home *** CD-ROM | disk | FTP | other *** search
- Path: EU.net!sun4nl!xs4all!usenet
- From: jtv@xs4all.nl (Jeroen T. Vermeulen)
- Newsgroups: comp.sys.amiga.advocacy,comp.sys.amiga.misc
- Subject: Re: New Press Release!
- Date: Wed, 27 Mar 96 03:31:03
- Organization: Leiden University, Mathematics & Computer Science, The Netherlands
- Message-ID: <19960327.7B3D420.3236@asd07-09.dial.xs4all.nl>
- References: <2937.6638T1404T1877@mozart.inet.co.th> <314B536E.5B1D@infohwy.com> <4im6oj$h9u@soleil.uvsq.fr> <4into7$136@nyx.cs.du.edu> <4j1mr0$dsb@valour.pem.cam.ac.uk>
- NNTP-Posting-Host: asd07-09.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 <4j1mr0$dsb@valour.pem.cam.ac.uk> cbrown@armltd.co.uk (Chris Brown) writes:
-
- > Remember that Spec is just as much a test of harddisk,
- > system bus, etc. as it is a test of CPU power.
-
- If I may put in my two cents': The influence hard disk performance is avoided
- as much as possible. That is why systems are tested with 64Mb of memory--they
- will swap otherwise. Not only would that adversely affect measured performance;
- it will also invalidate the benchmark in the eyes of the consortium.
-
- A very important factor you omit however is the compiler; companies like Intel
- put enormous amounts of work into achieving SPEC results that no real-world
- application would ever get without intolerable compilation times. They actually
- consider high SPEC results as the primary reason for developing their compiler
- and even then complain about the agonizing amount of time it takes to optimize
- its code.
-
- "Real-world" compilers do better work for the 68k than they do for the x86. The
- messier the architecture, the harder it is to generate good code for. PC
- advocates even used to say that C benchmarks aren't fair because their compilers
- are so lousy. That is why the Intel compiler contains optimizations that are so
- specific that their only practical use is to improve SPEC benchmark results.
-
- The recent Intel compiler bug was an extreme example: It gave a strong boost to
- SPECint92 performance, but produced code that only _happened_ to give the right
- results for the input used in the benchmark. Any "normal" testing or use of the
- same code by Intel should have revealed the bug.
-
-
- > /* _ */main(int k,char**n){char*i=k&1?"+L*;99,RU[,RUo+BeKAA+BECACJ+CAACA"
- > /* / ` */"CD+LBCACJ*":1[n],j,l=!k,m;do for(m=*i-48,j=l?m/k:m%k;m>>7?k=1<<m+
- > /* | */8,!l&&puts(&l)**&l:j--;printf(" \0_/"+l));while((l^=3)||l[++i]);}
- > /* \_,hris Brown -- All opinions expressed are probably wrong. */
-
- --
- ============================================================================
- # 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? #
-