Computing desk
< August 6	<< Jul \| August \| Sep >>	August 8 >

Welcome to the Wikipedia Computing Reference Desk Archives
The page you are currently viewing is an archive page. While you can leave answers for any questions shown below, please ask new questions on one of the current reference desk pages.

August 7

Efficiency of algorithms across OSs and architectures

If one algorithm is more efficient than another one running on an OS and using one microchip, would it be more efficient across all other OSs and single microchips? I'm excluding concurrent/multiple microchips architectures and so on. But I'd like to know about speculative future possible chips and OSs. --Hofhof (talk) 16:09, 7 August 2017 (UTC)[reply]

It depends on the granularity of the algorithm and the functionality of the chip. If one chip is customized to do something quickly, it will be efficient at doing that where another chip might need to perform a long series of steps. Simple example of functionality difference: If I have an array of 100 numbers and I want to add 1 to all of them, I have to perform at least 100 steps with a standard Intel processor by selecting the first number, adding one to it, then selecting the next number, and adding one to it, and so on. If I was using a cell processor, I could do it in one step by selecting the first number and telling it to add one to the 100 numbers from that number. Note: Cell processors can perform the same function to multiple locations in memory at the same time, so I can add 1 to 100 memory locations all at the same time. So, with two chips that you can find right now (Intel or similar in most computers and Cell in the PS3), you can easily have an algorithm that is efficient on one and not the other. 209.149.113.5 (talk) 16:28, 7 August 2017 (UTC)[reply]

Thanks. Is this the link: Cell (microprocessor) to the correct article?Hofhof (talk) 16:42, 7 August 2017 (UTC)[reply]

That is a general description of a cell processor. The example I mentioned is an application of vector processing. If you have a chip that cannot do vector processing, you have to iterate over the vector. If you have a chip that can do vector processing, you can perform a function to the vector in one step. 209.149.113.5 (talk) 16:59, 7 August 2017 (UTC)[reply]

The simple answer is "no". Quantum computers can use algorithms that are faster than any "classical" algorithm for solving the same problem. This is an incredibly broad question. If there's a problem you're trying to solve, try asking a more specific question. --47.138.161.183 (talk) 16:32, 7 August 2017 (UTC)[reply]

@talk You are missing the point.Hofhof (talk) 16:42, 7 August 2017 (UTC)[reply]

Not necessarily. CPUs vary in how quickly they execute various operations. They vary in how fast they access memory and how much cache memory they have. Etc. Bubba73 ^{You talkin' to me?} 03:42, 8 August 2017 (UTC)[reply]

Well theoretically, changing from one concrete computing model implementation to another will usually at most involve a polynomial overhead. In practice, moving from one processor type to another usually incurs only an overhead than can be bounded by a constant factor (which may appear to be large, but compared to what is available in N or R⁺, any constant number is small ;-). Of course, if you OS implements primitives you rely on stupidly (say qsort() as bubble sort or bogosort), all bets are off. --Stephan Schulz (talk) 11:45, 13 August 2017 (UTC)[reply]

How can I contact an administrator on this particular website

http://www.yhwiki.com/index.php?title=Main_Page Please help me. 50.68.118.24 (talk) 23:33, 7 August 2017 (UTC)[reply]

Go to the bottom of this page and email the owner. Of course they haven't edited since 2016 so you may not get an answer but he did edit on 26 July as User:Admin. CambridgeBayWeather, Uqaqtuq (talk), Sunasuttuq 00:22, 8 August 2017 (UTC)[reply]