home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.arch
- Path: sparky!uunet!ferkel.ucsb.edu!taco!gatech!emory!swrinde!zaphod.mps.ohio-state.edu!usc!howland.reston.ans.net!bogus.sura.net!udel!rochester!cornell!uw-beaver!tera.com!bob
- From: bob@tera.com (Bob Alverson)
- Subject: Re: How many vector registers are useful?
- Message-ID: <1993Jan26.173220.25989@tera.com>
- Followup-To: comp.arch
- Keywords: architecture, question
- Sender: news@tera.com (News Administrator)
- Organization: Tera Computer Company, Seattle, WA
- References: <93-01-174@comp.compilers>
- Date: Tue, 26 Jan 1993 17:32:20 GMT
- Lines: 15
-
- In article <93-01-174@comp.compilers> kirchner@uklira.informatik.uni-kl.de (Reinhard Kirchner) writes:
- >
- >The Fujitsu machines with their reconfigurable register file of 32 or
- >64kb, which is 4k or 8k words, being grouped from 256 register a 16/32
- >words to 8 registers a 512/1024 words.
- >
- >Now there is the question: is such a large register file useful at all ?
-
- Besides serving as a software cache, a register file also serves as a
- prefetch area (something caches are doing now, too). So, you might want
- more registers as your memory latency increases. Of course, more registers
- only help if your processor can use them. For a vector machine, you might
- need longer vectors to cover longer memory latency.
-
- Bob
-