home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!sun-barr!ames!claris!szebra!David.Fox
- From: David.Fox@Saigon.COM
- Newsgroups: comp.unix.bsd
- Subject: Math emulation quite slow on 386bsd
- Keywords: emulation, 387
- Message-ID: <1992Sep11.232514.12064@Saigon.COM>
- Date: 11 Sep 92 23:25:14 GMT
- Distribution: usa
- Organization: Station Zebra Corp, Sunnyvale CA Ph:408-739-1520
- Lines: 45
-
- Hi.
-
- I've noticed that the math emulation seems extremely slow compared to
- DOS with 386BSD. I suspected this for a while, but didn't have any
- real checks to test this. I just pulled off source code to the dhry-
- stone and whetstone benchmarks, and got the following results:
-
- With 386bsd, no tasks other than the standard ones running (i.e., single
- user):
-
- whetstones = 1903
- dhrystones (no register) = 3092
- dhrystones (register) = 3099
-
- QAPLUS under DOS 5.0:
-
- whetstones = 49.4K
- dhrystones = 3251
-
- QAPLUS under DOS 5.0, no QEMM/386:
-
- whetstones = 69.8K
- dhrystones = 3251
-
- This is rather interesting, especially the large jump (49.4K vs. 1903).
- Unless QAPLUS's idea of what a whetstone is is radically different than
- the whet.c program's idea, this shows a substantial decrease in emulator
- performance.
-
- What I got when I did 'time whet':
- 7.6u 525s 8:52.77 100.0% 0+0K 0+0io 0pf+0w (whatever this means) :)_
-
- I can't really interpret this but it looks like it took roughly 525 seconds
- to run it. Whet.c includes a define that says it's going to run for
- a million whetstones. Ergo, 1,000,000 / 525 gives approximately 1,903.
-
- Unfortunately, at this time, my dos partition is small, and rather flaky, so
- I wasn't able to restore a C compiler off of tape and run whet.c under DOS.
-
- If anyone has whet.c whetstone benchmark and has MSC or TC please run it
- and see if there is such a substantial difference between the two math
- emulators, and email me about it.
-
- If I can get the DOS partition to be less flaky (maybe a new HD?) I'll
- post a followup article on this point.
-