Simon Hayhurst (simon@zermatt.engr.sgi.com)
Wed, 18 Aug 1999 10:19:10 -0700 (PDT)
Although this is specifically Performer oriented, not Optimizer or
Volumizer, it has a very good summary of a lot that's going on. For those
of you who weren't at the FOP or on the info-performer list give it a
read.
Ciao,
Simon
--- Forwarded mail from --- Forwarded mail from allan@southpark (Allan
Schaffer)
Date: Wed, 18 Aug 1999 09:28:53 -0700 (PDT)
From: From: allan@southpark (Allan Schaffer)
To: info-performer@sgi.com
Subject: Friends of Performer Meeting Summary
Performers,
It was truly a delight to see so many of you at the Friends of
Performer meeting during SIGGRAPH last week. It's your interest and
support that drives our efforts, and having a chance to meet with you
and see IRIS Performer being used in so many ways around the show was
exhilirating and gratifying for us all.
For those who didn't make it to SIGGRAPH this year, I'd like to pass
along a report of my impressions of the show, what was said at the
meeting, and some of the big SGI news. All from the IRIS Performer
point of view of course. :-)
Before I begin, please allow me to apologize for the length, but I do
encourage you to read through the whole thing! All kinds of rumors
about changes at SGI have been flying around recently, some of which
started to bubble up here on info-performer last week. Many of us
here at SGI had spent the previous weeks just dying to respond to the
FUD and to expand upon our plans for success. So, I'd like to let
you all know what we're doing -- directly -- rather than force you to
deduce things from press releases and translations of them into
journalistic prose. On the flip side, please understand that my
interpretations are subject both to the possibility of error or
omission. Please refer to http://www.sgi.com for the official
announcements (and disclaimers).
The show was big, and seemed to me slightly larger and busier than
previous years. As usual the SGI booths were crowded. I can't do
justice to the many things we were showing around the booth (a
variety of Servers, Graphics systems, and products for entertainment
and the Internet) so for a more complete description please see
http://www.sgi.com/events/siggraph99
On the graphics side IRIS Performer was very well represented in the
RealityCenter section of the booth and also with two booth stations.
The RealityCenter was a 4-pipe Onyx2 InfiniteReality system driving a
(very wide and bright) panoramic display. Another single-pipe Onyx2
InfiniteReality with IRIS Performer 2.2.6 (the latest release) drove
the demos for the first of our two stations. For both Onyx2's the
demos included several from our partners at Aechelon Technology, BVR,
DePinxi, InfoByte, MathEngine, MPI, Open Worlds, RT-Set, and also
some that were SGI-made, including the 'Space-to-Face', Yosemite, and
others. The second Performer station featured an SGI 320 Visual
Workstation showing Performer on Linux, aka Mongoose; more on this in
a moment.
Friends of Performer began on Wednesday night with an enthusiastic
audience of 250+, the most ever! Fellow Performers Chris Insinger
and Angus Dorbie kicked off the meeting with a cute and
crowd-pleasing Linux-based demo of the (formerly flightless) Linux
Penguin getting a rocket-propelled boost from our favorite toolkit.
Performer's engineering manager Jenny Zhao then presented the State
of Performer and also made several announcements. Jenny began by
directly answering the questions people were dying to ask:
1. Is SGI getting out of the graphics business?
"HELL NO!!"
- Sharon Clay
No. This rumor mostly came from the SGI-NVIDIA business
alliance. Please read on for a more detailed answer.
2. Is SGI getting out of the High-End graphics business?
No. (HELL no!). We are hard at work on IR3, the successor to
IR2 in the InfiniteReality product line; and we are keeping our
commitment of producing a revolutionary next generation
high-end graphics machine: RE4. More on this below. And
Performer, as always, will be the first there, ready to bring
you the advanced features and optimized performance on all our
graphics machines.
3. Is Fahrenheit dead?
No, our efforts along with Microsoft continue. However, SGI
has decided not to support Fahrenheit on IRIX, and we have
reduced our resources there accordingly. Fahrenheit is still
an ongoing project for Windows-based systems. Microsoft will
handle the beta and release process.
4. Is IRIX going away?
No. IRIX continues as our primary platform for our MIPS-based
servers and workstations. SGI will continue to support and
enhance IRIX on MIPS, with a MIPS roadmap that provides
processor upgrades for several years yet. More on this below.
5. Which API should I use?
It depends on what problem you are trying to solve. If it is a
vis-sim application, our answer is, has always been, and will
be: IRIS Performer. Performer is the premier vis-sim API with
new releases scheduled for many years to come. We will keep
evolving Performer to support the visual simulation market for
today and tomorrow.
6. What is SGI's business strategy?
For the answer to this question, Jenny then welcomed Drew Henry,
Vice President & General Manager of the Visual Solutions Business
Unit at SGI. Drew heads our division, which is responsible for all
of SGI's Visual Computing and High-End Graphics systems (such as
InfiniteReality), our Advanced Media products (including the new
HDTV solutions and our digital media & video hardware & software),
Advanced Graphics Software (including Performer), our MS&I
business, and several other groups. Drew discussed many of the
dramatic announcements from the day before, which I will try to
capture here:
1. Strategic focusing of the company, centered around three
business areas, to deliver sustainable profitability and
growth:
- High Performance Systems: (aka: SERVERS)
Build the most scalable, high-performance servers for HPC,
technical computing, and Business Intelligence Applications.
- Visual Computing Solutions (aka: GRAPHICS)
Powerful solutions for collaboration, visualization of
complex data and media-rich content creation. This is where
IRIS Performer and our high-end graphics systems fit in.
- Broadband Systems (aka: INTERNET)
Internet infrastructure products with "appliance-like"
features for broadband content, applications, and services.
2. Invigorated, continuing, dedicated support for IRIX on our
MIPS-based systems.
IRIX continues as our primary platform for our MIPS-based
servers and workstations. SGI will continue to support and
enhance IRIX on MIPS, with a MIPS roadmap that provides
processor upgrades for several years yet. We have shipped a
number of 256 CPU Single-System-Image Origin & Onyx2 systems
running the standard IRIX OS and are developing a 512 CPU
version. We have already scheduled new systems with faster
CPUs to launch as far in the future as 2002 and will continue
long after that. These new systems will include both servers
and graphics workstations. IRIX, the best high-end Unix OS
there is, isn't going anywhere but up.
For those of you in the SGI Developer's Program, there's some
great reading about our OS strategy here:
http://www.sgi.com/developers/technology/
3. Position SGI to drive the market's acceptance of Linux and to
increase its feature set and scalability.
We will agressively push Linux forward as fast as we can, and
contribute technology from IRIX into Linux as quickly as the
Linux Community is willing to accept it. Our goal -- first, on
our new and upcoming IA-32 machines (both servers and
workstations), and secondly on our larger systems once Intel
brings IA-64 into the market -- is to allow us to have the
technology and robustness of IRIX with the Applications Capture
of Linux.
While still relatively immature, the Linux market is rapidly
shifting from early adopters to the early majority, primarily
in technically savvy markets. These markets have strong
affinity with SGI and we believe that being a recognized
innovator in the Linux community will further accelerate
Linux's adoption and that we can grow dramatically along with
it.
"We're using our ability to create and contribute
valuable code to Open Source in order to drive Linux
forward, not just as a viable operating system, but
as the premier UNIX-like operating system. If we
are successful in doing that, we will find ourselves
in the position of being a leader in operating system
technology--leaders not just in technology, but leaders
in the more important terms of the tools vast numbers
of people actually use. We will then have a viable
way of expressing our value in a way that benefits a
lot of people."
- Kurt Akeley, CTO, SGI
4. The formation of a new business unit to manage SGI's Visual
Workstation line of Windows NT platform-based workstations.
Our intention is to partner with another company who will
handle the sales, distribution, manufacturing, and marketing of
the 320 and 540 Visual Workstations (and their successors),
while SGI engineering talent continues to bring new innovations
to these platforms and the product line.
Drew said SGI had reached a preliminary understanding with
another computer systems company to form this joint venture;
Terms of the deal have not been finalized, so we don't have
more precise details to announce yet.
5. SGI announced further details of its previously announced
alliance with NVIDIA Corporation.
Under the alliance the two companies will work together on
next-generation midrange graphics products and an SGI graphics
engineering team will transfer to NVIDIA. The arrangement will
allow SGI to extend its leadership in its midrange graphics
systems while at the same time taking advantage of technology
and process developments in the high-volume graphics ASIC market.
This particular announcement was arguably the most widely
misunderstood. We're not bundling NVidia's game boards into
Octanes, we're not "moving all our graphics engineers" (or
most, or even very many) to NVidia, and we're not making boards
for beige PC's. The idea is for NVidia and SGI to develop in
close partnership -- even using some of our engineering talent
-- certain custom and exclusive-to-SGI ASICs which we can then
design into a comprehensive midrange graphics architecture for
our next-generation desktop workstations. Think of the 320,
the way we took Pentium chips and made a revolutionary new
system based on our IVC architecture. We'll be doing similar
things here but in Graphics. This partnership allows us to
focus our resources on core competencies (certain custom ASICs
for targeted features, and graphics architecture design);
improves our time to market; brings forth quicker product
"turns"; and allows us to build systems at a lower cost.
Press releases:
http://www.sgi.com/newsroom/press_releases/1999/july/nvidia_alliance.html
http://www.sgi.com/newsroom/press_releases/1999/august/nvidia.html
6. The formation of a business unit to manage SGI's Cray-branded
product family (T90, SV1, SV2, and T3E) and future vector-based
products.
SGI is in discussions with potential partners to assume the
operation of this business, which will carry forward the
current product roadmap of vector-based solutions to be
developed in conjunction with SGI's next-generation ccNUMA-
based server products.
Press release:
http://www.sgi.com/newsroom/press_releases/1999/august/cray.html
Jenny then proudly and officially unveiled the "Mongoose" project,
our effort to make IRIS Performer available for the Linux platform.
Mongoose, aka Performer-on-Linux, is:
1. Fully API-compatible with IRIS Performer 2.2.x
2. A full release:
- Core Performer library DSOs
- Source code included for libpfutil, libpfui, libpfdb, libpfdu
- Sample source code and Linux-ready Makefiles included
- man pages
3. Tested and supported on SGI Linux-based platforms.
4. Supports MESA, OpenGL 1.2, Redhat Linux.
5. Available for all IA32-based Linux systems.
The release schedule for "Mongoose 1.0" is:
Beta: Sept. 20th, 1999
MR: Nov. 22nd, 1999
We welcome everyone to try out the beta, which will be freely
available to download from the Performer website. We also welcome
contributions from our friends, such as loaders and sample code.
We have not yet reached a decision regarding pricing of the final
release, although we do expect it to be less expensive than the IRIX
version of Performer. The first release of Mongoose will be a
single-process version only. We do not plan for it to be open-source.
Next, Jenny had to remind everyone that amidst all the excitement
surrounding Mongoose, there is a lot happening in the overall world
of Performer too. Going forward, we have a lot of work and new
development planned to make Performer more robust.
The next full release of Performer, planned to come around the Spring
of 2000 will include:
1. New features for increased flexibility and ease-of-use
2. Improvement on scalability between APP and CULL
3. Better man pages and a new edition of the Performer Programmer's guide
4. More sample code and integration demos
5. Double precision transforms support
6. Cross-platform support of both IRIX and Linux
7. Multi-process support for the Linux version
We also have several further releases planned after this Spring 2000
version. As before, Performer releases tend to fall near the dates
scheduled for new SGI graphics systems, so that Performer can be
there right from the beginning to support the advanced features of
new hardware.
On the systems side, Jenny revealed that work is underway to complete
IR3, our next High-End graphics system in the InfiniteReality product
line, scheduled to ship in the fall of 2000. RE4, the next
generation High-End graphics machine with a radically new
architecture and feature set, will release in the fall of 2001.
We will announce the release schedule for Performer when we get
closer to those dates.
Jenny continued onward to my introduction and also introduced our
newest Performer Team member, Tom Flynn. I outlined the current
status and imminent plans for Performer on Linux. A personal note,
Jenny referred to those of us working on it as the Mongeese, _not_
the Monkeys!
I began with the Mongoose slogan, explaining our project name:
"Mongooses are famous for their snake-fighting ability,
and are almost always victorious because of their speed,
agility, and timing, and also because of their thick coat."
I then went on to explain and re-emphasize some of the details
surrounding Mongoose:
1. Based on IRIS Performer 2.2.x API, with _no_ API changes
2. Common code base -- as changes, innovations, bugfixes, etc are
made in Performer, they will appear simultaneously in both the
IRIX and Linux versions
3. Freely-downloadable Beta will be available September 20th;
will be distributed as rpm's and tarballs.
4. Runs with RedHat 6.0, using either MESA or the SGI-Linux
Accelerated OpenGL (not yet available)
5. The majority of the library is already ported:
- libpf, libpr, libpfutil, libpfui, libpfuiD, libpfdu
- ~45 of 71 loaders ported
- Sample programs & perfly "just compile"
- Minimal porting effort for user-level code
6. Not yet ported:
- Multiprocessing (Forked Cull, Draw, Isect, DBase, X Input)
- Multiple pipes or Multiple windows
- IRIX-Specific OS features:
- Shared Memory Arenas
- REACT features
- DataPools, Locks
- Direct File I/O
7. InfiniteReality-specific Performer features not available on Linux:
- SGI Large-Format Texture Extensions (ClipTexture)
- SGI Video Channel Extensions (for DVR, etc)
- SGI Fog modes
- SGI Graphics Pipeline Instrumentation
- SGI Calligraphics
8. We'll be adding more information about our progress and about
Mongoose in general to the Performer Web site in the coming
weeks.
After thanking those present from the SGI OpenGL group for their
tireless (and ongoing!) work in preparing OpenGL for Linux, Tom and I
showed perfly with the Performer Town database running on an SGI 320
Visual Workstation. Besides moving around in the Town itself, we
demonstrated the regular, graphics, and fill stats; phase and
overload control; fog, and several other GUI features in perfly.
The system had two 450Mhz Pentium-II processors and 512MB RAM, and
was running Red Hat Linux 6.0 with an SGI-developed accelerated
OpenGL library for the built-in Cobalt Graphics chipset. We saw
consistent frame rates of 30Hz -- soon to go higher! This was the
same technology-demo system we had on the show floor.
For a screen snapshot, please point your browser to:
http://reality.sgi.com/performer/images/mongoose-snap.jpg
We then ran the same demo on a Dell Inspiron 7000 laptop -- "great
for hacking on Performer code on the airplane." The system had one
266Mhz Pentium II processor and 128MB RAM, also running Red Hat Linux
6.0 and using MESA (a software-only implementation of OpenGL) as the
underlying OpenGL layer. Untextured frame rates were 5-8 Hz; with
texture there was some slowdown. :-) Both demos generated a rousing
amount of applause from those present.
Our next speaker was Kurt Akeley, an SGI founder and Senior Vice
President & Chief Technical Officer. After a brief pop quiz with a
hand-made Mongoose Polo Shirt as the prize, ("How many vertices could
the IRIS GTX system transform per second?" Could the winner please
identify yourself here?) Kurt discussed the thinking behind many of
the announcements and spelled out our continued vision for High-End
Graphics.
He re-emphasized that -- in spite of the rumors -- we have no
intentions to change what we do best: use our graphics heritage and
expertise to bring new and revolutionary high-end visualization
systems to market. What IS changing is how we intend to build them.
Said simply, the goal is to build high-end graphics systems using
more off-the-shelf technology than in previous product designs. We
did it with RealityEngine (the Geometry Engine ASICs were Intel i860
processors), we did it with parts of the display back-end for
InfiniteReality, and now we're doing it again. It's a shift from
architecting every last part of our graphics systems, instead
focusing on creating brilliant graphics system architectures which
can take best advantage of OTS parts. Similar to what was said
above, we can focus our resources on core competencies (doing custom
ASICs for targeted features, and graphics architecture design);
improve our time to market; bring forth quicker product "turns"; and
build systems at a lower cost. It's worth rementioning that this
initiative is independent of what's happening with NVidia, which is
intended for our next-generation midrange systems.
Kurt then led the Q&A session, and when it was done almost called the
meeting to an end -- but there was more to come!
We had originally intended to have our fellow Performer Don Burns
speak next, but due to the packed schedule we unfortunately didn't
have an opportunity to hear from him. The topic of Don's talk was
"Integrating Volumizer into Performer"; he has created an example to
show how to integrate Volumizer data into IRIS Performer by using
Performer's pfFlux feature. A set of pfFlux objects are generated by
Volumizer at run-time which are then be handled by Performer and
integrated into the scene graph. The generation can be carried out
either in the main application or in the COMPUTE process. We'll be
adding Don's sample code and presentation slides to the Performer
website in the next couple weeks.
Our next speaker was Carter Emmart, Astrophysics artist from the
Digital Galaxy project at NASA, who along with Performer Friend Nacho
Sans-Pastor of Aechelon Technology, gave us a sneak-preview of their
virtual universe, and a passionate discourse of how visualization can
bring understanding of the gifts of science and humanity to us all.
The demo itself was a scientifically accurate digital galaxy
containing close to five *Billion* objects, based on actual
astronomical data and being delivered in real time. It will be shown
in its completed form in early 2000 when the Museum opens the Hayden
Planetarium, the centerpiece of the new Frederick Phineas and Sandra
Priest Rose Center for Earth and Space. Honest folks, it was
spectacular.
Our final speaker, Performer Friend Ran Yakir from BVR, brought us
gently back to earth with a discussion of the issues involved with
using IRIS Performer for ground simulation. He focused on a very
challenging problem: planting cultural features such as trees,
fences, roads, and rivers to preserve realism. Since a simulation
with thousands or perhaps tens of thousands of individual trees can
overwhelm the rendering system, one traditional method of avoiding
this is to create walls or boxes with tree-wall textures that viewers
can not walk or drive through. Ran's solution is to plant individual
features dynamically in real-time using OpenGL. As the viewer
marches in the forest, trees fade in as they get closer and trees
behind the user are faded out.
Ran finished with a self-modifying ground simulation demo within
which small trees could fall if pushed over, and craters, ditches,
and other cultural features could be placed within the database in
real-time as the user moves across the landscape.
Chris Insinger then returned to the podium to call the meeting to an
end, and so ends this message as well. So on behalf of the Performer
team: thank you to all who attended, and thank you all for helping
us have such a successful meeting and SIGGRAPH '99! See you at I/ITSEC!
Allan
----
Allan Schaffer allan@sgi.com
Silicon Graphics http://reality.sgi.com/allan
-----------------------------------------------------------------------
List Archives, FAQ, FTP: http://www.sgi.com/software/performer/
Submissions: info-performer@sgi.com
Admin. requests: info-performer-request@sgi.com
---End of forwarded mail from ---End of forwarded mail from allan@southpark
(Allan Schaffer)
--
--------------------------------------------------------------------------------
Simon Hayhurst | Phone (1)-650-933 6258
Onyx2 Product Management and Applied Engineering | Fax (1)-650-964 8671
"Opening Minds, Closing Deals" | V-mail 933-6258
"Leadership comes from superior insight" | E-mail simon @ sgi.com
--------------------------------------------------------------------------------
This archive was generated by hypermail 2.0b2 on Mon Nov 01 1999 - 14:21:40 PST