config.inc.sh

# Configuration file

# Required configuration for the platform specific properties
declare -A platform_routines

# Uncomment for Raptor Lake system
source "${SCRIPTDIR}/configs/raptor.conf"

# Uncomment for Odroid system
#source "${SCRIPTDIR}/configs/odroid.conf"

# All the configuration options for TETRiS
source "${SCRIPTDIR}/configs/tetris.conf"

export BINDIR="${SCRIPTDIR}/bench"

# Important variables that influence the benchmark

# 1) The list of applications (== scenarios) that should be benchmarked
# This benchmark suite can run 3 application dialects
#   a) A simple binary without any additional arguments or configuration
#   b) A binary that expects additional arguments which are not TETRiS specific
#      Such scenarios should be prefixed with a '?'.
#   c) A binary/script that expects as argument the current benchmarking mode
#      and will run and configure the application by itself. Importantly, this binary/script
#      will also do the necessary TETRiS specific configuration.
#      Such scenarios should be prefixed with a '!'.
scenarios=("abc|bc" "!foobar|bar|!foo" "?blub|foo")

# 2) The number of repetitions per application and benchmarking mode in order to
#    produce statistically reliable values. At least 10 should be chosen here.
runs=10

# 3) The modes at which the applications should be benchmarked.
# Currently the script supports the following modes:
#   a) *cfs* -- The application will be executed on without any additional configuration/management
#      directly on the Linux default scheduler (which might be EEVDF on newer Linux kernels).
#   b) *td* -- The Thread Director extension of modern Intel CPUs will be used to manage the applications
#      at runtime.
#   c) *tetris* -- The TETRiS resource manager will control the applications at runtime.
#   d) *tetris-offline* -- The TETRiS resource manager will control the applications at runtime based on
#      offline generated application characteristics.
#   e) *tetris-grouped* -- The same as *tetris* but with a learning phase where applications are executed
#      in parallel and not separately
#   f) *tetris-noscaling* -- The same as *tetris* but dynamic application scaling is disabled
#   h) *learning* -- Measure the learning phase of tetris
#   g) *overhead* -- Measure the overhead of the whole TETRiS approach without applying the resource decisions
modes=("cfs" "td" "tetris" "tetris-offline")


# Other optional configuration variables

# Try to continue the benchmark where we last left of. (Default = 0)
#continuation=1

# Output directories for results, logs and other intermediate data. (Default = auto)

# Directory where the final benchmark result should be stored
#result_dir=auto

# Directory where log output of the clients and the server should be stored for the run
#log_dir=auto

# Directory where resource allocation traces should be stored
#trace_dir=auto

# Directory where the OP learning database should be stored
#learn_dir=auto