|
|
|
| Authored by: Anonymous on Monday, October 15 2012 @ 05:59 PM EDT |
Eliza is exactly the type of software that POIR's article describes.
The problem with the article is that in POIR's world, computers do what they are
supposed to. You run the same algorithm over and over on the same data and
always get the same results.
I used to live in that world. Now I've seen too many interesting things go
wrong.
Like every Intel P4 class processor I tested could potentially drop a bit in the
address calculations. Which bit got dropped seemed to depend on the exact type
of processor, but I didn't do enough testing to consider the result
statistically valid.
The fun part was that the Linux 2.4 kernel memory management code was really
good at tweaking the bug. In the end, I advised the client to switch to Linux
2.6. It wouldn't fix the problem, but it would turn it into noise.
[ Reply to This | Parent | # ]
|
- Another angle? - Authored by: Anonymous on Monday, October 15 2012 @ 07:12 PM EDT
- Another angle? - Authored by: PolR on Monday, October 15 2012 @ 07:25 PM EDT
- patentable? - Authored by: Anonymous on Monday, October 15 2012 @ 07:55 PM EDT
- Another angle? - Authored by: Anonymous on Monday, October 15 2012 @ 08:01 PM EDT
- Another angle? - Authored by: PolR on Monday, October 15 2012 @ 08:13 PM EDT
- Another angle? - Authored by: Anonymous on Monday, October 15 2012 @ 08:58 PM EDT
- Another angle? - Authored by: PolR on Monday, October 15 2012 @ 09:21 PM EDT
- Another angle? - Authored by: Anonymous on Monday, October 15 2012 @ 09:34 PM EDT
- What is math - Authored by: Anonymous on Wednesday, October 17 2012 @ 01:56 PM EDT
|
|
|
|
