Benchmark Runner Scripts

[Quincunx271]

I've updated my benchmark running scripts and figured I might as well PR them.

Sample:

export SHOCDRIVER="/path/to/shoc/build/bin/shocdriver"
export OPTSCHEDCFG="/path/to/optsched-cfg"
export VALIDATE_CMD="~/scripts/validate"
export ANALYZE_CMD="~/scripts/analyze"
export RUNNER_GIT_REPO="/path/to/OptSched"

# ~/scripts/validate
#!/usr/bin/env bash
# This is a modified validation-test.py script that I use
python3 ~/OptSched/util/misc/validation-test.py --benchsuite=$BENCHSUITE no.*/ yes.*/ | tee validation.txt

# ~/scripts/analyze
#!/usr/bin/env bash
~/OptSched/util/gt_analysis/gt_cmp.py --benchsuite=$BENCHSUITE no.*/ yes.*/ | tee nodesx.csv
~/OptSched/util/gt_analysis/gt_cmp.py --benchsuite=$BENCHSUITE --pass-num=1 no.*/ yes.*/ | tee nodesx1.csv
~/OptSched/util/gt_analysis/gt_cmp.py --benchsuite=$BENCHSUITE --pass-num=2 no.*/ yes.*/ | tee nodesx2.csv

The actual run command (~/runners is a symlink to ~/OptSched/util/runners):

BENCHSUITE=plaidml python3.8 ~/runners/multirun.py -o ~/local-results/output/$(~/runners/datename) -L yes,no --with STATIC_NODE_SUPERIORITY=YES 2ND_PASS_ILP_NODE_SUPERIORITY_PRESERVING_OCCUPANCY=YES --with STATIC_NODE_SUPERIORITY=NO 2ND_PASS_ILP_NODE_SUPERIORITY_PRESERVING_OCCUPANCY=NO -- ~/run-plaidbench.py {{label}}.logs -b xception

This does two runs of the ~/run-plaidbench.py command, one with the first --with's settings, and one with the second --with's settings.

---

It does require Python 3.8; Python 3.6 doesn't have capture_output on the subprocess.run(...) command.

I'll also probably rebase this to simplify the git history if we do want to merge this.