gh-117151: optimize BufferedWriter(), do not buffer writes that are the buffer size - #118037
Conversation
Sorry, something went wrong.
… are equal to the buffer size. avoid extra memory copy. BufferedWriter() was buffering calls that are the exact same size as the buffer. it's a very common case to read/write in blocks of the exact buffer size. it's pointless to copy a full buffer, it's costing extra memory copy and the full buffer will have to be written in the next call anyway.
|
All commit authors signed the Contributor License Agreement. |
Sorry, something went wrong.
|
Most changes to Python require a NEWS entry. Add one using the blurb_it web app or the blurb command-line tool. If this change has little impact on Python users, wait for a maintainer to apply the skip news label instead. |
Sorry, something went wrong.
|
Regarding the CLA, the CLA was signed by my organization yesterday. |
Sorry, something went wrong.
serhiy-storchaka
left a comment
There was a problem hiding this comment.
LGTM.
It can potentially affect the performance of BufferedRandom in rare cases, but I think that they are too rare, and it worst cases it only affects performance, not the behavior. If it affects behavior, then there are other existing bugs in the code.
Sorry, something went wrong.
|
How much difference in speed is caused by this PR? If it is more than say 10%, it makes sense to add a short NEWS entry. |
Sorry, something went wrong.
|
It's single percent difference, it might not be news worthy. |
Sorry, something went wrong.
|
I've reconfirmed the CLA with both emails. is it okay now? |
Sorry, something went wrong.
|
Thank you. |
Sorry, something went wrong.
gh-117151: optimize BufferedWriter(), do not buffer writes that are the buffer size
Hello,
I am having a look at the buffering code as part of the linked discussion, #gh-117151
One thing that was puzzling me is finding the buffering code slightly slower than the non-buffering code, given the same conditions, so I started having a deeper look into it.
Could I get a second pair of eyes on this old piece of code?
BufferedWriter() is buffering calls that are the exact same size as the buffer.
It's a very common case to read/write in blocks of the exact buffer size.
I think this might actually be the primary use case, reading in chunks from disk/network and writing to disk.
I think it shouldn't buffer in that case. It's costing an extra memory copy and the buffer will be full and have to be written in the next call anyway.
Thoughts?