Prev: Missing SA entry with racoon setup.
Next: Prevent loading items from /boot/loader.conf
From: Michael Sierchio on 19 Jun 2008 13:39
jpd wrote:

>> The assertion of 1/10th the throughput is absurd, and gives reason
>> enough to refute the assertion.
>
> Well, don't forget to do so. While at it, provide some numbers.

Distinguish throughput from transfer rate, among other
conceptual errors in the assertion. At some point, esp.
for large transfers, spindle speed is the limiting factor --
which is gives a maximum data rate far lower than the
theoretical transfer rate.

DMA is not "zero copy" either -- throughput implies application
throughput, which requires copying buffers to user space.


From: Martin Etteldorf on 19 Jun 2008 17:32
Yep <yep(a)yep> wrote:
> Martin Etteldorf wrote:
>> Yep <yep(a)yep> wrote:
>>> Martin Etteldorf wrote:
>>>> What about sysctl hw.ata.atapi_dma?
>>>> I assume it's set to "1". If that's the case, try setting it to "0".
>>> Wouldnt this make drive access nice and slow, PIO data transfers require
>>> a significant amount of CPU time too and give about 1/10th of the
>>> throughput?
>>
>> No, it wouldn't.
>
> So lets break this down, disabling DMA means that PIO data transfers are
> used right?

Yes.

> So, you are disagreeing that PIO is slower that DMA mode
> transfers?

No, but I do disagree with the numbers you gave and with your
statement about "significant amount of CPU time".



Martin
--
"For the Snark's a peculiar creature, that won't
Be caught in a commonplace way.
Do all that you know, and try all that you don't;
Not a chance must be wasted to-day!"
From: Ted Nolan <tednolan> on 19 Jun 2008 18:57
In article <g3ej8t$fma$1(a)kadath.gruftie.net>,
Martin Etteldorf <etteldor(a)email.lu> wrote:
>
>
>Yep <yep(a)yep> wrote:
>> Martin Etteldorf wrote:
>>> Yep <yep(a)yep> wrote:
>>>> Martin Etteldorf wrote:
>>>>> What about sysctl hw.ata.atapi_dma?
>>>>> I assume it's set to "1". If that's the case, try setting it to "0".
>>>> Wouldnt this make drive access nice and slow, PIO data transfers require
>>>> a significant amount of CPU time too and give about 1/10th of the
>>>> throughput?
>>>
>>> No, it wouldn't.
>>
>> So lets break this down, disabling DMA means that PIO data transfers are
>> used right?
>
>Yes.
>
>> So, you are disagreeing that PIO is slower that DMA mode
>> transfers?
>
>No, but I do disagree with the numbers you gave and with your
>statement about "significant amount of CPU time".
>
>
>
> Martin

Does it even matter? If that's what it takes to make it work, you either
accept the speed hit or switch OSes.

Ted
--
------
columbiaclosings.com
What's not in Columbia anymore..
From: Steve O'Hara-Smith on 19 Jun 2008 17:46
On Thu, 19 Jun 2008 10:39:32 -0700
Michael Sierchio <kudzu-usenet2(a)tenebras.com> wrote:

> jpd wrote:
>
> >> The assertion of 1/10th the throughput is absurd, and gives reason
> >> enough to refute the assertion.
> >
> > Well, don't forget to do so. While at it, provide some numbers.
>
> Distinguish throughput from transfer rate, among other
> conceptual errors in the assertion. At some point, esp.
> for large transfers, spindle speed is the limiting factor --
> which is gives a maximum data rate far lower than the
> theoretical transfer rate.

Hmm lessee now DVD-ROM at 16x rate is somewhere around 20MB/s
whereas it's hard to get above about 8MB/s with PIO. Dropping to PIO will
reduce throughput to any fast drive (by a factor of 2-3 not 10 for optical
drives) but what is worse is that at maximum throughput it will use all
available CPU in interrupt response.

--
C:>WIN | Directable Mirror Arrays
The computer obeys and wins. | A better way to focus the sun
You lose and Bill collects. | licences available see
| http://www.sohara.org/
From: jpd on 20 Jun 2008 01:49
On Thu, 19 Jun 2008 22:57:34 GMT,
Ted Nolan <tednolan> <ted(a)loft.tnolan.com> wrote:
>
> Does it even matter? If that's what it takes to make it work, you either
> accept the speed hit or switch OSes.

Or fix the issue, or assist in resolving it: file a PR.


--
j p d (at) d s b (dot) t u d e l f t (dot) n l .
This message was originally posted on Usenet in plain text.
Any other representation, additions, or changes do not have my
consent and may be a violation of international copyright law.
First  |  Prev  |  Next  |  Last
Pages: 1 2 3 4
Prev: Missing SA entry with racoon setup.
Next: Prevent loading items from /boot/loader.conf