129374 - [perf] we should cached the read buffer in nsStreamConverter and QuotingOutputStreamListener

Status: RESOLVED WONTFIX

(MailNews Core :: Composition, enhancement)

Description

[Jean-Francois Ducarroz]

everytime we process a block in data during the quoting part of reply, we
allocate a buffer, and then destroy it. I would be more efficient to try to
reuse the same buffer.

Comment 1

[Jean-Francois Ducarroz]

same problem occurs in mime nsStreamConverter::OnDataAvailable. Would be very
nice if we could fix this two case and check if we don't have more...

Comment 2

[scottputterman]

How big is a block?  What kind of performance improvement would you expect to
see in reply and message display?

Comment 3

[Jean-Francois Ducarroz]

I need to figure out the max size of a block but it should be somewhere between
4K and 32K. The speed improvement will be noticable only on low-end
configuration (slow CPU/low memory) with large message. That will affect message
display, attachment saving and reply (twice for reply). I don't have numbers but
the fix is very simple to do.

Comment 4

[Michael Buckland]

Discussed in Mail News bug mtg w Engineering Mktng PjM.  Decided to plus the
bug.  Making P2 also.

Comment 5

[Michael Buckland]

Discussed in Mail News with Mktng, QA, Engineering and PjM.  Decided to minus
this bug.

Comment 6

[Wayne Mery (:wsmwk)]

(In reply to comment #3)
> I need to figure out the max size of a block but it should be somewhere between
> 4K and 32K. The speed improvement will be noticable only on low-end
> configuration (slow CPU/low memory) with large message. That will affect message
> display, attachment saving and reply (twice for reply). I don't have numbers but
> the fix is very simple to do.

bienvenu feels this isn't worth keeping. If someone feels differently, please reopen (as ENH) with details - or a patch! :)