home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.programming
- Path: sparky!uunet!psgrain!neon!nestmoi!jim
- From: jim@nestmoi.rain.com (Jim Chorn)
- Subject: Re: Semaphores, Swap, or Test_And_Set
- Message-ID: <1992Dec24.065103.7263@nestmoi.rain.com>
- Followup-To: comp.arch
- Lines: 1860
- Organization: none - Portland, OR
- References: <1992Dec22.055727.8056@nestmoi.rain.com> <1h7htbINNdg6@usenet.INS.CWRU.Edu> <1992Dec22.192229.17569@pasteur.Berkeley.EDU> <1h87slINNshf@rodan.UU.NET>
- Date: Thu, 24 Dec 1992 06:51:03 GMT
-
- In Article <1h87slINNshf@rodan.UU.NET> Vadim Antonov writes:
- -> Never mind DMA. So far i saw no x86 machines (even real
- -> multiprocessors like Sequent i currently enjoy looking at
- -> through my office door) which have circuitry protecting
- -> DMA from colliding with locked memory locations.
- Like I stated before, It depends on the hardware.
- The Sequent cache has bus watching circuitry which causes DMA's to
- invalidate cache lines on each transfer. Have you removed it? ;-)
-
- -> LOCK prefix is useful only for multiprocessing, period.
- Again, it depends on the hardware.
-
- -> 386 isn't really good for multiprocessing (no per-page
- -> cacheability control), it was amended in 486 (but they
- -> added non-coherent internal cache to keep life funnier,
- -> sigh).
- Again, (sigh) it depends on the hardware. Per page cache controls
- make multiprocessor software easier for some cases but this can
- also be accomplished in the cache controller. As for the 386 not
- being good, then why did Sequent select the 386 for a multiprocessor
- system? ;-)
-
- With the advent of today's ASIC hardware technology, almost *any*
- general purpose cpu can be used in a multiprocessor system. How
- fast the system really is depends on several factors including bus,
- cache and memory bandwidth as well as processor topology and what
- application the system design is targeted to solve. Each of these
- areas are complex and should be discussed in other news groups.
-
- I've put follow-ups to comp.arch for which I have no news feed
- so please email me if you wish to continue this discussion.
-
- -jim jim@nestmoi.rain.com
-