Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sdd.hp.com!zaphod.mps.ohio-state.edu!brutus.cs.uiuc.edu!psuvax1!rutgers!cbmvax!daveh From: daveh@cbmvax.commodore.com (Dave Haynie) Newsgroups: comp.sys.atari.st Subject: Re: TT speed figure Message-ID: <12659@cbmvax.commodore.com> Date: 13 Jun 90 20:09:56 GMT References: Reply-To: daveh@cbmvax (Dave Haynie) Organization: Commodore, West Chester, PA Lines: 49 In article cmm1@CUNIXA.CC.COLUMBIA.EDU (Christopher M Mauritz) writes: >l86@nikhef.nl (Hugo Burm) writes: >>The developers version of the TT (16 MHz) runs >>at 4100 Dhrystones >4100 Dhrystones? ACK! That is awful. You can get that kind of >performance from a vanilla 386 clone. Benchmarking is kind of a black art, and Dhrystone is one of the blackest forms of this art. First, saying "Dhrystone" is kind of meaningless in this era of optimizing compilers -- some compilers can produce far faster numbers with Dhrystone 1.1 vs. Dhrystone 2.x, on the same hardware. So first thing, you want to compare like with like. Then of course there's the compiler used. You can get a nearly 2:1 increase going from early C compilers to the latest Lattice with all the "go faster" flags set, on the same [Amiga] hardware. Then you get to the issue of "memory model". For 68030s, you should be running with 32 bit integers, just like RISC machines or VAXen or other 32 bit CPUs that show up in the comparisons. But you can get higher numbers in many cases choosing 16 bit integers. Benchmarks for Clones will just about always use 16 bit integers. Then consider the operating system. Most machines have an OS eating part of the CPU time during any benchmark. On an MS-DOS machine, you can throw out MS-DOS and use a '386 program loader to further boost the numbers. And on a '386 with cache, the entire Dhrystone program will fit in a 16K cache. All of which means that current '386 machines under MS-DOS tend to give pretty high Dhrystone numbers. It doesn't say too much about what such machines will do with actual programs, or even moreso, actual programs in a real operating system. So don't pay too much attention to Dhrystone; it's best at comparing Amigas to Amigas, Ataris to Ataris, etc. where the software is invarient. As move toward more and more dissimilar architectures, you often find unusual differences, especially with Dhrystone 1.1. >I'm not sure how much of this difference is because of the faster clock or >from the 68882 in the A3000. I wonder what kind of Dhrystone figure a Mac >IIcx would produce? Dhrystone has no floating point component. The '882 won't affect the results. Other than clock speed, the compiler will have a noticable difference. The caches, burst memory, etc. do too. >Chris Mauritz |Donde hay una cerveza -- Dave Haynie Commodore-Amiga (Amiga 3000) "The Crew That Never Rests" {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: hazy BIX: hazy "I have been given the freedom to do as I see fit" -REM