A distributed 3D graphics library
Proceedings of the 25th annual conference on Computer graphics and interactive techniques
| Hi-index | 0.00 |
Good morning, ladies and gentlemen. Welcome to the panelentitled Distributed Graphics: Where to Draw the Lines?My name is Dick Phillips. I'm from Los Alamos NationalLaboratory and I'll be your chair this session. I'll be joined by agreat group of panelists --- friends and colleagues all.Our second speaker following me will be Michael Pique fromScripps Clinic. Following him will be Cleve Moler from ArdentComputer. After Cleve we'll hear from Jay Torborg who is associatedwith Alliant Computer. And batting in the clean-up position isgoing to be Don Greenberg from Cornell University.I have to give you one administrative announcement. You probablyknow this by now if you've been attending panel sessions all week.But once again, these proceedings are being audio taped forsubsequent transcription and publication. That means that when weopen up the session for question and answer, which will be inanother 30 or 40 minutes, if you would like to ask a question, youmust come to one of the microphones that's situated in the aisles.They are just about in every aisle, part way back and close to thefront. And to be recognized, please state your name andaffiliation, and I'll remind you of that when we get into thequestion and answer session.The title of our panel begs a question --- where to draw thelines. Well, the trivial answer to that question is obviously onthe display that you have available. The real implication of thattitle was where to draw the lines of demarcation for graphicsprocessing. You're going to hear from me and from my otherpanelists several different points of view. Just when you thoughteverything was settling down and it was clear that all graphicsprocessing was moving out to workstations or graphicsupercomputers, you're going to hear at least two different pointsof view that may sound a bit nostalgic.Let me take you back in time just a bit, and this is a greatlyoversimplified graphics time line --- where we have been and wherewe are and where we're going in the evolution of visualizationcapability.I'm not going to dwell too much on the part of this time line tothe left. We're really interested in what's up at the right handside. But I can't resist pointing out that back in the days which Ihave labeled pre-history here, a lot of us can remember gettingexcited about seeing output in the form of a printer plot, thinkingthat we were doing visualization and that that was really computergraphics. And I for one can remember the first time I had 300 bandavailable to me on a storage tube terminal and I thought this isblazing speed. I cannot believe what kind of graphics capability Ihave got now.Where things really get interesting though, if you move alongthat time line to the right, up into the mid 1980s, I have put someI think seminal events on there --- Silicon Graphics introducingthe geometry engine in the workstation. Well, workstations ingeneral. That was a real watershed event that has changed the waythat we do graphics and where we do graphics considerably.Then as we move into the later part of the 1980s, I have notedthe appearance of graphics accelerators for workstations. These arespecialized plug-in boards that have all of the graphics featureslike Phong shading and high speed transformations built into them.Graphic supercomputers like Ardent and Stellar and HP/Apollo haveappeared in that time frame. Then we look a little bit further intothe '90s and I have indicated the occurrence of very high speednetworks is going to have a profound effect on the way we dographics display and how we distribute the activities that areassociated with it.Let me give a very oversimplified couple of statements on whatgave rise to the need for specialized graphics hardware --- theaccelerators that I talked about and indeed the graphicsupercomputers. As I've said, to terribly oversimplify, it wascertainly the need for real time transformations and rendering. Allof the advances in computer graphics over the last 10 or 15 years,many of them we can now find built into the hardware of theworkstations and graphic supercomputers that we have available tous.One of the other reasons for wanting to bring all of that highspeed computational capability right to the desktop, as it is, wasto compensate for the lamentably low communication bandwidths whichwe had then --- which we have now, as a matter of fact. And I'meven including Ethernet and I'll be bold enough to say that theFDDI, which is not really upon us, is also in that lamentably slowcategory for many of the kinds of things we'd like to do.It turns out --- in my view, at least --- that that specializedhardware, wonderful as it is for many, many applications, and makeno mistake, it has revolutionized the way that we can dointeractive graphics --- it's not useful for all applications.One application that I've listed as a first bullet is one wherewe're doing specialized rendering --- research rendering let's callit. Not everything we wanted --- not all the research in renderinghas been done --- right? So Gouraud shading and Phong shading andso on is not the be-all end-all necessarily. There's a lot ofinteresting work being done. It has been reported at thisconference, as a matter of fact.That is really a small reason for wanting to do the graphicscomputing on yet another system. But the next one that I've listedis a very compelling reason in many installations, particularlywhere large scale heavy-duty simulations are being done. I'vementioned that I'm from Los Alamos and that's certainly one centerwhere there are computations that are done on supercomputers andthat need to be visualized, and because of the nature of thecomputations all of the specialized hardware in accelerator boardsand in graphic supercomputers is not necessarily useful. Indeed,I'll argue that in many cases it's of no value whatsoever.The last point I want to make here --- before I show you acouple of specific slides of these simulations that I'm referringto --- is that what will happen is that the emergence of very highspeed networks --- both local networks and international andnational networks --- is going to provide a way for these largescale simulations to take advantage of graphics hardware that doesnot necessarily have the specialized capabilities we just talkedabout.At Los Alamos a group of folks in our network engineeringdepartment have taken the lead in defining what is called the HighSpeed Channel specification. Before I get to that, let me just giveyou an idea of the kinds of computations that are being done at LosAlamos --- and I know at many other places --- that simply can'ttake advantage of the specialized hardware that I've just beenreferring to. This happens to be the mesh that's associated with afinite difference computation for some simulation. It doesn'treally matter what it is, but I just wanted to show you that we'retalking typically tens of thousands of individual mesh points, andI can guarantee you this is a fairly sparse mesh compared to thekinds of things that most of our users encounter.The point in showing you this is that as the simulation evolvesin time, there is a different version of this mesh for every singletime step. The scientists who are doing the simulation would liketo be able --- either after the fact or perhaps if the timing isappropriate --- to steer the computations that are going on bybeing able to visualize the evolution in time of meshes like this.And they need to be sent to some display device. And ideally you'dlike to do that at the rate of 24 frames per second, but we can gothrough some computations and find that's simply not feasible withthe kind of network bandwidths that are available today.The specialized hardware that I've just been talking about givesus no help at all here, because what I need to be able to do is tosend one instance of this mesh to the display device for every timestep, as I mentioned a moment ago.In addition, the scientists at Los Alamos and other places wouldlike to be able to have the counterpart of a numerical laboratory.This is completely synthesized, but you can --- and many of you mayhave had experience in the past with visualization techniques andfluid flow, where you can actually see shock waves by variouslighting techniques. The intent here is to be able to simulate thatsituation and be able to show the flow evolving --- not necessarilyas it's being computed, but perhaps after the fact --- but be ableto pick out important points by seeing a temporal evolution of thatparticular simulation.So those are just a couple of examples that have given rise tothe development of a high speed channel specification and anaccompanying network at Los Alamos, and I wanted to say right now--- just so you don't think oh, great, a special purpose solutionfor a national laboratory that no one else will ever be able to use--- not so.Many of you out there I am sure know --- and I know several ofour panelists are either aware of or working on high speed channelhardware for their particular products. There are about 30 vendorsthat have signed on to the high speed channel specification.In addition, Digital Equipment Corporation is building thecorresponding network, which is called CP*. I'm not going to gointo network details here because that's not my point. I reallywanted to describe what is now a new highway for data transmissionthat facilitates my job, which is to help the scientists do thevisualization that they need to do.So what we're seeing here is a very simplified view of how thishigh speed network, which is spec'ed at 800 megabits and acorresponding cross bar switch-style network that is going to alloweffective point-to-point connections between the various componentsof the computing environment --- the supercomputers, the datastorage devices, and the display devices. And each --- unlike witha bus structure --- each user will effectively have that completebandwidth available to him or her.A larger view of that network is shown here and it gives us anidea of how we might interconnect the various devices and again. Idon't want to go through the details here, but you notice thatwe're accommodating FDDI gateways, so that the FDDI LANs can beused easily in this environment, and various workstations. I'veshown a Sun workstation there. I described vendors who are signingon to this concept, and Sun is providing a high speed interface totheir TAAC board, which can then be put into a Sun workstation, andconnected at 800 megabits directly over the network.I mentioned earlier that this is not necessarily limited to justour local area networks. Many of you are probably aware of the workthat's going on now to establish these so-called national datahighways. The Corporation for National Research Initiatives iscoordinating an activity to establish centers throughout the UnitedStates that will participate in a test bed of what is to become inthe 1990s a four gigabit data highway spanning the UnitedStates.So I'd like to leave you with the thought that while we havemigrated a lot of the graphics computing capabilities --- graphicsrelated computing requirements to workstations and specializedgraphic supercomputers --- the emergence of extremely high speeddata communications makes one rethink these situations ---particularly when you are faced with the kinds of computing tasksthat I just mentioned that we have for large scale simulations atnational laboratories.I'm going to stop my heretical remarks here and I'm going toturn it over to the panel to describe several different points ofview.As I mentioned earlier, the next speaker will be Michael Piquefrom Scripps Clinic and Research Foundation. Michael? Thankyou.