Thanks to Knuth SKA's Merging Stage 1 is 90% faster

**scotspine** · 20 January 2006, 18:44

good for you atw. we knew you'd get there. you can give me a hand with some reed-muller and binary hamming stuff when you've fnished.

**xoggoth** · 20 January 2006, 18:56

Hmmm 10 times faster? Seen people say that before about assembler coding the most used functions, speed up was nothing like. (Though original code Was in C)

**OrangeHopper** · 20 January 2006, 19:15

What's assembler?

**AtW** · 20 January 2006, 19:36

Not sure what you on about scots...

xoggoth - 10 times faster is my target since I expect to keep most of stuff in registers.

**scotspine** · 20 January 2006, 19:37

ya, oh! as if you wouldn't know! can i tell you that it is a new form of the emerging nymph?

**Fungus** · 20 January 2006, 19:41

Originally posted by AtW

I doubt many of you even opened his awesome book "Art of Computer Programming", yet alone understood and applied successfully some of the algorithms from it.

Come to think of it most of you (especially SandyDown) deal with primitive problems that can be done using VB's of this world. I mean, how many of you worked on tasks that require efficient sorting of a few billions of numbers?

Next week I will be using assembly language for the first time in many years to optimise yet another part of SKA - I am smiling at the pleasure I will have making a few functions run 10 times faster

What a big boy you are.

Can we stop grovelling now or would you like us miserable and unworthy thickos to genuflect in humber admiration until our pathetic knees drop off?

What a wanker!

Psion re-wrote the word application on the Series 3 in assembler language but the gain in performance was minimal. Previously it was in an object oriented version of C.

I'm spending my time rewriting unintelligible code written by someone who thinks that clarity means "resembling a wine from the Bordeaux region of France". Why oh why does he have to do everything the hard way, and and why can't he learn what threads and sychronisation objects are. It would be nice if he serialised access to resources used by multiple threads. Sigh. At least I get to have a £200,000 3G protocol simulator on my desk. Power crazed, me? Naahhh. Not much.

Fungus

**scotspine** · 20 January 2006, 19:46

good for you fungus. we knew you'd get there. you can give me a hand with some reed-muller and binary hamming stuff when you've fnished.

**Fungus** · 20 January 2006, 19:48

Originally posted by scotspine

good for you fungus. we knew you'd get there. you can give me a hand with some reed-muller and binary hamming stuff when you've fnished.

Come again?

**AtW** · 20 January 2006, 19:54

Originally posted by Fungus

Psion re-wrote the word application on the Series 3 in assembler language but the gain in performance was minimal. Previously it was in an object oriented version of C.

Looks like whoever did that was incompetent, since you know internal detail so well I presume you were involved?

I will report back performance improvement that I have achieved by rewriting selected functions in high-performance x86 assembly

Thanks to Knuth SKA's Merging Stage 1 is 90% faster