Graphics Software Standard Interfaces
by Roy Latham

An Applications Programming Interface (API) is a standardized set of computer subprogram calls designed to help a programmer incorporate a proprietary product into a larger application. For example, if one wants to write a CAD/CAM graphics package, one may write a program which makes subroutine calls to an API called OpenGL. OpenGL is a standard interface developed by Silicon Graphics and subsequently endorsed by Microsoft. Code written in OpenGL would typically include subroutine calls to do things like "draw a triangle." The details of exactly how the triangle is drawn are inside OpenGL and are hidden from the applications programmer. This leaves open the possibility of having different implementations of OpenGL, all of which work with the application because they all have the same subprogram calls.

Different implementations of OpenGL are written for different graphics accelerators. If a computer running Microsoft software does not have a graphics accelerator, Microsoft provides a software implementation that runs on the CPU. If the computer is upgraded with a hardware accelerator, the maker of the accelerator board may supply a version of OpenGL than routes the OpenGL commands to the board, converting the control sequences to commands appropriate to that particular hardware.

Without a standard API, every application would have to be rewritten to work with each new set of proprietary hardware. Rewriting applications is expensive. By using a standard API, accelerator manufacturers can get a much larger base of applications software at low cost. Computer users benefit from having the many applications and by having fewer worries about whether their software investment will be maintained when hardware is upgraded.

A limitation of using a standard API is that there will often be inefficiency in having the extra layer of software between the application and the hardware device. If the API is well written and a close match to the capabilities of the hardware, the inefficiency will be within acceptable bounds. Also, the standard must be established and widely accepted for the whole scheme to work.

OpenGL is a widely accepted standard API for high-end graphics applications. It was designed with CAD/CAM applications in mind rather than simulator applications, so it is not perfectly suited to real time graphics, but in many cases it suffices.

Microsoft is so large they can afford to have many standards. In May, they were set to release a new graphics API called Direct3D. Direct3D is aimed at applications like 3-D games. The emerging market for low-cost 3-D graphics sorely needs a standard API, and Direct3D has arrived none to soon.

Presently, games are being written directly for proprietary graphics boards, and users will only be able to use games written for the proprietary product they happen to own. Consequently, each of the more than 30 vendors hoping to cash in on the home PC 3-D game market must convince game producers to write for its particular product line. One insider tells us that competition for game developers' attention has reached the point where developers now demand around US $250,000 per game from a graphics vendor to develop a game version compatible with that graphics vendor's product.

If there were 30 proprietary interfaces fighting for game developers' attention and for consumer acceptance, the market place would be chaotic and we would expect to ultimately see many orphaned products. At times like this one must acknowledge, however begrudgingly, that it is good to have Microsoft as plenipotentiary. Microsoft's market dominance all but ensures the rapid acceptance of Direct3D as the standard API. Thus, the lives of 30 or more chip vendors may be spared. Or at least they will live or die based upon cost and quality rather than upon their ability to charm or bribe the support of game developers.

Limitations of Low-Cost Graphics

Limitations of most of the current crop of low cost graphics products are likely to postpone application to many "industrial strength" simulators. Good polygon performance numbers and texture implementations conjure up dreams of using emerging PC products for low cost driver trainers, military simulators, and location-based entertainment attractions. The biggest problem is lack of antialiasing. Poor image quality is deemed acceptable for home PC games, the target of the initial marketing of the 3-D boards. However, for training purposes the digital artifacts are quite distracting and unlikely to be accepted for day-in, day-out use. A second limitation is the general inability of low cost products to support high resolution screen at real time update rates. There are few strategies for paralleling low cost chips to use with million-plus pixel displays.

We are likely to wait about three years for a new generation of products to overcome these limitations at low cost.

Aerospace Consolidation Hurts Professional Organizations

Let's recall the consolidation of the aerospace simulation industry one more time. Five years ago, for example, Loral teamed with GE Aerospace to bid the Close Combat Tactical Trainer, and other bidders were IBM Federal Systems, Hughes, Link, and Martin. Then Martin acquired GE Aerospace, Loral acquired IBM Federal Systems, and Hughes acquired Link. Next Martin acquired GE Aerospace, after that Martin merged with Lockheed, and the combined Lockheed Martin acquired Loral. This leaves Lockheed Martin and Hughes as the two large prime contractors in U.S. aerospace simulation. In Europe, Thomson acquired Link-Miles then later acquired Reflectone, which Hughes had previously acquired. Loral acquired Solartron, and, as we mentioned, has since itself been acquired by Lockheed Martin.

A major side effect of the consolidation has been to hurt professional and trade organizations, such as the National Security Industrial Association (that co-sponsors the I/ITSEC simulator trade show) and the IMAGE Society, which sponsors the IMAGE Conferences that are focused on visual simulation. In Europe, consolidation is hurting ITEC especially with Thomson choosing not to exhibit. With fewer large corporate sponsors, the associations and societies receive less support. One may be so bold as to ask why we need aerospace simulation conferences in the US at all if in fact there are only two serious aerospace competitors.

A partial answer is that the uses of simulation technology are expanding. There is more work available for smaller companies aside from the giant companies that seek correspondingly giant contracts. Another partial answer is that distributed interactive simulation is a whole new market which is developing its own set of serious players, Scientific Applications International Corporation (SAIC) and TASC to name two. The DIS Workshops are booming. So a partial answer to the problems of aerospace simulation organizations is for the organizations to move in the direction of new aerospace applications. They seem to be trying to do that, but these efforts are not enough to compensate for the effects of aerospace consolidation as a whole.

Another route is for professional organizations to attempt to tap the growth in non-aerospace simulation. Growth applications include driving simulators, industrial simulators, and especially entertainment simulators. The organizations pursuing these new applications tend not to be the traditional aerospace suppliers, with some major exceptions of course. Tapping non-traditional applications is easier said than done. It's very difficult to get new people to participate in any organization, let alone one that they do not perceive as being single-mindedly designed for their industry. Even people who once worked in aerospace seem not to think of participating in their old forums once they have left for a job in entertainment.

Oddly, an obstacle to participation by entertainment developers is that as a whole they are far more secretive than aerospace or military systems developers. Governments have procedures for determining what is a secret and what is not, whereas, entertainment developers tend to be paranoid about everything.

We do not have a good solution to this circumstance. We do urge professionals at organizations large and small to understand and value the role of professional organizations. By exchanging information, we each can do a better job in fulfilling the needs of our clients. If you happen to make the switch from a traditional branch of simulation to an emerging one, don't forget the professional organizations to which you have contributed and from which you have benefited. It is appropriate to point out these benefits to the management of a new enterprise. Sometimes you will be surprised by their enthusiasm, once told.