aboutsummaryrefslogtreecommitdiff
Commit message (Collapse)AuthorAge
* More graph fine-tuningHEADmasterFranklin Wei2018-09-29
|
* Change graphs a little bitFranklin Wei2018-09-29
| | | | Labels are now modified in the text file (not code).
* Tweak graph styleFranklin Wei2018-09-20
| | | | Smaller size to fit ACM template; also moved key inside plots
* Test build/compose file tracking in test scriptsFranklin Wei2018-09-19
|
* Change graph layoutFranklin Wei2018-09-18
| | | | Now 2 pages of 2x2 each
* Update READMEFranklin Wei2018-09-18
|
* Change profiling breakdownFranklin Wei2018-09-18
| | | | | Coarser intervals across the board; also profiles F_verify() for both file and info retrievals.
* Further tweak graph styleFranklin Wei2018-09-02
|
* Complete client checking of module request fieldsFranklin Wei2018-09-02
| | | | | | The client now checks all fields of the TM request the server sends before signing it. This requires the client to know in advance the state of the container (by doing a info retrieve first).
* Change graph style to lines only, without fill, and fix generation bugFranklin Wei2018-08-04
| | | | | | | This also fixes a issue with the graph generation. We were summing all the columns, 2-n, which would include the confidence error and logleaves value. This would cause a consistent bias in the produced graph. Should be fixed now.
* Update README.mdFranklin Wei2018-08-03
|
* Adapt dummy service and client for new profiling methodFranklin Wei2018-08-03
| | | | | | This moves the profiling code to crypto.c (which should probably be renamed util.c at some point). Test scripts and data processing scripts have been updated as well to deal with the new data.
* Make y-axis title more accurateFranklin Wei2018-08-01
|
* Make test scripts take logleaves range and trial count on command lineFranklin Wei2018-07-31
| | | | This was long overdue.
* Finish fixing off-by-one errorFranklin Wei2018-07-31
|
* Change font to LiberationSerifFranklin Wei2018-07-31
| | | | | Apparently Times New Roman has the same misalignment issue. This one should actually work.
* Reverse nesting order of testing loopsFranklin Wei2018-07-31
| | | | We can now get progressively refined graphs, which is nice. :)
* Change graph font to Times New RomanFranklin Wei2018-07-31
| | | | Should fix the alignment issues seen with the BitstreamVera family.
* Fix off-by-one in test scriptsFranklin Wei2018-07-31
| | | | | This would cause slightly incorrect results because modify and retrieve would each do 501 operations, not 500 as the scripts expected.
* Improve graph generation and add timing code to sp_modifyfile()Franklin Wei2018-07-31
| | | | | The number of timing intervals is no longer hard-coded. Also added some profiling instrumentation to the modify function.
* Add README.mdFranklin Wei2018-07-31
| | | | | About time, huh? It's still incomplete and difficult to read, but it's a start.
* Improve graph generationFranklin Wei2018-07-31
| | | | | Graphs are now stacked and labeled according to the timing points in the source.
* Further instrument create code for profilingFranklin Wei2018-07-30
|
* Finish up data processing scripts for new profiling methodFranklin Wei2018-07-30
|
* WIP on processing data from new profileFranklin Wei2018-07-30
| | | | | tabulate.sh is partially updated; dummy hasn't been updated, though, and we still can't produce graphs.
* Add profile option using clock()Franklin Wei2018-07-29
| | | | | | This makes the service provider record the timestamps at certain points in each request, and return that information to the user. Still need to get the data processing scripts adapted.
* Update .gitignore to include various temporary/output filesFranklin Wei2018-07-28
|
* Add LOffice Draw diagramFranklin Wei2018-07-20
|
* Add scripts for testing with prepopulated databaseFranklin Wei2018-07-20
|
* Implement dummy prepopulation (with a separate script)Franklin Wei2018-07-19
|
* Add a compile-time option to do database prepopulation and exitFranklin Wei2018-07-18
| | | | | | Enabling the PREPOPULATE macro in service_provider.c will cause it to initialize a database with 2^logleaves - RUNS_TEST files, and exit. This allows fast testing on the resulting database at near full load.
* Work on saving prepopulated databasesFranklin Wei2018-07-17
|
* Add dot diagramFranklin Wei2018-07-17
|
* Change testing procedure, once againFranklin Wei2018-07-17
| | | | | Create is still done 2^logleaves times, but all other operations (modify, retrieve) are done a fixed number of times to speed testing.
* Allow tabulation of data from partially complete runsFranklin Wei2018-07-17
|
* Make test scripts test 5 runs of 4-20Franklin Wei2018-07-16
| | | | I really should make these a command-line argument.
* Add some commentsFranklin Wei2018-07-12
|
* Fix memory leak in find_empty_slot()Franklin Wei2018-07-12
| | | | Wasn't freeing statement before. Impact should be minimal.
* Make sp_createfile() allocate ACL IOMT in memory, rather than DBFranklin Wei2018-07-12
| | | | | sp_request() will copy it into the database anyway, so there's no point in doing twice the work.
* Comment out unused functionsFranklin Wei2018-07-12
| | | | Should silence some warnings.
* Make caller allocate memory in call to merkle_update()Franklin Wei2018-07-12
| | | | Hopefully this is slightly faster... should probably profile.
* Inline bintree_sibling() and bintree_parent()Franklin Wei2018-07-12
| | | | No point in incurring a function call overhead for such simple functions.
* Increase robustness of measurement processing scriptsFranklin Wei2018-07-11
| | | | | Makes graph.gnu executable to save some keystrokes, and allows producing graphs even with partially finished runs (good for quick feedback).
* Add omitted nonce in client's request_verinfo()Franklin Wei2018-07-11
| | | | | | I forgot this at first, so it always passed the same all-zero "nonce": hardly a good thing! Thankfully the performance impact should be minimal, so there's no need to re-run any tests.
* Handle measurements longer than 1 hour gracefullyFranklin Wei2018-07-11
|
* Improve dummy performanceFranklin Wei2018-07-11
| | | | | | The old code was creating a brand-new statement (in dynamic memory) on each operation, which slowed down over time. Replaced with persistent prepared statements.
* Add tests for IOMT search operations (disabled by default)Franklin Wei2018-07-11
| | | | Uncomment the run_tests() call in main.c to run the tests.
* Fix whitespace, indentation; minor changes to plotting scriptFranklin Wei2018-07-10
|
* Track build code/compose file as a hashFranklin Wei2018-07-10
|
* Produce graphs of dummy performance to compare with secure versionFranklin Wei2018-07-10
|
generated by cgit v1.1 at 2023-10-01 21:42:50 +0000