encoding/json: remove comment about performance of scanner.step #37386
Conversation
In an experiment I could not reproduce the 10% performance loss when using an integer constant for scanner.step. Instead the performance seemed to be about the same or slightly better than before.
googlebot
added
the
cla: yes
|
This PR (HEAD: fc6a49f) has been imported to Gerrit for code review. Please visit https://go-review.googlesource.com/c/go/+/220582 to see it. Tip: You can toggle comments from me using the |
|
Message from Ian Lance Taylor: Patch Set 1: Please update the CL description to describe the machine you tested on. Thanks. Please don’t reply on this GitHub thread. Visit golang.org/cl/220582. |
|
Message from Richard Ulmer: Patch Set 2:
Done. Let me know if you need more info. Please don’t reply on this GitHub thread. Visit golang.org/cl/220582. |
|
Message from Richard Ulmer: Patch Set 2: I just learned that I have to use I feel a bit stupid now, but as you can see in the updated benchmark (see CL's description), the slowdown is just about 1% instead of 10% on my machine. Please don’t reply on this GitHub thread. Visit golang.org/cl/220582. |
In an experiment I could not reproduce the 10% performance loss when
using an integer constant for scanner.step. Instead the performance
seemed to be about the same.
The experiment can be seen here: codesoap@c7b685b
Less allocations are made in my experimental version, but the difference
is probably insignificant and depends on the used hardware. I'd be
interested to see the results on other systems.
I tested on a ThinkPad T450:
I know this PR is somewhat pedantic, but I think the comment could
potentially keep people from trying to optimize this piece of code and
thus hinder development.