Page MenuHomePhorge

regression in decryption performance
Open, NormalPublicBUG

Description

Compared to 17ms in https://iron.lily-is.land/T20

] KAZV_OUTPUT_LEVELS='client=info' ./src/tests/client--encryption-benchmark-test '[!benchmark]'
Filters: [!benchmark]
Randomness seeded to: 3325471601
Generating data

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
client--encryption-benchmark-test is a Catch2 v3.12.0 host application.
Run with -? for options

-------------------------------------------------------------------------------
tryDecryptEvents() benchmark: small
-------------------------------------------------------------------------------
/home/user/Code/libkazv/src/tests/client/encryption-benchmark-test.cpp:77
...............................................................................

benchmark name                       samples       iterations    est run time
                                     mean          low mean      high mean
                                     std dev       low std dev   high std dev
-------------------------------------------------------------------------------
tryDecryptEvents()                             100             1     39.6463 s
                                        398.315 ms    397.231 ms    399.626 ms
                                        6.05396 ms    5.13979 ms    7.16717 ms

Event Timeline

Switching to vodozemac base64 does not improve the benchmark. Which is expected because most of the time we are decrypting megolm events, not olm events.