Message from discussion
The fastest 68k Amiga ever....
Path: supernews.google.com!sn-xit-02!supernews.com!nntp-relay.ihug.net!ihug.co.nz!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!news.mel.connect.com.au!news1.optus.net.au!optus!news0.optus.net.au!wombat.cs.monash.edu.au!not-for-mail
From: bme...@cs.monash.edu.au
Newsgroups: comp.sys.amiga.advocacy
Subject: The fastest 68k Amiga ever....
Organization: This is innd taking over...
Lines: 93
Sender: bme...@bruce.cs.monash.edu.au
Message-ID: <8nbkst$ta9$1@wombat.cs.monash.edu.au>
Reply-To: bme...@cs.monash.edu.au
Date: 16 Aug 2000 00:42:37 +1000
NNTP-Posting-Host: 130.194.64.240
X-Trace: news0.optus.net.au 966351644 130.194.64.240 (Wed, 16 Aug 2000 01:00:44 EST)
NNTP-Posting-Date: Wed, 16 Aug 2000 01:00:44 EST
.. is a version of UAE, at least as far as integer instructions are
concerned.
It is with great pleasure that I can finally make the first Alpha release
of the Just In Time compiler for UAE/linux available. At
http://byron.csse.monash.edu.au/uae-JIT.tar.gz
you can download 2.2MB archive that contains two executables (one
targetting XF86 4.0 specifically, the other addressing all other versions
of X) as well as the sources they were built from.
If you install that stuff on a reasonably fast linux system, you get
something that puts most any 680x0 Amiga in existence to shame. Here are
a few benchmarks to spark your interest:
Test Celeron2/566@876MHz Celeron333@458 A3000/CSPPC
Trident 3DImage975 S3 Trio3D/2X 060-66
============================================================================
RC5 (latest 812kkeys/s 407kkeys/s ~200kkeys/s
client)
OGR (latest 497knodes/s 252knodes/s ????
client)
DoomAttack, 99.6 fps 58.6fps ~40fps
-timedemo demo3
SysSpeed ImageStudio Test (in s):
---------------------------------
* JPEG Load 0.24 0.48 0.38
* Scale 0.15 0.24 0.25
* BlurHigh 0.76 1.44 0.90
* Cross 0.43 0.80 0.49
* Diagonal 0.40 0.74 0.53
* Focus 0.32 0.58 0.43
* Shake 0.46 0.86 0.51
* Texture 0.54 1.00 0.61
* EmbossHigh 0.49 0.90 0.53
SysSpeed Cruncher Test (in s):
---------------------------------
* LhaCrunch 0.75 1.46 1.64
* LhaTest 0.15 0.29 0.17
* LhaDecrunch 0.17 0.30 0.24
* XPKCrunch 1.86 3.37 4.13
* XPKDecrunch 0.28 0.58 0.61
* PPCrunch 1.03 2.08 3.23
* PPDecrunch 0.08 0.16 0.16
WarpRace:
--------------------------------- (060-50)
* Pixelomania 0.04s 0.08s 0.08s
* ByteCopy 94.05M/s 62.00M/s 15.01M/s
* TurboCopy 122.01M/s 80.02M/s 34.02M/s
* C2P 1.60ms 3.05ms 10.33ms
* Cybermand 5.17s 10.25s 1.38s
Do I have your attention? Great! Now it's time for someone to test the
compiler on a wide variety of software. If you have some software to
test with, and have a linux machine handy, why not give it a spin?
The executables were built on a Linux RedHat 6.1 system, and link
against the following libraries:
libSM.so.6 => /usr/X11R6/lib/libSM.so.6 (0x4001a000)
libICE.so.6 => /usr/X11R6/lib/libICE.so.6 (0x40024000)
libXext.so.6 => /usr/X11R6/lib/libXext.so.6 (0x4003a000)
libX11.so.6 => /usr/X11R6/lib/libX11.so.6 (0x40047000)
libpthread.so.0 => /lib/libpthread.so.0 (0x40109000)
libgtk-1.2.so.0 => /usr/lib/libgtk-1.2.so.0 (0x4011a000)
libgdk-1.2.so.0 => /usr/lib/libgdk-1.2.so.0 (0x4023a000)
libgmodule-1.2.so.0 => /usr/lib/libgmodule-1.2.so.0 (0x4026d000)
libglib-1.2.so.0 => /usr/lib/libglib-1.2.so.0 (0x40270000)
libdl.so.2 => /lib/libdl.so.2 (0x40292000)
libm.so.6 => /lib/libm.so.6 (0x40295000)
libc.so.6 => /lib/libc.so.6 (0x402b1000)
/lib/ld-linux.so.2 => /lib/ld-linux.so.2 (0x40000000)
But if you lack any of those, you can always compile from the sources.
Looking forward to your feedback,
Bernie
P.S.: I posted this to csaa separately to keep the resulting flamewars
(wanna bet?) out of csam.
--
Two things are infinite: the universe and human stupidity; and I'm not
sure about the the universe
Albert Einstein