-
Notifications
You must be signed in to change notification settings - Fork 2.6k
Conversation
The FRAME benchmarking does not handle the case well where a linear component as only a range of [0, =1] since in this case it has to calculate a linear formula from just two points which is unstable. Now bumping these upper limits back API_BENCHMARK_BATCHES should improve stability. This partially reverts 2eb8ad2 Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>
These also show up as asymptotically unstable. There are some more, but let's see first if we go into the right direction. Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>
This reverts commit e023626.
bot clean |
@ggwpez https://gitlab.parity.io/parity/mirrors/substrate/-/jobs/2508367 was started for your command Comment |
@ggwpez https://gitlab.parity.io/parity/mirrors/substrate/-/jobs/2508368 was started for your command Comment |
@ggwpez Command |
@ggwpez Command |
Looks much better now (SWC link). Note that it is using the On |
Closing in favour of #13595 |
The FRAME benchmarking does not handle the case well where a linear component as only a range of [0, =1] since in this case it has to calculate a linear formula from just two points which is unstable.
Now bumping these upper limits back
API_BENCHMARK_BATCHES
should improve stability. There are some more benches which could profit from this, but let's first see if we go into the right direction by re-benching.This partially reverts #12526