We have targets like:
|
plot-dacapo-results: |
|
bin/csv_to_krun_json -u "`uname -a`" -v HotSpot -l Java dacapo.hotspot.results |
|
bin/mark_outliers_in_json -w ${WINDOW_SIZE} -t ${OUTLIER_THRESHOLD} dacapo.hotspot.json.bz2 |
|
LD_LIBRARY_PATH=${R_LD_LIBRARY_PATH}:${LD_LIBRARY_PATH} bin/mark_changepoints_in_json -s ${STEADY_STATE_EXPECTED} dacapo.hotspot_outliers_w${WINDOW_SIZE}.json.bz2 |
|
bin/plot_krun_results --wallclock-only -w ${WINDOW_SIZE} -m -t --with-outliers -o dacapo.hotspot_${PLOTS_NO_CPTS} dacapo.hotspot_outliers_w${WINDOW_SIZE}.json.bz2 |
|
bin/plot_krun_results --wallclock-only -w ${WINDOW_SIZE} -m -t --with-outliers --with-changepoints -o dacapo.hotspot_${PLOTS_WITH_CPTS} dacapo.hotspot_outliers_w${WINDOW_SIZE}_changepoints.json.bz2 |
|
if [ -e dacapo.graal.results ]; then \ |
|
bin/csv_to_krun_json -u "`uname -a`" -v Graal -l Java dacapo.graal.results && \ |
|
bin/mark_outliers_in_json -w ${WINDOW_SIZE} -t ${OUTLIER_THRESHOLD} dacapo.graal.json.bz2 && \ |
|
LD_LIBRARY_PATH=${R_LD_LIBRARY_PATH}:${LD_LIBRARY_PATH} bin/mark_changepoints_in_json -s ${STEADY_STATE_EXPECTED} dacapo.graal_outliers_w${WINDOW_SIZE}.json.bz2 && \ |
|
bin/plot_krun_results --wallclock-only -w ${WINDOW_SIZE} -m -t --with-outliers -o dacapo.graal_${PLOTS_NO_CPTS} dacapo.graal_outliers_w${WINDOW_SIZE}.json.bz2 && \ |
|
bin/plot_krun_results --wallclock-only -w ${WINDOW_SIZE} -m -t --with-outliers --with-changepoints -o dacapo.graal_${PLOTS_WITH_CPTS} dacapo.graal_outliers_w${WINDOW_SIZE}_changepoints.json.bz2; \ |
|
fi |
Ideally each of the intermediate files would be its own target and dependencies could then be specified.
Note also that dacapo/octane targets assume the plotting targets have run and that the residual intermediate files are existing. These targets should use the dependencies introduced by the above step.
We have targets like:
warmup_experiment/Makefile
Lines 93 to 105 in 6330fa2
Ideally each of the intermediate files would be its own target and dependencies could then be specified.
Note also that dacapo/octane targets assume the plotting targets have run and that the residual intermediate files are existing. These targets should use the dependencies introduced by the above step.