Re: WrMeta
Re: WrMeta
- Subject: Re: WrMeta
- From: Shantonu Sen <email@hidden>
- Date: Fri, 06 Jul 2012 14:30:11 -0700
It's a write of filesystem metadata. This is likely updates to the "atime" last accessed time for every file you enumerate, which causes writes to the HFS+ catalog (as well as journal when journaling is active).
You can try an experiment to see if this is the case with:
$ mount -t hfs | head -1
/dev/disk0s4 on / (hfs, local, journaled)
$ sudo mount -t hfs -u -o noatime /dev/disk0s4 /
$ mount -t hfs | head -1
/dev/disk0s4 on / (hfs, local, journaled, noatime)
$
Since this should be asynchronous, I'd be surprised if this is causing a significant performance degradation...
Shantonu
On Jul 6, 2012, at 2:22 PM, Dave Keck <email@hidden> wrote:
> Hey list,
>
> I'm optimizing a program that reads image metadata using
> CGImageSourceCreateWithURL(). This program should strictly read from
> disk, but as it iterates over thousands of photos, the kernel is
> writing data to disk at >1 MB/sec. fs_usage reports many instances of
> the following:
>
> WrMeta[asyncP] D=0x... B=0x2000 /dev/disk... 0.097343 W kernel_task
>
> This is not the kernel paging-out to disk, as Activity Monitor reports
> a constant value for "Page outs" for the lifetime of the program.
>
> What is 'WrMeta'?
> _______________________________________________
> Do not post admin requests to the list. They will be ignored.
> Darwin-dev mailing list (email@hidden)
> Help/Unsubscribe/Update your Subscription:
>
> This email sent to email@hidden
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Darwin-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden
| References: | |
| | >WrMeta (From: Dave Keck <email@hidden>) |