Quick Start: 1D Shock Tube - hyschive/gamer-fork GitHub Wiki

This page includes three demos:

  1. CPU-only without OpenMP
  2. CPU-only with OpenMP
  3. Hybrid OpenMP/GPU

CPU-only without OpenMP

1. Move to the source directory.

cd src

2. Generate Makefile by configure.py.

cp ../example/test_problem/Hydro/Riemann/generate_make.sh ./
sh generate_make.sh --openmp=false
Execution results 
   ...
   ...
========================================
Makefile is created.
========================================

Note

We have set --mpi=false, --gpu=false, and --openmp=false to run in a CPU-only mode without OpenMP and MPI. See Option List for a complete list of all available options of configure.py.

3. Compile the code.

make clean
make
Execution results 
   ...
   ...
Compiling GAMER --> Successful!

4. Create a working directory.

cd ../bin
mkdir shocktube
cd shocktube

5. Copy the GAMER executable and example files of the test problem to the working directory.

cp -r ../../example/test_problem/Hydro/Riemann/* .
cp ../gamer .
ls
Execution results 
clean.sh  gamer  Input__Flag_Lohner  Input__Parameter  Input__TestProb  plot__hydro_dens.gpt  plot__mhd.gpt  README  ReferenceSolution

6. Run the code. It will display ~ GAMER OVER ~ if it succeeds.

./gamer
Execution results 
   ...
   ...
Time: 9.0000000e-02 -> 9.3057198e-02,   Step:      29 ->      30,   dt_base: 3.0571979e-03
Time: 9.3057198e-02 -> 9.6446514e-02,   Step:      30 ->      31,   dt_base: 3.3893166e-03
Time: 9.6446514e-02 -> 9.9834923e-02,   Step:      31 ->      32,   dt_base: 3.3884083e-03
Time: 9.9834923e-02 -> 1.0000000e-01,   Step:      32 ->      33,   dt_base: 1.6507717e-04
Output_DumpData_Part (DumpID = 10) ...
Output_DumpData_Part (DumpID = 10) ... done
End_GAMER ...
End_MemFree ... done
End_GAMER ... done


~ GAME OVER ~

7. The code will generate several log files Record__* and a series of 1D text data files Xline_y0.000_z0.000_*.

ls
Execution results 
clean.sh              plot__mhd.gpt       Record__PatchCount          Xline_y0.000_z0.000_000001  Xline_y0.000_z0.000_000007
gamer                 README              Record__Performance         Xline_y0.000_z0.000_000002  Xline_y0.000_z0.000_000008
Input__Flag_Lohner    Record__Dump        Record__TimeStep            Xline_y0.000_z0.000_000003  Xline_y0.000_z0.000_000009
Input__Parameter      Record__MemInfo     Record__Timing              Xline_y0.000_z0.000_000004  Xline_y0.000_z0.000_000010
Input__TestProb       Record__NCorrUnphy  ReferenceSolution           Xline_y0.000_z0.000_000005
plot__hydro_dens.gpt  Record__Note        Xline_y0.000_z0.000_000000  Xline_y0.000_z0.000_000006

8. Plot a 1D text data file. You can use the sample gnuplot script plot__hydro_dens.gpt (replace display by other image viewers if necessary).

gnuplot plot__hydro_dens.gpt
display Fig__Riemann_Density_000010.png
Execution results

shocktube

9. Check the performance. See Log Files for more detailed timing analysis.

tail -n 3 Record__Note
Execution results 
Total Processing Time : 75.954923 s

CPU-only with OpenMP

Next, we enable OpenMP for the same test problem. Repeat the steps above with the following modifications.

1. Re-generate Makefile by configure.py and recompile gamer.

sh generate_make.sh --openmp=true
make clean
make -j4

Caution

Remember to copy the new executable to bin/shocktube.

2. Set the number of OpenMP threads by editing the runtime parameter OMP_NTHREAD in the input file Input__Parameter. The following example uses 4 threads.

OMP_NTHREAD      4      # number of OpenMP threads (<=0=auto) [-1]

Caution

All input files Input__* must be put in the same directory as the executable gamer.

3. Remove all old log and data files, if necessary. You can use the helper script clean.sh.

sh clean.sh
ls
Execution results 
clean.sh  gamer  Input__Flag_Lohner  Input__Parameter  Input__TestProb  plot__hydro_dens.gpt  plot__mhd.gpt  README  ReferenceSolution

4. Run the code in the same way as without OpenMP.

./gamer

5. Validate the OpenMP settings by searching for the keyword "OpenMP" in the log file Record__Note. You should see something like

OpenMP Diagnosis
***********************************************************************************
OMP__SCHEDULE                   DYNAMIC
OMP__SCHEDULE_CHUNK_SIZE        1
OMP__NESTED                     OFF

CPU core IDs of all OpenMP threads (tid == thread ID):
------------------------------------------------------------------------
 Rank        Host  NThread  tid-00  tid-01  tid-02  tid-03
    0    golub123        4       2       5       4       7
***********************************************************************************

Check the following things:

  • The number under NThread is the same as the runtime parameter OMP_NTHREAD you just set
  • Different threads use different CPU cores

6. Check the performance. It should be about OMP_NTHREAD times faster than the case without OpenMP.

tail -n 3 Record__Note
Execution results 
Total Processing Time : 20.460586 s

Hybrid OpenMP/GPU

To enable both GPU and OpenMP, repeat the steps in CPU-only with OpenMP with the following modifications.

1. Re-generate Makefile by configure.py and recompile gamer.

sh generate_make.sh --openmp=true --gpu=true
make clean
make -j4

Caution

  • Please make sure that the GPU_COMPUTE_CAPABILITY is set properly in your machine configuration file
  • Remember to copy the new executable to bin/shocktube.

2. Remove all old log and data files, if necessary. Run the code with the new executable.

sh clean.sh
./gamer

3. Validate the GPU settings by searching for the keyword "Device Diagnosis" in the log file Record__Note. You should see something like

Device Diagnosis
***********************************************************************************
MPI_Rank =   0, hostname =   golub123, PID = 47842

CPU Info :
CPU Type        : Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
CPU MHz         : 2499.982
Cache Size      : 25600 KB
CPU Cores       : 10
Total Memory    : 63.0 GB

GPU Info :
Number of GPUs                    : 2
GPU ID                            : 0
GPU Name                          : Tesla K40m
CUDA Driver Version               : 8.0
CUDA Runtime Version              : 7.0
CUDA Major Revision Number        : 3
CUDA Minor Revision Number        : 5
Clock Rate                        : 0.745000 GHz
Global Memory Size                : 11439 MB
Constant Memory Size              : 64 KB
Shared Memory Size per Block      : 48 KB
Number of Registers per Block     : 65536
Warp Size                         : 32
Number of Multiprocessors:        : 15
Number of Cores per Multiprocessor: 192
Total Number of Cores:            : 2880
Max Number of Threads per Block   : 1024
Max Size of the Block X-Dimension : 1024
Max Size of the Grid X-Dimension  : 2147483647
Concurrent Copy and Execution     : Yes
Concurrent Up/Downstream Copies   : Yes
Concurrent Kernel Execution       : Yes
GPU has ECC Support Enabled       : Yes
***********************************************************************************

This example shows that we are running on the computing node golub123 with 2 GPUs, and we are using the one with GPU ID = 0, a Tesla K40m GPU.

4. Check the performance. It should be noticeably faster than the case without GPU.

tail -n 3 Record__Note
Execution results 
Total Processing Time : 9.417532 s

Links

Add a custom footer
⚠️ **GitHub.com Fallback** ⚠️