[Coco] trig.h

Steven Hirsch snhirsch at gmail.com
Tue Jan 11 18:34:42 EST 2011


On Tue, 11 Jan 2011, Willard Goosey wrote:

> On Mon, Jan 10, 2011 at 08:48:37PM -0500, gene heskett wrote:
>> On Monday, January 10, 2011 08:47:19 pm Willard Goosey did opine:
>
>>> That's just not something talked about much in modern
>>> "abstract the problem away" CS courses.
>>
>> To the detriment of the students acquired education if I can opine on that.
>
> Oh, yes.  I really would have liked a little more low-level "this is
> how the code works" and a little less Top-Down Design.

University of Vermont (my alma-mater) dropped Assembly Language and 
Machine Organization as a required course in the CS curriculum.  Just 
incredible.  No wonder we have applications that occupy half the disk and 
run like a snail.  Conventional programming course wisdom is "..don't fuss 
with the code structure so much, let the compiler optimize it".

Speaking as one who develops VLSI design-verification tools, I can assure 
you that when dealing with 100+ GB in-memory data sets (no, I'm not 
kidding) it VERY MUCH matters how you form your C/C++ code.  The way 
variables are passed can cause all manner of detrimental side-effects by 
forcing TLB and cache misses.  Do it 100 times and the 1-2usec. 
differences are livable.  Do it in one hundred places several billion 
times and things add up quickly.  Point is, unless you have a sense of 
what the CPU is doing at the register and address level you're not going 
to know how to code such things defensively.

> The worst was a database theory class I took.  Half the class was
> designing databases in some complicated model that can't even be
> implemented (that we had to transform into a relational database when
> we went to actually code it in), and the other half the class was the
> professor whining that he can't change the hardware sector size on
> modern hard-drives to match his database's record-size. :-(

I suppose I was fortunate.  The instructor for my graduate-level database 
course (at RPI) had years of practical experience and had recently 
consulted on a re-write of the Connnecticut motor-vehicles database 
system.  It was a perfect balance of set-theory and nuts-and-bolts that 
has stood me in good stead.

Steve


-- 




More information about the Coco mailing list