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
(Filing at the request of the JB folks for referencing)
I re-profiled our builds with KSP2 with 2.0.21-1.0.27. Unfortunately, across the board in a bunch of benchmarks I found that
abi change - KSP2 is ~40-50% slower
noabi change - KSP2 is 80-100% slower
clean build - negligible/amortized by the clean build
If I just benchmark expensive our two most expensive KSP tasks with `, KSP2 is around 20-30% slower.
Other notes:
We give the kotlin and gradle daemons the same memory, and these benchmarks were run on overpowered AWS boxes
None of these builds were experiencing memory pressure with the settings they had (somewhere in the dozens of GB for xmx)
time-to-task-start is more or less the same, assumes a hot configuration cache
Here's a few samples. Looking again there is some GC time creeping into some of these that I'll look closer into, but not enough to majorly skew the larger trend
The text was updated successfully, but these errors were encountered:
(Filing at the request of the JB folks for referencing)
I re-profiled our builds with KSP2 with 2.0.21-1.0.27. Unfortunately, across the board in a bunch of benchmarks I found that
If I just benchmark expensive our two most expensive KSP tasks with `, KSP2 is around 20-30% slower.
Other notes:
time-to-task-start is more or less the same, assumes a hot configuration cache
Here's a few samples. Looking again there is some GC time creeping into some of these that I'll look closer into, but not enough to majorly skew the larger trend
The text was updated successfully, but these errors were encountered: