Talk:Run-time algorithm specialization

[Untitled]

Isn't "specialization" the right spelling for "specialisation"? -—The preceding unsigned comment was added by 201.51.8.36 (talk • contribs) .

It is the British spelling, according to the dictionary, which is equally valid on Wikipedia. -Goldom ‽‽‽ ⁂ 21:10, 24 July 2006 (UTC)

Then Wiktionary is lacking that information. - 201.51.28.56 15:01, 25 July 2006 (UTC)

Article based on isolated (and faulty) research

From my perspective, this article is merely a rehashing of one small part of one PhD student's dissertation, as opposed to an encyclopedic summary of an established topic. Furthermore, I believe that this article misrepresents partial evaluation and run-time algorithm specialization. For example, partial evaluation at run-time was done before Riazanov's work (I'm guessing he just didn't do a proper literature review?). He seems to claim in his paper that doing partial evaluation at run time is a "new thing" and thus he introduces "run-time algorithm specialization". Nothing about partial evaluation dictates when it is performed (compile-time vs. run-time), thus doing so at run-time doesn't justify a distinction. Therefore this article, based on this mistake, is likewise flawed.

There is a plethora of literature on the topic of partial evaluation at run-time, but the most condemning to this article (IMO) is Scott Draves PhD dissertation, wherein he claims (years before Riazanov):

"...this was the first research to explicitly apply partial evaluation to run-time code generation."

and references his 1995 paper ("Lightweight Languages for Interactive Graphics"). Following this paper's references (and those that reference it) will lead you to the other examples that precede Riazanov's (and others in his research group) work.

Clearly doing partial evaluation at run-time does not deserve the distinction of a new name, nor a new topic. But the root of the problem here is that this article is based on the isolated work of a small group of people instead of a mature field from an established community. The faulty information is only a side-effect of this mistake.

I would like to recommend this article for deletion.

Dan 23:32, 6 November 2007 (UTC)

false claim: this idea is from the 1980ies, named "customization" back then.

The article states a false claim: creating specialized code variants for costly computation tasks at run-time has been first exercised in the late 1980ies by object-oriented compilers. E.g. see publication list of http://www.cs.ucsb.edu/~urs/oocsb/self/papers/papers.html, the methodology on which later the Java Hotspot virtual machine was based. In this context, the methodology was called "customization". This is much earlier than the literature cited in the article, from the late 1990ies even to 2000s. — Preceding unsigned comment added by 2001:4CA0:0:F240:917F:F701:C14D:5C25 (talk) 13:37, 22 January 2015 (UTC)