home *** CD-ROM | disk | FTP | other *** search
- Path: surfnet.nl!sun4nl!xs4all!usenet
- From: jtv@xs4all.nl (Jeroen T. Vermeulen)
- Newsgroups: comp.sys.amiga.misc
- Subject: Re: Speed: 68040 vs. 68060
- Date: Sun, 25 Feb 96 15:34:52
- Organization: Leiden University, Mathematics & Computer Science, The Netherlands
- Distribution: world
- Message-ID: <19960225.7AF9790.E534@asd10-22.dial.xs4all.nl>
- References: <4foi00$60t@gondor.sdsu.edu> <3125E74D.3390@gih.no> <19960223.425E10.10CBD@an100.du.pipex.com>
- NNTP-Posting-Host: asd10-22.dial.xs4all.nl
- Mime-Version: 1.0
- Content-Type: text/plain; charset=iso-8859-1
- Content-Transfer-Encoding: 8bit
- X-NewsSoftware: GRn 2.1 Feb 19, 1994
-
- In article <19960223.425E10.10CBD@an100.du.pipex.com> m.hendry@dial.pipex.com (Mathew Hendry) writes:
- >
- > Try the BYTEMark tests on Aminet, which contain compiled algorithms designed
- > to simulate real world applications. They are scaled to performance - faster
- > CPUs are given bigger tasks. The results are benchmarked relative to a Dell
- > P90, which was quite humbling for my poor Amiga and its aging 40MHz '030 :(
- > (approx. 10% of a P90 for integer routines, BTW. Even slower for FP)
-
- On my own port I got results that said the Pentium is about equal to the 68040
- *per cycle*. This is without touching the timing code; I don't think it takes
- multitasking into account so the Amiga may get artificially low results because
- of this.
-
- As for FP performance, I didn't look through the source all that closely but it
- seemed to me that the FP tests happened to hammer mainly on the few FP
- instructions that aren't implemented on the 040 (and are trapped by SW
- emulation). Here too the Amiga could be getting results that can be said to be
- artificially low by a very large factor.
-
- Comparing compilers (SAS vs. gcc) gcc wins for FP (probably because it knows how
- to inline code for those emulated instructions) but SAS generally gets some
- performance wins out of integer code.
-
-
- > : If code is optimised for an '060 is has more chance of running 2 or more
- > : instructions at once then unoptimised code.
- >
- > Only when a compiler appears which produces code optimised for the '060 can
- > you run such tests. Contriving code in assembly language to take advantage of
- > new CPU features provides misleading results unless a current compiler (as
- > used in the REAL WORLD) optimises code in the same way. Since SAS has
- > apparently abandoned support for SAS C/C++ I don't see this happening anytime
- > soon. So, your only real world estimates at the moment should be based on
- > code compiled for the 68040.
-
- At the introduction of the 060, Motorola disected a piece of SPECmark code (some
- important loop) as generated by existing compilers, and showed how it ran at a
- consistent 0.7 cycles per instruction.
-
- Of course they wouldn't exactly be showing the worst case here, but OTOH an
- optimizer for the new chip could conceivably still improve on that loop (and
- certainly in the general case).
-
- Anyway, I do think gcc has a scheduler.
-
-
- > BTW, as far as I know the current AmigaOS itself doesn't recognise an '060.
- > The CPU flags in execbase.h only extend up to the 68040. They also show a
- > possible lack of forethought -
- >
- > /* Processors and Co-processors: */
- > #define AFB_68010 0 /* also set for 68020 */
- > #define AFB_68020 1 /* also set for 68030 */
- > #define AFB_68030 2 /* also set for 68040 */
- > #define AFB_68040 3
- > #define AFB_68881 4 /* also set for 68882 */
- > #define AFB_68882 5
- > #define AFB_FPU40 6 /* Set if 68040 FPU */
- >
- > Surely they should have left a gap after AFB_68040 to allow for new additions
- > to the 680x0 line ;)
-
- Don't forget though...
-
- /* #define AFB_RESERVED8 8 */
- /* #define AFB_RESERVED9 9 */
-
- Lack of forethought? AFB_RESERVED8 must be for the 68060 and AFB_RESERVED9 for
- the POWERAmiga. Note how they predicted the *exact* transition point.
- ;-)
-
-
- > -- Mat.
-
- --
- ============================================================================
- # Jeroen T. Vermeulen \"How are we doing kid?"/ Yes, we use Amigas. #
- #--- jtv@xs4all.nl ---\"Oh, same as always."/-- ... --#
- #jvermeul@wi.leidenuniv.nl \ "That bad, huh?" / Got a problem with that? #
-