Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 24 Apr 2007 18:20:27 -0400
From:      Bill Moran <wmoran@potentialtech.com>
To:        Ivan Voras <ivoras@fer.hr>
Cc:        freebsd-questions@freebsd.org
Subject:   Re: Memory >3.5GB not used?
Message-ID:  <20070424182027.33d16b28.wmoran@potentialtech.com>
In-Reply-To: <f0lrup$18a$1@sea.gmane.org>
References:  <01d301c78699$d6a36820$0300020a@mickey> <20070424140528.95287ff4.wmoran@potentialtech.com> <021201c7869f$ee90fd70$0300020a@mickey> <3ee9ca710704241144n4ab349c6m901586e427b1ae0d@mail.gmail.com> <021c01c786a0$fe7e5510$0300020a@mickey> <20070424145433.734761db.wmoran@potentialtech.com> <f0lrup$18a$1@sea.gmane.org>

next in thread | previous in thread | raw e-mail | index | archive | help
In response to Ivan Voras <ivoras@fer.hr>:

> Bill Moran wrote:
> > In response to "Don O'Neil" <lists@lizardhill.com>:
> > 
> >> I never had this problem before when I built the kernel the first time.
> >> Could my module source be corrupt? If so, how do I re-install just the
> >> kernel sources for 6.1?
> > 
> > Not all modules work with PAE.  Read the example PAE kernel file for
> > information.
> > 
> > PAE is an awful hack, BTW.  I've heard a number of people complain that
> > performance sucks under PAE.
> 
> It greatly depends on the workload. For example, these are my results
> with unixbench:
> 
> 
> PAE:
>                      INDEX VALUES
> TEST                                        BASELINE     RESULT      INDEX
> 
> Dhrystone 2 using register variables        116700.0  6404191.9      548.8
> Double-Precision Whetstone                      55.0     1444.6      262.7
> Execl Throughput                                43.0     2374.5      552.2
> File Copy 1024 bufsize 2000 maxblocks         3960.0    47618.0      120.2
> File Copy 256 bufsize 500 maxblocks           1655.0    41809.0      252.6
> File Copy 4096 bufsize 8000 maxblocks         5800.0    58002.0      100.0
> Pipe Throughput                              12440.0  1018477.4      818.7
> Pipe-based Context Switching                  4000.0    32811.6       82.0
> Process Creation                               126.0     4491.9      356.5
> Shell Scripts (8 concurrent)                     6.0      638.0     1063.3
> System Call Overhead                         15000.0   798137.5      532.1
>                                                                  =========
>      FINAL SCORE                                                     317.2
> 
> 
> 
> NO PAE:
>                      INDEX VALUES
> TEST                                        BASELINE     RESULT      INDEX
> 
> Dhrystone 2 using register variables        116700.0  6673515.4      571.9
> Double-Precision Whetstone                      55.0     1475.1      268.2
> Execl Throughput                                43.0     2335.9      543.2
> File Copy 1024 bufsize 2000 maxblocks         3960.0    43796.0      110.6
> File Copy 256 bufsize 500 maxblocks           1655.0    39474.0      238.5
> File Copy 4096 bufsize 8000 maxblocks         5800.0    57819.0       99.7
> Pipe Throughput                              12440.0   998089.5      802.3
> Pipe-based Context Switching                  4000.0    25928.4       64.8
> Process Creation                               126.0     5043.9      400.3
> Shell Scripts (8 concurrent)                     6.0      697.0     1161.7
> System Call Overhead                         15000.0   792628.3      528.4
>                                                                  =========
>      FINAL SCORE                                                     312.7
> 
> The final score is better in PAE case because IO performance measured
> better, but in this case I know this particular benchmark can be
> ignored, but the rest of the numbers should be fine.
> 
> In short, PAE is worse, but not horribly so.

Does this test demonstrate usage of memory over 4G?  It's my understanding
that PAE starts to suffer when it has to look at the memory over 4G (which
is the problem it's intended to solve)

If your entire test fits in under 4G, you're not seeing the worst of it.
At least, that's my understanding of the issue.

-- 
Bill Moran
http://www.potentialtech.com



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20070424182027.33d16b28.wmoran>