Subject: Re: FreeBSD Bus DMA
To: None <,,,>
From: Jonathan Stone <jonathan@DSG.Stanford.EDU>
List: tech-kern
Date: 06/12/1998 00:26:06
Jonathan Stone said:
> Amdahl's Law laves plenty of room for skepticism. Whatever it is that
> FreeBSD is doing to get a 3X speedup on compiles, it must be taking
> 75% of the total wall time on NetBSD.
> That's a staggering difference for a kernel change which, allegedly,
> doesn't involve heavy paging.  What kind of application (which isn't
> thrashing the VM system) is spending 75% of its time in something
> fixed by a merged block/buffer cache?
>It is actually reasonable considering the total dearth of
>disk seeks when using FreeBSD as the build engine.

No, I dont think it is.  Not if that number misleadingly excludes
other relevant costs _necessary_ to obtain the quoted numbers.  I find
it very difficult to imagine any good kernel hacker with a background
in performance issues failing to understand that.

To put back the text which John Dyson snipped without marking
and without answering (thereby tacitly signyfing agreement?)

>From the description, it sound suspiciuosly like maybe is doing a
>"make depend", and then timing the results of a "make build", and
>reporting the difference in times for only the "make build" stage.
          [[NB: I should have explicitly said 
	  Thus excluding the time `make depend'" takes loading 
	  everything into the cache." ]]
>That would simply not be accepted in, for example, a refereed publication.

In case it wasn't clear, I'm saying that _if_ this is what is being
done, then claiming 2.5x-3x _without_ also saying clearly that your
numbers excluded the necessary cache-load time, is dishonest.  If
someone from NetBSD did the same thing on a FreeBSD list, I'm sure
John and others would call them for it.  Rightly so, too.

But I'm not sure that's what is going on here.  Its hard to tell while
John is being so coy on the question and with  the details of the test
in general.