Run the gpu applications with both ariel and mmio balar.

1. Two SST installations
   1. One with MMIO balar
   2. One with ariel balar
2. Need to download gpu-app-collection from accel-sim repo
   1. Compile with both CUDA 9.1 and 10.1
3. Need to download NVBIT tool in cuda_api_tracer/ folder
4. Need to have a python environment installing the packages in requirements.txt, could use virtualenv to create a sandbox
5. get_traces.py script
   1. Need to source gpu-app-collection/src/setup_environment before calling
   2. Runs with CUDA 10.1 and GCC 5.5.0
   3. But higher GCC version should also work
6. run_traces.py script
   1. Need to source setup_environment before calling
   2. Need CUDA 9.1 and GCC 4.9.2 for the ariel balar to works
   3. MMIO balar works with CUDA 10.1 and GCC 5.5.0

# Get trace
python3 get_traces.py -B [BENCHMARKS]

# Run traces
python3 run_traces.py -B [BENCHMARKS]

# Example
python3 run_traces.py -B GPU_Microbenchmark,rodinia_2.0-ft --original_balar_sst_exe=~/SST-Integration/sstcore-11.0.0-release/bin/sst

# Collecting stats
## MMIO example
python3 convert_results.py -B GPU_Microbenchmark,rodinia_2.0-ft --statfile=mmio_stats.out --gpgpusim_statfile=gpgpu_inst_stats_mmio.log -o mmio_stats.json
## Ariel/Original balar example
python3 convert_results.py -B GPU_Microbenchmark,rodinia_2.0-ft --statfile=original_stats.out --gpgpusim_statfile=gpgpu_inst_stats_original.log -o original_stats.json

1. sst testBalar-simple.py --model-options='-c ariel-gpu-v100.cfg -v -x run -t trace/cuda_calls.trace'
2. Use sst 10.1 env setup script

1. ~/SST-Integration/sstcore-11.1.0-release/bin/sst cuda-test.py --model-options='-c ariel-gpu-v100.cfg -v -x vectorAdd/vectorAdd'
   1. Use the other sst installation
   2. Need to use cuda 9.1?

1. NVBit tool should compiled with nvcc version >= 10.2
   1. tracer_tool setup script
2. Test app are compiled with CUDA 10.1 version
   1. test_app setup script
3. Need both 9.1 and 10.1 app
   1. 9.1 for simulation run
   2. 10.1 for tracer tool run

1. Use gpu app from accel sim
   1. Need automatic script to do the work
2. Python script launch cannot find cudart for app
3. Rodinia backprop benchmark cannot trace, kernel parameter is null
4. Rodinia 2.0 Backprop
   1. second kernel has issue
   2. Due to alignment setting
      1. Irregular argument size need to align properly?
      2. Implement this in the testcpu
5. Link the python run script and cfg file for SST in run script?
6. Create a clean script to clear all tmp files and option to not dump test data?
   1. Due to disk space concern
7. In hw traces, save multiple cuda version traces?
8. Parallelize the run traces
9. Minimize the python requirement file? (iPython not needed, just the YAML file?)