9,762 reputation
12143
bio website en.wikipedia.org/wiki/…
location United States
age 70
visits member for 4 years, 8 months
seen 16 hours ago

BS Mechanical Engr.
PhD CS(AI)
CS Prof (4yr)
Numerous consulting jobs.
15 yr at http://www.pharsight.com
Published book on CS & several articles
4 kids, 2 grand
Pilot(student)

P.S. The picture is a Beta-prime distribution. It shows the program speedup factors you can get if you see a problem twice in 2, 3, 4, and 5 samples.


Apr
23
comment Why do we need “callback functions”?
+ That's what callbacks are good for, but I hate having to write them :-)
Apr
16
comment Implicit optimization versus explicit optimization
+ Library routines should not be sluggards, but their main purpose is ease of use, correctness, and reliability. I discovered using LAPACK that it was actually not very efficient in the case of small matrices, but it still serves its purpose.
Apr
12
comment What to do when your colleagues don't value code maintainability
I saw things like that during my contracting days. It seems to me the real problem is the high turnover, and that is caused by a management style, starting from the top, that doesn't actually value its people. I could always tell, just by sizing up the CEO, whether this would be a good company to work for.
Apr
4
comment “Too object-oriented”
@ThuneGrill: Karl's right. Stick to pragmatic reasons, not religious ones. OOP is certainly a good idea, but I've seen it carried to ridiculous extremes. The result is making mountains out of molehills. Things that could be done in 1000 lines of code end up being 10,000 lines of code with classes galore. Then, Gee, it's hard to maintain, and the performance sucks. (No matter what collection classes get used.)
Apr
2
comment Performance overhead of standard containers and boost
Here's what astounds me. The problem with all such recommendations is they are flying blind, like a plane in the clouds with no instruments. Just having wings and a motor won't get you where you want to go, without guidance. Sure it might be that new & delete are costly, but just avoiding those won't make the code fast. You need to know how to find out what to fix. Sure - profile - everybody says. They say it, but they don't do it, and it doesn't work very well anyway. Here's what works.
Mar
9
comment Connection between programming language and solutions
+ At least for making me look up Sapir-Whorf :)
Feb
14
comment A question regarding linked list vs vector insert/remove results comparison
+ Right. I hate it when I hear blanket, categorical assertions like "use lists" or "use arrays". It's a way of putting part of your brain on a vacation/holiday. I let the program itself tell me what's taking the most time, as in this example.
Feb
11
comment Why is the main memory for object allocation called the 'heap'?
@Robert is right, and don't forget, the folks inventing these terms were just ordinary slobs like us. So what do you call a pile of stuff that has no particular order to it? Well, "heap" is one thing. (I can think of others :)
Feb
11
comment Usefulness of pre and post order traversal of binary trees
@Kilian: sdg is right. With inorder, you have to be concerned with precedence, unless you put parentheses around everything.
Feb
4
comment How to best approach planning and budgeting of software performance optimization efforts?
@Paul: And these people are good. It's just the team dynamics and how programmers are taught. I often wish there could be a revolving-door situation between academia and industry, just to keep the profs (I was one) from teaching idiotic stuff (like gprof :) There's tons more wisdom in the pages of stackexchange.
Feb
3
comment How to profile a Perl + FastCGi + StarRunner web program?
You're right. 20 years ago I worked on a factory-floor system with multiple machines TTY-ing each other. It wasn't easy, but I managed to get time-stamped logs, from which I could trace the history of individual message flows across machines/processes. It wasn't fun, but I did manage to dig out some tumors. Glad you found something that works.
Feb
2
comment How to profile a Perl + FastCGi + StarRunner web program?
Hi @Paul. This is a case where, in general, my favorite method only takes you part way, because you need visibility into round-trips between server and client, although here's a case where it was effective, even with a protocol in the middle.
Jan
29
comment How can I plan optimization tasks in Scrum?
Possibly helpful ...
Jan
23
comment What performance topics should I be aware of?
That's my point. A lot of people think that, and they blissfully assume the other 99% is as optimal as it could possibly be. That's the issue.
Jan
23
comment What performance topics should I be aware of?
You mentioned caching, aliasing, branching, loop unrolling, etc. That's the last 1% of the job. The lion's share is not in things you know ahead of time, it's in things you discover as you work.
Jan
23
comment What performance topics should I be aware of?
There's what you do before you write it, and what you do after. What you do before is KEEP IT SIMPLE, especially in the class and data structure, so you avoid putting in performance problems. What you do after is TAKE OUT THE PROBLEMS YOU DID PUT IN, and be willing to do large re-writes. It's possible to reduce time by a factor of 10-100. gnat's link gives an example.
Dec
15
comment Is there any evidence that lisp actually is better than other languages at artificial intelligence?
@Giorgio: The ancestors and cousins of Prolog are mostly written in what? Lisp. Fischer Black's deduction program, Micro-Planner, etc. For me, a good book on this is Deductive Foundations of Computer Programming.
Dec
7
comment KISS principle applied to programming language design?
I would just compare C vs. PL/1
Nov
23
comment When does “proper” programming no longer matter?
Since performance is your concern, you should be asking the program what you should be looking at, not asking if your preconceived question makes a difference. Please click on this link, and do what it says. It will tell you what the program is really spending time at, and that will tell you how to make it faster.
Oct
2
comment What causes bad performance in consumer apps?
@Crash: If you're still interested, here's probably the most math-heavy explanation I've put together.