You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Ok, now that #23 is in, I guess it's time to start looking at those performance regressions. I have basically zero experience with profiling, so sorry if the rest of this is a bit incoherent!
To start tackling this, I just did a malloc analysis for this simple function on julia v1.6.5 and v1.7.1:
On v1.6.5, this was nice and allocation free, and on v1.7.1... not so much, so this is my attempt to start the hunt for what the cause might be. Going line by line, it looks like the following bits of code from KeywordCalls.jl are allocating in v1.7.1 now (shown by the arrows):
Below is the Coverage.jl report for each julia version and script that I used to generate the .mem files. Is this already looking like a wild goose chase, or do you think these are the places that we should start looking into?
Ok, now that #23 is in, I guess it's time to start looking at those performance regressions. I have basically zero experience with profiling, so sorry if the rest of this is a bit incoherent!
To start tackling this, I just did a malloc analysis for this simple function on julia v1.6.5 and v1.7.1:
On v1.6.5, this was nice and allocation free, and on v1.7.1... not so much, so this is my attempt to start the hunt for what the cause might be. Going line by line, it looks like the following bits of code from
KeywordCalls.jl
are allocating in v1.7.1 now (shown by the arrows):Below is the
Coverage.jl
report for each julia version and script that I used to generate the.mem
files. Is this already looking like a wild goose chase, or do you think these are the places that we should start looking into?v1.6.5 mallocs
v1.7.1 mallocs
malloc script
The text was updated successfully, but these errors were encountered: