home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!auspex-gw!guy
- From: guy@Auspex.COM (Guy Harris)
- Newsgroups: comp.arch
- Subject: Re: BCD Addition
- Message-ID: <15395@auspex-gw.auspex.com>
- Date: 8 Nov 92 23:51:51 GMT
- References: <32580141@hpcuhe.cup.hp.com> <1992Nov4.151413.4203@odin.diku.dk> <1992Nov07.175000.14703@ksmith.uucp>
- Sender: news@auspex-gw.auspex.com
- Organization: Auspex Systems, Santa Clara
- Lines: 43
- Nntp-Posting-Host: auspex.auspex.com
-
- >>To be more precise: PA-RISC maintains carry bits for each BCD digit,
- >>which allows BCD addition/subtraction using a decimal correct
- >>instruction after each normal ADD/SUB. (It has IDCOR to re-bias
- >>intermediate results, and DCOR to un-bias final results.)
-
- ...
-
- >Which probably explains why HP is doing well in the business markets.
-
- It might, or it might not. It depends on how much it speeds BCD
- arithmetic up.
-
- There may be other explanations (I assume here that HP-PA machines are,
- in fact, doing better in business markets than other RISC machines; I'd
- have no problem believing it's true, but I have no evidence for or
- against that claim).
-
- For example, other vendors of RISC machines may simply not be aiming
- very hard at that market; IBM may or may not want to have RS/6000's
- competing with 3[679]0's and S/38s or AS/400's, and DEC may have
- considered DECstations to be for technical markets and VAXes for
- commercial markets.
-
- Sun hasn't any non-SPARC machines with which SPARC machines might
- compete, but Sun hasn't been in the business market as long as HP.
-
- >I find it amusing that probably WAY more than half of the CPU cycles
- >expended on this globe are probably for some business purpose,
-
- Are there any benchmarks for commercial code (e.g., the US Steel COBOL
- benchmarks?) that are considered to give numbers that indicate how well
- a machine will execute commercial code and, if so:
-
- how well do HP-PA machines do on them relative to
- SPARC/POWER/MIPS/Alpha/etc. machines?
-
- how does the relative performance of the PA machines relative to
- the others compare to their relative performance on other
- benchmarks and applications? (I.e., a 100 MHz PA7100 machine
- will probably outdo a 25 MHz 7C601 SPARC machine or a 25 MHz
- R2000 machine on COBOL benchmarks, but it'll probably out do
- them when running compilers, engineering applications, etc. as
- well.)
-