on=1

*endclover

Did any one else get this error ?

—
Reply to this email directly or view it on GitHub
#1.

[gihanmudalige]

Actually this is OpenACC.

On my Machine -
Field summary at step 460:
Volume Mass Density Pressure Internal Energy Kinetic Energy Total Energy
step: 460 0.1000E+04 0.2342E+03 0.2342E+00 0.1144E+00 0.2861E+03 0.9795E+01 0.2959E+03

Field summary at step 460 (from the reference version):
Volume Mass Density Pressure Internal Energy Kinetic Energy Total Energy
step: 460 0.1000E+04 0.2340E+03 0.2340E+00 0.1167E+00 0.2917E+03 0.4349E+01 0.2960E+03

So it seems something is wrong. Both are compiled using the pgi Fortran compilers