Impossible
Mar. 12th, 2005 08:23 amTen things which make my job impossible:
- the basic concept of converting a complicated serial algorithm (which nobody understands well) to run in parallel is flawed
- none of the people who originally wrote the serial code have been at The Company in years
- in fact due to massive programmer rotation there are several layers of coding styles involved
- there is actually one coding style-aspect which is persistent: poor to non-existent documentation
- Adil and I are the only ones who has any motivation to keep parallel abstraction working
- parallel code is, by its very nature, fragile, and developers who don't have any motivation to keep it working are going to break it
- Jeff says "zeroth order" too often, making me want to slap him
- generally Jeff opens his mouth far too often
- the customers have unrealistic expectations about what parallel will actually do for timing and abstraction
- and ... most of all BASIC ASSUMPTIONS THAT THEY MADE AND TOLD ME ALL ALONG TURNED OUT TO BE TOTALLY FALSE YESTERDAY.