Date: Wed, 26 Feb 1997 17:30:36 -0500 (EST) From: Mark Mayo <mark@quickweb.com> To: Satoshi Asami <asami@vader.cs.berkeley.edu> Cc: Brett_Glass_at_POST-IW1@infoworld.com, bde@zeta.org.au, jin@george.lbl.gov, freebsd-hardware@FreeBSD.ORG, kuku@gilberto.physik.rwth-aachen.de, robsch@robkaos.ruhr.de Subject: Re: Memory speed of P6-200 (256k) Message-ID: <Pine.BSF.3.95q.970226172737.12865C-100000@vinyl.quickweb.com> In-Reply-To: <199702262218.OAA02909@vader.cs.berkeley.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, 26 Feb 1997, Satoshi Asami wrote: > * How about MMX, guys? Is there any way to speed up copies by bursting in and > * out of the MMX registers? The FPU? > > The P5-optimized copyin/out already uses the FPU. We've verified that > it doesn't help the 486 and P6 (at least with Orion and Natoma). > > I don't know about MMX, but from what Bruce has observed, it may help > lower the cost of load/store (which is pretty tough on CPUs for > floating-point ops). If someone with an MMX can step forward to be > the guinea pig, I'm sure Bruce will be happy to help. :) The problem is that (AFAIK) there aren't any C compilers iwth MMX support yet.. Even the big guys (Watcom and Microsoft) only have inline assembler support. Of course, if this stuff is being coded in assembler, you're on your way! I'll be getting a MMX machine at school soon to play with (working on vector stuff - going to prove that Intel MMX is still slower than the native vectoring operations of the PPC and Alpha) so I can test any code that shows up. -Mark > > Satoshi > ---------------------------------------------------------------------------- Mark Mayo mark@quickweb.com RingZero Comp. http://vinyl.quickweb.com/mark ---------------------------------------------------------------------------- Nature shows that with the growth of intelligence comes increased capacity for pain, and it is only with the highest degree of intelligence that suffering reaches its supreme point. -- Arthur Schopenhauer
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.3.95q.970226172737.12865C-100000>