rich@pro-exchange.cts.com (Rich Sims) (04/14/89)
Since there's been some discussion about the TransWarp GS, I thought I'd pass on the results of some (admittedly, very un-scientific) performance testing I did. I'm sure all the neat benchmarks in the various magazines are very important and accurate, but I don't often have a need to make my computer do a lot of NOP's or compute prime numbers a few thousand times, so I tried something more in keeping with what I do use it for. I created a text file of about 408k, and did a "global search and replace" on it with several different editors to get a comparison. The test was to replace the word "destruction" with the word "construction" wherever it appeared. There were 67 replacements carried out in each test, since the test file was just 67 concatenated copies of a shorter file. The computers I used were an Apple IIgs with 1.75 megs of RAM, a Macintosh Plus with 1 meg of RAM, and an IBM PS/2-50 with 1 meg of RAM (but in true IBM fashion, only 640k of that was useable). When the load times were taken, the GS was reading from a CMS 60SD SCSI drive, the Mac from a DataFrame XP-20 SCSI drive, and the IBM from a 3.5" floppy in the internal drive. Here's the results: (the '*' indicates the 'TransWarped' results) Computer Speed Editor Load Replace ------------------------------------------------------------------------- Apple IIgs 1.0 mHz AppleWorks 2.0 134 sec. 92 sec. Apple IIgs 2.6 mHz AppleWorks 2.0 56 sec. 66 sec. *Apple IIgs 7.0 mHz AppleWorks 2.0 25 sec. 35 sec. Apple IIgs 2.6 mHz MicroEMACS 3.9 146 sec. 165 sec. *Apple IIgs 7.0 mHz MicroEMACS 3.9 53 sec. 76 sec. Macintosh+ 8.0 mHz QUED/M 2.07 14 sec. 16 sec. IBM PS/2-50 10.0 mHz QEdit 2.07 23 sec. 1.5 sec. The choice of editors (aside from that's what I have) was based on the fact that they all work on RAM-based files. The file read by AppleWorks was in AWP format, the rest were plain ASCII files with no formatting codes. That added about 25k to the "base" file. (AppleWorks was also a bit slower when loading the original text file.) In all fairness, the IBM actually only did 66 replacements, since I kept the file lengths the same, and the line feeds in the IBM file replaced some of the text that was in the other files. I also tried it with Microsoft Word, v. 3.01, on the Macintosh. If anyone is really interested, I'll let you know how that turned out, as soon as it gets done! :-) Rich Sims UUCP: crash!pro-exchange!rich ARPA: crash!pro-exchange!rich@nosc.mil INET: rich@pro-exchange.cts.com