Marinus Theodoor Verelst (Theo)
van Bossestraat 99
1051JW Amsterdam
born: the Hague
marital state: unmarried
pre-university: Zandvliet ongedeeld VWO, Den Haag 1978 - 1984
University: Technical University of Delft Fac. of Electrical
Engineering
Master thesis in: Network Theory Section, E. Deprettere, Prof P. Dewilde.
Experience: University: during graduation project and subsequently as University employee with amoung others as purpose a PhD: the design and implementation of a system for rendering complex computer graphics scenes.
Keywords: state of the art computer system design, graphics issues: ray tracing, radiosity, also algorithmically, modeling, graphics model languages, large interactive (in core and on disc, also persistent) databases Nurbs (Rational Cubic Bezier Splines) subdivision based rendering chips and software; run-time scheduled non-regular system design and simulation, also methodology, including process theory, 3D math, including a (new) bounding box estimation for spherical coordinates, concept and implementation of (efficient) distributed rendering system, and hardware system simulation. Graphical user interface for distributed radiosity software, decoupled over (ascii) a network link. Design of messaging protocol between host and (DSP based) accelerator engine, and its (working) distributed workstation (Hewlett Packart and Sun) simulation.
I did the design and implementation of an automated measurement system for measuring the analog line transmission properties of corporate telephone exchange units (PABX), with the dutch telecom company (at that point passing on from being a gouvermental organization PTT to be a independent company KPN), section DWB2 'directorate of corporate telecom systems'
I've set up a private company (2 persons) 'The Source Computing', for teaching Unix/C courses, and writing and starting to commercialize the software packet 'PULS' I wrote, an integrated database maintenance, fast search engine, and real time MIDI interface package for categorizing transfering, and experimenting with DX-7 synthesizer sounds.
I have a lot of experience with design and implementation of electronical digital (computer) circuitry, already before university.
I have major or reasonable knowledge of the following programming languages and environments: C, Unix, objective C, communicatie (sockets, client server), shells, Tcl/Tk, Basics, Lisp, Z80 en 68000 assembly, AVS applications, Khoros, Bones network simulator, C++, Java, (P)Spice.
Lately I've focussed on some long-term interests, and set up the following for that:
An interactive 'block' editor supplied with with tcl/tk functions and blocks
A multithreading environment with a fast, socket-compatible, communication interface between threads and low thread switch latency.
Couplings between various programming environments such as tcl/tk, and C and Java.
3d graphics examples with the 'Mesa' (opengl) environment.
Audio processing frame and examples, including physical modeling and additive synthesis, sample processing , drum track generator and various effects.
A number of digital and electronical designs and circuits, succesfully simulated (latter part of the page) with Microsim (containing PSPICE), including analog synthesizer basic building blocks (see also a synthesizer company aimed cv ).
Prototype (physical) implementation of various circuits: a microprocessor design, a switched filter, a PWM example, a fast DA converter and sample and hold, and some more.
Lately I've been working on (windows95) PC's, with, amoung others, the Cygnus (GNU based) C(++) compiler and UNIX-like environment, before I've worked with trs80, BBC electron, and Atari ST (Z80,6502,68000,all including design and implementation of interface circuitry and added analog and digital hardware), various workstations and mini's, mainly HP (720) and Sun (Sparc).
'pcom', is a tcl/tk tool to connect up two parties for combined mutual file system access and 'talk', that I've done in a few work days, it nicely shows what tcl/tk can be up to.
I've been working with the TMS320C62(7)02 simulator that Texas Instruments recently made available for their evaluation board (200MHz, 1.6 Gflops max, .7 Giga byte /s bus, PCI interface with complete set of DMA functions, will add a few pages), that even allows the counting of clock pulses for critical code segments for pipelines (parallel) code, and has Ansi C.
My interest is based on connection-core to application coverage of distributed processing. I would like to asses how General Purpose network elements function as substrate for Distributed Processing, with both scientific and upcoming general application level multiprocessing in low and medium (price) level distributed systems in mind, such as for graphics, multimedia, multimedia-conferencing, and entertainment software.
Various network types allow more or less efficient distributed processing, but the quest for bandwidth and short delay times remains a challenging one, especially considering the rapid increase in processing power even on reatively cheap PC's, which are often still connected over ethernet.
Even when small ethernet segments are connected with high 'lump sum' (overall) bandwidth the maximum per-node data rate is still at best a little under 1 mega byte per second, which is not too much, compared to AGP bus specs, to mention a current-day standard used for instance in high performance graphics rendering.
Upgrading to higher performance LANS is relatively costly, and will normally not approach real high overall multiprocessing bandwidth. ATM type of links offer significant bandwidth, and do allow fairly fine grained messages to be sent, but are not primarily intended for DP applications, though efficient path and channel setup facilities and flexible, preferably near non-blocking QOS resquest handling could deliver a good DP infrastructure, albeit relatively costly (I think).
Maybe dedicated (closed) local networks with limited wire span width could be appropriate, say a PCI based straightforward DMA packet switch configuration. It would be a challenge to consider and look into such questions, in a top level networking environment.
The general form to make such facilities available at session, presentation, and application layer is added challenge especially considering that even though TCP packet handling is not too far from DMA type of interconnect (except for preferred high level only, course grained reliability / error detection facilities), the common 'socket' interface has several shortcomings in that respect, even though the read/write functions passing iov structures could be used advantageously.
There lies a main technological and management challenge to give many consumers this possibility for a fair price, with sufficient flexibility, and with an interesting growth path, ranging from the choice of carrier and good use of the existing one's to the application of (tcp/ip ?) to a Residential BroadBand network that will continue to very rapidly increase in size and required bandwidth.