DEC.BANKS@MARLBORO.DEC.COM (12/24/85)
From: Dawn Banks <DEC.BANKS@MARLBORO.DEC.COM> Mr Freed: Have your really ignored the half dozen postings to this list explaining that the poor drystone performace on the Amiga is more due to bad compiler output? I think the 1000 ballpark figure (from another compiler) is about what you'd expect in relation to the ST given the Amiga's slower clock and no explanation of cycle stealing is needed. Plea to the rest of the net: We've all seen dismal numbers for a drystone benchmark run with Lattice C for the Amiga. We've also seen more impressive numbers from a better compiler. Could we all please get off this stupid benchmark war? --------
freed@aum.UUCP (Erik Freed) (12/27/85)
> Mr Freed: > > Have your really ignored the half dozen postings to this list explaining > that the poor drystone performace on the Amiga is more due to bad compiler > output? I think the 1000 ballpark figure (from another compiler) is > about what you'd expect in relation to the ST given the Amiga's slower > clock and no explanation of cycle stealing is needed. > > Plea to the rest of the net: We've all seen dismal numbers for a drystone > benchmark run with Lattice C for the Amiga. We've also seen more impressive > numbers from a better compiler. Could we all please get off this stupid > benchmark war? > -------- First of all, this is pretty close to a flame and should have been a Email message, Second the timing of articles on the west coast can be quite a bit different than on the East coast. When I asked this question, there was, as far as I know, no explanation for the difference in benchmarks. There is, as far as I know, still nothing other than assumptions that the reason is the compiler, other than one helpful message that there is a presently unreleased compiler which puts to rest the discussion. I think that it is important to realize that this is a discussion not a war. If anyone could mail me the article that shows the Amiga equaling the ST in the drystone I would be very interested. I do not care whether the Amiga or the ST runs the benchmark faster. It is only interesting to figure out why they aren't relatively equal as they logically should be. And to the poster of this rebuke; I would appreciate a little more maturity in this whole matter. I read every article that I receive at my site regarding both the ST and Amiga. I honestly did not see an answer to my question *IF THERE WAS ONE*. I have no bone to pick with people who have decided to buy a Amiga. I simply like what the ST has to offer for my needs. I wouldn't throw either off my desk. -- ------------------------------------------------------------------------------- Erik James Freed Aurora Systems San Francisco, CA {dual,ptsfa}!aum!freed
DEC.BANKS@MARLBORO.DEC.COM (12/27/85)
From: Dawn Banks <DEC.BANKS@MARLBORO.DEC.COM> This message is a public apology for my rude behavior directed towards Mr. Freed. Anyone is free to ignore this message if they prefer to think of me as a rude person: You, of course are absolutely correct that I should not have posted that to the entire net (excepting the plea to get everyone off this subject without mentioning anyone in particular). On the other point, the drystone benchmark that provided the in excees of 1000 number has been discussed (and posted) several times over the last month. In fact, it was posted over a month ago shortly after the first person quoted the less appealling number which came from the (most appalling) Lattice C benchmark. It has also been uniformly ignored by anyone who doesn't want to hear that the Amiga can live up to its clock speed. Whether you are one of these individuals who chose to ignore the prior postings, or whether you entered the conversation late enough to have missed the other postings is unknown to me, and I appologize again for assuming wrongly if you are indeed in the latter case. The mail message was written in the "heat of the moment" when I'd been absolutely fed up with the hundreds of postings claiming that the Amiga is no good, holding out a single benchmark as proof. Believe me, if you've ever read code produced by the Lattice C compiler (as I have), you'd have no doubt that it is a major cause for the poor benchmark numbers we've seen. As straight machine speeds go, the only valid comparison for raw, usable CPU power would be arrived at if the benchmark had been coded in assembler, with the machine independent parts being identical on both machines. There is a valid discussion to be had, however, as to the usefulness of a machine whose sole compiler produces such horridly slow (and large) code. This is a reflection on the machine in that anything compiled using the Amiga's default C compiler would probably be slower than the same program compiled on one of the competition's machine, using the competition's compiler. While being a valid concern, it would have little or nothing to do with the actual hardware it's running on other than guilt by association with sleazy compilers. With any luck, someone will introduce a native compiler that corrects these problems. In the mean time, my personal preference is assembler anyway (which could get me started on what I think of the Amiga's assembler ...). Again, my appologies. D. Banks --------