Path: csiph.com!x330-a1.tempe.blueboxinc.net!usenet.pasdenom.info!selfless.tophat.at!news.glorb.com!border3.nntp.dca.giganews.com!Xl.tags.giganews.com!border1.nntp.dca.giganews.com!nntp.giganews.com!local2.nntp.dca.giganews.com!news.giganews.com.POSTED!not-for-mail NNTP-Posting-Date: Thu, 12 May 2011 21:26:38 -0500 Message-ID: <4DCC96D2.40603@SPAM.comp-arch.net> Date: Thu, 12 May 2011 19:26:26 -0700 From: "Andy \"Krazy\" Glew" Reply-To: andy@SPAM.comp-arch.net Organization: comp-arch.net User-Agent: Mozilla/5.0 (Windows; U; Windows NT 6.0; en-US; rv:1.9.2.11) Gecko/20101013 Thunderbird/3.1.5 MIME-Version: 1.0 Newsgroups: comp.arch Subject: Re: Anyone take a serious look at partitioned register sets? References: <670793a0-7d16-4163-b717-e0d53dc82d54@glegroupsg2000goo.googlegroups.com> In-Reply-To: <670793a0-7d16-4163-b717-e0d53dc82d54@glegroupsg2000goo.googlegroups.com> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Lines: 31 X-Usenet-Provider: http://www.giganews.com X-Trace: sv3-G7a+2TnjWi+D/QYgiyhMriKNB/3vWwDyM9nZydHtGn8T12DSfIQYdBlEOC8ePy+GtAv44STI1PKNl3O!lCB5nhtvCdT9yddJ9vDudoYm0kuB+Nt4w4HjHDTCK/iRrS/dTiRwfn/L0P3FJA== X-Complaints-To: abuse@giganews.com X-DMCA-Notifications: http://www.giganews.com/info/dmca.html X-Abuse-and-DMCA-Info: Please be sure to forward a copy of ALL headers X-Abuse-and-DMCA-Info: Otherwise we will be unable to process your complaint properly X-Postfilter: 1.3.40 X-Original-Bytes: 2709 Xref: x330-a1.tempe.blueboxinc.net comp.arch:1515 On 5/12/2011 8:26 AM, Paul A. Clayton wrote: > On Thursday, May 12, 2011 3:43:34 AM UTC-4, Michael S wrote: > [snip] >> Sounds like you are talking about different things. Or, may be, you >> and Paul are talking about the same thing, but Brett was talking about >> something else. >> Duplicated integer register file in Ev6/Ev7 is just an implementation >> detail. Brett, on the other hand, was arguing in favor of software- >> visible partitioning, similar to TI C6000 family, but more far- >> fetched. > > I think Brett is looking at microarchitecture splitting as > an indication of the appropriateness of Architectural > splitting. I.e., if EV6-style clustering is so useful, > why not Architect the split and avoid some of the > overhead. The answer is: compatibility. Lifetime of the architecture. If your architecture is going to last many years, it may be built on many different microarchitectures. Some of which have split RF, some which don't. If your architecture is likely to only ever be used in microarchitectures that benefit from split RF, expose it. By the way, the TI DSP architecture has lasted a long time. For many years the DSP architectures were characterized by split X/Y registers. (By the way, I think the meaningful split noawadays is not integer versus FP, but scalar (64b) versus SIMD packed vector (128-256-512 bit).)