Benchmark between make and ninja build for kvrocks #625
Replies: 3 comments 3 replies
-
@PragmaTwice I guess this can be a reply on #621 or since this issue has been created, we can move it to Discussions? It doesn't seem like an issue to resolve but a general topic about benchmark. |
Beta Was this translation helpful? Give feedback.
-
Oh yeah, I also think moving to Discussions is better. |
Beta Was this translation helpful? Give feedback.
-
That sounds good. We can introduce ninja build option as you describe and if developers find it a frequent better choice, switch it to default later. |
Beta Was this translation helpful? Give feedback.
-
Motivation
In #621, @tisonkun came up with a great idea: benchmark ninja build and make build of kvrocks.
I did this test after making some fixes to make cmake configuration of kvrocks support ninja.
A PR for that will follow.
Preparation
Clean Build
No-op Incremental Build
Incremental Build after modifying a source file
Incremental Build after modifying a header file
Conclusion
As you can see, ninja is slightly slower than make build in clean build, but has huge advantages in incremental build, which is very friendly to kvrocks developers who need to do frequent incremental builds.
I would like to make kvrocks use make by default, but have a
--ninja
option to enable ninja.Beta Was this translation helpful? Give feedback.
All reactions