Date: Sun, 24 Mar 1996 13:20:33 +0200 (EET) From: Narvi <narvi@haldjas.folklore.ee> To: Frank Durda IV <uhclem@nemesis.lonestar.org> Cc: hackers@freebsd.org Subject: Re: FreeBSD and MMX Message-ID: <Pine.BSF.3.91.960324131653.21822B-100000@haldjas.folklore.ee> In-Reply-To: <m0u0hRL-000CPxC@nemesis.lonestar.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Eat good food, preserve nature, be nice to all nice people :) On Sat, 23 Mar 1996, Frank Durda IV wrote: > Is anybody looking ahead six to nine months when the Pentium and > Pentium-Pro chips with the MMX extensions start arriving in machines? > > For those of you who haven't hward anything about MMX, here is a real > quickie summary: > > MMX (not an acronym) is a set of additional opcodes Intel is adding to > new Pentium and Pentium Pro chips later this year. In effect, there will > be Pentiums that are smarter than other Pentiums, and the same will be true > in the Pentium Pro line. This already has computer-maker marketing > people worrying about the possibility of ending up with the "dumber" > Pentium systems left in stores when on the shelf next to them are systems > with the new "smarter" Pentiums. Intel will apparently try to prevent some > of this apples-to-apples comparison by bringing out the MMX chips at higher > speeds than plain Pentiums (initially), and offering lower speed versions > of the MMX chips later on. > > The MMX processor adds several new 64-bit registers to the system > that are off in a corner, similar to the way the floating point > registers and opcodes are handled. There are new opcodes for gettings > things in and out of the new registers, plus new opcodes that perform > operations on the new registers. (Again, this is real similar to how > the floating point subsystem works.) According to Intels WWW FAQ, they > didn't have mess with the integer microcode much, thus lowering the risk > of a compatibility flaw of some sort. > > The new operations are all geared to things found in graphics, > compression, and other repetitive algorithms. For example, it > is possible to load eight 8-bit values into one of these new > registers and perform eight adds simultaneously, without having the Carry > bits roll over from one eight-bit value to the next. There are > also some codes to do max/min type functions to avoid signed rollover. > > There is a flag in the chip that lets software identify the presence of > the MMX subsystem, and (according to the WWW FAQ), if MMX isn't there > and an application uses those opcodes, "an Intel-provided .DLL will Too bad they only think about .dlls... How about machines which don't run windows (or doesn't do it all the time?) > emulate those opcodes". Yeah, if you are running Windows. > > - - - - - -End of summary > > > No doubt, eventually someone will want to use these opcodes for > something under FreeBSD, and the assember/compiler will get updated to > support them. > > However, it appears that the MMX subsystem interfaces very similarly > to the way the floating point operations are performed, including using > the same exception mechaniams. Since it seems that FreeBSD has always > kept floating point support on the back of the back burner with > regard to error handling, GDB support, big fixes, etc, these weak areas > may bite us when it comes to trying to handle MMX. > > I am no expert on the floating point situation (and really don't want to > be), and things may not be as bad as people on the mailing lists have > portrayed, but if things are broken or not completely implemented, we need > to find some people willing to dig into floating point support and get its > house in order so that when MMX becomes available, we will have most of > the pieces already working. > > We might even have to think about doing MMX emulation, much like > we do floating point emulation now. Ugh. > > > Comments? > > If you are interested in MMX, the MMX Programmers Manual is available > electronically at www.intel.com (under a non-obvious button on the > right side of the home screen - sorry I am at a place where I can't look > right now). It's in Acrobat PDF format, five or six chapters plus > four appendixes, but there is a Windows viewer available you can use to > print it out, or you can wait until Intel has the exact same material in > available printed form. You don't have to use Windows or acrobat viewers for that - just use Ghostscript 3.5.1... > > FTC, All information in this posting came from material Intel disclosed > in the www.intel.com site or from the Programmers Manual. > > Frank Durda IV <uhclem@nemesis.lonestar.org>|"I can't tell the difference > or uhclem%nemesis@rwsystr.nkn.net | between Whizzo butter (pointing) > | this dead crab (pointing)." > or ...letni!rwsys!nemesis!uhclem | - Monty Pythons' Flying Circus > >
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.3.91.960324131653.21822B-100000>