Date   : Thu, 04 Nov 2010 02:56:39 +0100
From   : rick@... (Rick Murray)
Subject: No wonder CompSci graduates are unemployed

On 03/11/2010 20:50, Mark Usher wrote:

> You have to remember also, we are lucky that we had our minds trained
> with problems at those levels and honed our own skills

Fair point. Though you would have thought a waltz around the insides of 
an 8 bit micro would be a required part of CompSci ... 8 bit to be 
powerful enough to actually work, as opposed to fake "this is what a CPU 
looks like" exercises; and 8 bit so as to be simple enough to be easily 
grasped without obfuscation. I'd like a CompSci course that introduced 
me to the ARM, but before you can do diddly on a modern ARM there's a 
load of hoops to jump through.
Any older chip - 6502, Z80, that sort of thing, runs a basic flat 
addressing scheme that is really easy to grasp, doesn't require address 
translation, paging, or segmentation. And in the case of the 6502, if 
you assemble code with the three vectors defined, getting going is not 
that much more than:
   SEI
   CLD
   LDX 255
   TXS
   ...poke hardware, etc...
   CLI
   ...app code...

Try *that* on the TMS320DM320 I mentioned in another thread. [hint, the 
basic system init code is jtag_start.s (2.3K) and platform.s (11.9K), 
and that is only setting up basic memory mapping and the PLLs....


> Ask us to fix a mainframe problem and well, we might start scratching
> our heads too ;-)

<scratch><scratch> You guys DO know my iPhone has more processing power 
than this heap? <scratch>

:-)


> These days they aren't usually taught that sort of thing, as what relevance
> does it actually have in the market place.

How can you claim to be a programmer if you don't *understand* what your 
code does? Some grounding in the hardware would explain why a program 
written for an Acorn/ARM which uses int all over would fail badly when 
ported to an x86 machine. Because as ARM registers are 32 bit, int=long, 
but historically this wasn't the case for x86 for int!=long.

Not to mention code that does fancy things to bitfields. Will that code 
still work if the orientation of the numbers is different (endian issues)?


We have, at work, a quality control girl who just sorta rolled out of 
some university or other with a fancy-ass bit of paper. She has poor 
people skills and seems to annoy half the people she deals with (I, 
myself, am dangerously close to telling her something extremely rude). 
She may have been top of the class, but here everybody must do their 
part or it gets messy. I'm not going to say any more, except to say that 
a qualification doesn't make you any use. Its when you have the hands on 
that you become useful. Its when you leap into a problem and come up 
with realistic ideas (instead of hiding, passing the buck, or wild 
nonsensical ideas) that you become useful. This depends on the 
personality, and it depends on their confidence in what they were 
taught. A piece of paper doesn't necessarily reflect this.


Best wishes,

Rick.

-- 
Rick Murray, eeePC901 & ADSL WiFI'd into it, all ETLAs!
BBC B: DNFS, 2 x 5.25" floppies, EPROM prog, Acorn TTX
E01S FileStore, A3000/A5000/RiscPC/various PCs/blahblah...