| Age | Commit message (Collapse) | Author | Lines | |
|---|---|---|---|---|
| 2021-03-08 | documentation++ | Daniel Friesel | -1/+48 | |
| 2021-03-08 | pelt: Add tail_state_only option (exclude all but the last changepoint) | Daniel Friesel | -0/+21 | |
| 2021-02-24 | pelt: allow contraction (averaging) of data as well as stretching | Daniel Friesel | -2/+23 | |
| 2021-02-23 | fix drift compensation and reduce pelt + caching overhead | Daniel Friesel | -19/+29 | |
| 2021-02-22 | PELT: Increase parallelism | Daniel Friesel | -77/+94 | |
| 2021-02-19 | start modeling of individual sub-states | Daniel Friesel | -3/+16 | |
| 2021-02-18 | pelt: Set minimum penalty to 1, improve cache layout | Daniel Friesel | -11/+19 | |
| 2021-02-18 | pelt: add tail substate | Daniel Friesel | -3/+5 | |
| 2021-02-18 | pelt: cache changepoint detection results | Daniel Friesel | -5/+65 | |
| 2021-02-15 | pelt: Fix stretch != 1 support | Daniel Friesel | -2/+4 | |
| 2021-02-15 | debug output | Daniel Friesel | -6/+11 | |
| 2021-02-15 | Limit PELT via --with-substates=name_filter=state_or_transition_name | Daniel Friesel | -0/+1 | |
| 2021-02-15 | needs_refinement: short traces cannot be reliably refined | Daniel Friesel | -1/+3 | |
| 2021-02-12 | PELT: use variable penalty for benchmark repetations | Daniel Friesel | -31/+42 | |
| Previously, the first measurement of each by_param entry determined its penalty, and all subsequenc measurements in it were analyzed with this penalty. This lead to some of them coming up with a different number of sub-states and being ignored. Now, all by_param entries are passed to the kneedle / plateau algorithm, and each gets its own penalty. In most cases, this leads to all entries rporting the same number of sub-states, so far less need to be ignored. | ||||
| 2021-01-15 | pelt: add stretch parameter | Daniel Friesel | -0/+20 | |
| 2021-01-15 | pelt: Fix Dynp support | Daniel Friesel | -1/+5 | |
| 2021-01-14 | EnergyTrace drift compensation: optimize via shortest paths | Daniel Friesel | -4/+20 | |
| Right now performance is worse than the previous greedy approach as the dijkstra variant does not take into account that a transition may be missing from the set of detected changepoints (i.e., it assumes that the set of detected changepoints contains the transition timestamp, which is not always the case). This will be fixed in the next commit by adding nodes for the expected transition timestamp (with a slightly higher weight to ensure there are no proper nodes in the vicinity) -- or something similar. | ||||
| 2021-01-13 | pelt: Allow overrides via environment variables | Daniel Friesel | -0/+10 | |
| 2020-12-14 | energytrace: add pelt-based drift compensation experiment. | Daniel Friesel | -5/+12 | |
| Enable with DFATOOL_COMPENSATE_DRIFT=1 so far it's pretty unreliable. | ||||
| 2020-11-30 | pelt: allow disabling multiprocessing | Daniel Friesel | -2/+6 | |
| 2020-11-26 | some comments | Daniel Friesel | -2/+1 | |
| 2020-11-03 | analyze number of substates per state | Daniel Friesel | -70/+60 | |
| 2020-11-02 | PELT: Do not set jump when len(signal) < num_samples | Daniel Friesel | -1/+2 | |
| 2020-11-02 | Simplify PELT usage. remove kneedle, refactor code | Daniel Friesel | -273/+67 | |
| 2020-10-09 | add sub-state generation | Daniel Friesel | -1/+133 | |
| 2020-10-08 | BA Janis import: working detection of optimal number of changepoints | Daniel Friesel | -0/+261 | |
 |
index : dfatool | |
| Performance Model Generation for Configurable System Components | derf |
| summaryrefslogtreecommitdiff |