Dont use benchmark range on constant functions - #12456
Conversation
shawntabrizi
added
B0-silent
ggwpez
left a comment
ggwpez
left a comment
There was a problem hiding this comment.
Did you manually deploy this to the Substrate runtime to update the weights?
yes, did it manually, just to get a new file. The weight itself is not very important. |
We also never updated it. Could be that it never ran on reference hardware and therefore incorrectly detected a slope. |
I ran it locally, and it detected a very small slope of 1 ns. |
koute
left a comment
koute
left a comment
There was a problem hiding this comment.
Makes sense to me.
It's probably not worth it to try and detect these kinds of constant functions and special-case them during weight generation; such tiny slopes should not make much difference in practice anyway, and there's always the possibility of getting it wrong.
| // This is the benchmark setup phase. | ||
| let value = 1000u32.into(); |
There was a problem hiding this comment.
Since this is supposed to be an example maybe it'd be a good idea to explicitly say that "the set_dummy always takes constant time hence we're hardcoding the value here" or something like that?
|
bot merge |
|
Waiting for commit status. |
|
Merge cancelled due to error. Error: Statuses failed for 1600913 |
|
bot rebase |
…-example-benchmarking
|
Rebased |
|
bot merge |
* dont use benchmark range on constant function * update weights * fix * new weights * Update frame/examples/basic/src/benchmarking.rs Co-authored-by: parity-processbot <>
4 participants
Trying to address the confusion here: https://substrate.stackexchange.com/questions/5249/why-would-benchmarking-a-simple-setter-has-a-linear-behaviour
These are constant functions, and should have constant weight.