 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
 |
|
||||||||||||||
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
|
 |
 |
 |
 |
|
|||||||||||
 |
 |
 |
|
||||||||||||
|
|
|
|
|
 |
|
|||
 |
 |
|
||
 |
|
|||
 |
|
|||
 |
|
|||
 |
|
|||
 |
|
|||
 |
|
|||
 |
|
|||
 |
|
|||
 |
 |
 |
|
|
 |
|
|||
1995-1996 Distinguished Lectures
RICHARD L. SITES
Digital Equipment Corporation
Tuesday
AbstractAs high-speed computers and operating systems get more complex, software performance no longer increases predictably. The CPU chip industry has now reached the point that instructions can be executed more quickly than the chips can be fed with code and data. Future chip design is memory design. Future software design is also memory design. We have been studying the performance of the Windows NT operating system running on Alpha computers. The technique we use, PatchWrx, is focussed on understanding operating system performance, but may be applied to user code also. The technique involves patching all the binary executables of the operating system to record all the branching behavior, then postprocessing to reconstruct the entire dynamic instruction and data streams. Patched programs run at about 1/4 of normal speed. Direct measurements of the SQLserver database application reveal that its memory access pattern is nearly indistinguishable from white noise, and that caches therefore don't work. Executing SQLserver and similar 2 GB/sec of main memory bandwidth, unless the software is restructured to treat main memory like disks -- accesses are very slow, and once you get one piece of data, you want to use an entire block of data. Without extreme care in memory access patterns, pointer structures are death. This is in stark contrast to what one might "learn" about computer design by studying the SPEC benchmarks. Controlling memory access patterns will drive hardware and software designs for the foreseeable future.
CARLO H. SEQUIN
University of California, Berkeley
Tuesday
AbstractReal-time virtual walk-through exploration of computer graphics models of fully furnished buildings comprising millions of polygons are possible on today's high-end workstations. However, real-time rendering of such virtual environments requires a suitably organized database, the use of furniture models at various levels of detail, and efficient determination of worst-case visibility ranges. To make such environments truly interactive, so that the user can displace furniture during a walk-through session, raises additional computational challenges and user interface issues. This paper discusses these issues, and shows results obtained on a model comprising more than a hundred furnished rooms.
NORMAN ABRAMSON
Aloha Networks
Thursday
AbstractDigital networks serving large numbers of users sharing a common channel require a multiple access communications architecture in order to regulate user access to the shared channel resource. When the user traffic is rapidly varying, two basic multiple access techniques have been employed -- Code Division Multiple Access (CDMA) and ALOHA. CDMA requires the use of a wide bandwidth spread spectrum channel while ALOHA access has generally been employed in narrowband channels. When ALOHA channels are designed to make the same profligate use of bandwidth as CDMA channels a new form of ALOHA access called Spread ALOHA results. Spread ALOHA is equivalent to conventional CDMA with but a single code shared by all users rather than a different code for each user. Spread ALOHA can be used to greatly simplify the generally baroque architecture of conventional CDMA implementations. In addition it can be shown that in the case of a bandlimited average power limited channel operating at a low signal to noise ratio, it is not possible to find a multiple access technique with a higher capacity than Spread ALOHA.
FOREST BASKETT
Silicon Graphics, Inc.
Tuesday Sorry, no abstract available.
CHRISTOS H. PAPADIMITRIOU
University of California, San Diego
Thursday
Abstract As the field of computational complexity grows and matures, it is tackling
broader and more subtle questions than just how hard functions are to compute.
Complexity of a computational problem is often used as an indirect indication
that an area or approach is mathematically nasty or devoid of favorable structure.
Such a conception of complexity is particularly productive and informative in
interdisciplinary research. I shall present examples in which complexity is
used as an "allegory" for chaos in dynamical systems, for unfair distribution
in economics, and for unbounded rationality in game theory. |
|||||||||||||||||||||||||||||||||||||||||||
|
