Architecture Life

2 Responses on What knowledge is essential to work in computer architecture field?

• 
  By amansscientiae at June 29th, 2008

  My best guess would be that you need to write a really great thesis on computer architecture if you want to play with the big boys. At least that could be your entry ticket. You will still have to deal with the politics to get ahead in corporations like that, but you can figure that part out once you get trough the interview process and get a job offer.

  For now, I would think that you need to work out something with your adviser that is bleeding edge AND that you can complete successfully (failing on a bleeding edge topic is still failing).

  I had a really interesting experience at the Design Automation Conference, once. There was this student who had done something really amazing on VLSI design tools. While working with the usual workstation tools he got fed up with how slow the display was when he zoomed to the whole chip level (this is no wonder, a typical VLSI chip display representation can consist of tens of millions of polygons). So he went back to what he knew about computer games and 3d scene graphs.

  In 3d games one can not update all the scene all the time, it would use way too much rendering time. So the games render parts of the scene at different zoom level and cache them as textures in a tree. When the player moves in the scene, most of the time a pre-rendered 2d texture is already available which can be distorted and mapped on the parts of the scene that have just become visible. By keeping enough pre-rendered textures available for the low-overhead projection engine to map, the game can be made orders of magnitude faster (this also explains why early 3d games like Doom liked to be in a maze… that way the player has very limited choices where to go and what to look at… thus easing the computational complexity for the game engine :-))

  So this student had taken a copy of Magic (an open source VLSI viewer/editor) and added this type of texture mapping to it (there is plenty of open source for that, too). He then demonstrated the difference during his talk: plain Vanilla loaded a file in fifteen seconds but zooming in and out of the chip took between a fraction of a second and approx. 10s.

  In comparison, his precalculated texture mapping version needed approx. 2 minutes at the beginning of the program to calculate all views of the chip at once and store them as compressed bitmaps. Once it was done… he could “fly through” the chip in real time as if it was a first person shooter game!

  When he did that, the audience went “ahhh” and “ohhhh” and there were standing ovations. After the talk I saw the plenary speaker from Cadence run to the front to talk to the young man. I can only assume that he did not have a problem finding a job after he had finished his thesis.

  I suppose you are more than smart enough to understand what my little real world story has to do with your desire…

  Good luck!

• 
  By rckrllrfg at July 1st, 2008

  To be honest with you, going to work for the “big boys” is simply “hype,” and I recommend that you don’t get all starry eyed over them. Think beyond the logos and the name recognition. These are mega size corporations with thousands of people who have built their bureaucracies around their founder(s)’s concepts and dreams. Even if you had ideas for THE next major breakthrough for the microprocessor, chances are that you will not be able to get the attention of the powers-that-be to get it implemented. You will quickly learn that there is a pecking order, office politics drive a great deal of decisions, and that free thinkers are generally not appreciated.

  You should still apply for these companies and they are great for a college grad since they provide a stable work environment; however, you may quickly find yourself restless and bored, especially if you do have a true desire to explore and invent. To be brutally honest, only 10% of the engineering comminuty gets to design. The rest of the engineers impliment it.