Physical synthesis combines logic optimization, placement, and routing into one step. Traditionally, synthesis is done first, creating a netlist with estimated wire delays, but you don't know the real wire delays until after placement and routing, and the placement and routing may not be appropriate for the design constraints. Physical synthesis can do a better job, since it deals only with real wire delays.
http://www.fpgajournal.com/articles/20041026_mentor.htm On 5/16/05, Mike Matera <[EMAIL PROTECTED]> wrote: > Hi, > > Xilinx has a tool called XST (a Verilog/VHDL synthesizer) that comes as > a part of the free webpack. Is that what you mean by physical > synthesis, or am I way of (again) ;-) > > Cheers > ./m > > Timothy Miller wrote: > > On 5/14/05, Peter Karlsson <[EMAIL PROTECTED]> wrote: > > > >>On Fri, 13 May 2005, Hugh Fisher wrote: > >> > >> > >>>I want a 3D graphics card with OK performance which anyone can > >>>write a device driver for. I really, really, don't care about > >>>whether or not the RTL/Verilog/hardware design would be "free" > >>>according to the FSF. > >> > >>I'm not "zealous" about opening up the RTL but there are some benefits to > >>it; an obvious one would be that educating a new generation of hardware > >>hackers could lead to hereto unknown inventions in the graphics area and, > >>if you ever have played with the C64, sinclair spectrum, amiga etc. you > >>would know that there are quite a few games and "demos" that pushes the > >>hardware much further than what the hardware manufacturers thought > >>possible, all because hackers had access to the hardware specs. Releasing > >>the RTL would enable the same process taking place on an fpga board... > > > > > > You'll be able to do this already. Honestly, I don't think the RTL is > > going to tell anyone anything they don't already know. You know the > > register structure, and you know the general pipelining. What you > > don't know (and don't care about) is exactly how many pipeline stages > > are required for, say, the texture unit. Some readers will be > > interested in how I break it down, but the RTL will be much less > > useful as documentation than the documentation that you'll already > > have, in terms of useful programmers' reference. > > > > > >>>At the recent LinuxConf in Australia Wayne Piekarski, augmented > >>>VR guru, said that he uses nVidia because they work, and that's > >>>the criteria that matters to a 3D developer. I don't recall > >>>Keith Packard complaining about "closed hardware" graphic cards > >>>either when he was talking about the future of X Windows. > >> > >>No arguments there as long as you mean that only the RTL is closed. But > >>for Keith (and other X hackers) I would imagine having full specifications > >>on the hardware would help when making new extensions etc... > > > > > > You'll have all that. > > > > > >>>When one of these "release the RTL" zealots starts emailing > >>>from a box that runs on a free CPU design rather than something > >>>by Intel/AMD/IBM, maybe I'll pay some attention to them. > >> > >>Be careful what you wish for. ;-) > >> > >>http://www.opencores.org/projects.cgi/web/or1k/openrisc_1200 > >>http://www.estec.esa.nl/wsmwww/leon/ > >>http://www.fpgacpu.org/links.html > >> > >>But of course you would have to manufacture them yourself... :-) > > > > > > The prototype board will be a huge enabler in this area. > > > > And you know what Xilinx needs? A good physical synthesis tool. I'm > > sure Xilinx wouldn't mind the free software community developing > > software to get more bang out of their chips. > > > > _______________________________________________ > > Open-graphics mailing list > > [email protected] > > http://lists.duskglow.com/mailman/listinfo/open-graphics > > List service provided by Duskglow Consulting, LLC (www.duskglow.com) > > > > _______________________________________________ > Open-graphics mailing list > [email protected] > http://lists.duskglow.com/mailman/listinfo/open-graphics > List service provided by Duskglow Consulting, LLC (www.duskglow.com) > _______________________________________________ Open-graphics mailing list [email protected] http://lists.duskglow.com/mailman/listinfo/open-graphics List service provided by Duskglow Consulting, LLC (www.duskglow.com)
