Re: MapViewOfFile v ReadFile experiments and results
- From: "Lawrence Rust" <nospam@xxxxxxxxxxxxxxxxxxxxxx>
- Date: Wed, 18 Jan 2006 19:34:23 +0100
<news@xxxxxxxxxxxxxxxxxx> wrote
[snip]
> > You might try disabling buffering. Read up on FILE_FLAG_NO_BUFFERING -
> > it places various constraints on how you call ReadFile. I don't know
> > whether it will actually speed things up though.
>
> Yes, I played with this option many moons ago, but only with ReadFile -
> it killed performance! However, I'm now wondering what it might do to
> MapViewOfFile. I'm guessing that in theory it should make no
> difference. Hmm - I might give that a go. If I do, I'll re-post the
> results.
Setting FILE_FLAG_NO_BUFFERING & FILE_FLAG_OVERLAPPED when creating the
mapping file give maximum performance for MapViewOfFile.
> I suspect, despite all this, that I'm at the limits of the speed of the
> disk subsystem. Not that that has ever stopped my trying. :)
Have you tried benchmarking the disk I/O using perfmon? 347 MB in 23.3s
(16.9 MB/s) is well short of current disk performance. Back in 1999 I was
seeing a steady 8 MB/s writing video to a Seagate Medallist disk on NT4.
-- Lawrence Rust, Software Systems, www.softsystem.co.uk
.
- Follow-Ups:
- References:
- MapViewOfFile v ReadFile experiments and results
- From: news
- Re: MapViewOfFile v ReadFile experiments and results
- From: Tom Widmer [VC++ MVP]
- Re: MapViewOfFile v ReadFile experiments and results
- From: news
- MapViewOfFile v ReadFile experiments and results
- Prev by Date: Re: Disabling Serial Ballpoint Mouse when GPS connected
- Next by Date: Re: MapViewOfFile v ReadFile experiments and results
- Previous by thread: Re: MapViewOfFile v ReadFile experiments and results
- Next by thread: Re: MapViewOfFile v ReadFile experiments and results
- Index(es):
Relevant Pages
|
