Wikipedia:Peer review/Fast inverse square root/archive1

Fast inverse square root

 * A script has been used to generate a semi-automated review of the article for issues relating to grammar and house style; it can be found on the automated peer review page for March 2009.
 * A script has been used to generate a semi-automated review of the article for issues relating to grammar and house style; it can be found on the automated peer review page for March 2009.

This peer review discussion has been closed. I've listed this article for peer review because I'm hoping to send it to FAC. It may need some work clarifying the "proof" section--I'm kind of stalled on doing that right now. This article received some ce help from others and a good GA review, but I feel it needs some close attention from an expert or interested party. It's neither strictly mathematics nor computer gaming, so I placed it here.

Thanks, Protonk (talk) 17:47, 21 March 2009 (UTC)
 * As an addendum, I know the sourcing is borderline, but I think that a cursory review of the sources advanced will show that they are reasonable for the subject matter. Protonk (talk) 17:49, 21 March 2009 (UTC)
 * Specifically, Lomont, Eberly and McEniry. Eberly, interestingly enough, has the worst analysis of the code but is the only SPS (AFAIK)--he has a book published on 3-D computer graphics (which is also references, completely by accident, actually).  Lomont's paper is referenced a half dozen times, in scholarly papers (one of which is cited in this article) and on most of the web articles.  McEniry's paper is later, so it missed the craze, but I am willing to stand by it as a good source.  It is web available and it is easily the deepest look of the bunch. Protonk (talk) 18:24, 21 March 2009 (UTC)

Ruhrfisch comments: I will be the first to admit that I do not understand much of the article, so take my suggestions with a grain (or shaker) or salt. Here are some suggestions for improvement. Hope this helps. If my comments are useful, please consider peer reviewing an article, especially one at Peer review/backlog (which is how I found this article). Yours, Ruhrfisch &gt;&lt;&gt; &deg; &deg; 20:35, 31 March 2009 (UTC)
 * I would try to provide context to the reader wherever possible. Some places do this well already, but just in the first sentence "(IEEE 754-2008)" is a complete mystery without some context - see a method of calculating the reciprocal of a square root for a 32-bit floating point number (IEEE 754-2008). I know it is linked, but could it be a method of calculating the reciprocal of a square root for a 32-bit floating point number (the IEEE Standard for Floating-Point Arithmetic, or IEEE 754-2008).
 * I would link 32-bit, perhaps in the second paragraph of the lead (I know it is also in the first paragraph)
 * Would this sentence read better as Next, a logical shift right of one bit is performed and the result [is] subtracted from the "magic" [hexadecimal] value 0x5f3759df. I know 0x is linked, but most people have trouble understanding binary and do not know or recognize hex
 * I am not sure the casual reader would know that "the code" in  Specifically, when the code was developed in the early 1990s, most floating point processing power lagged behind the speed of integer processing.[1] refers to the Fast inverse square root algorithm.
 * I am unsure about refs in math / computer articles - my normal rule of thumb is that every quote, every statistic, every extraordinary claim and every paragraph needs a ref, but when there are many one sentence paragraphs with equations, I am not sure this applies. Still, the article feels under-referenced to me. For example the whole part in "A need for fast..." from the beginning "To normalize a vector,..." to the end "...the inverse square root of the distance components." seems like it should have a ref. Or these sentences later in the Overview section This integer subtraction and bit shift results in an integer number which when converted back into floating point notation is a rough approximation for the inverse square root of the input number. One iteration of Newton's method is performed to gain some precision, and the code is finished.
 * Since this is taking a square root (albeit an inverse one) is "division" correct in This bit of code proved faster than table lookups and approximately four times faster than regular floating point division.[4]?
 * Why the use of quotes in The advantages in speed offered by the "fast inverse square root" hack came from converting the floating ...? The term is not in quotes elsewhere.
 * My eyes glazed over from the Magic number section down to History, so I am skipping over them here too - sorry. I do note that File:Float example.svg uses "fraction" where the article uses "significand", which is confusing.
 * In History I have no sense of time - could years be added? Also the chronology seems to go backwards (most recent to most distant in the past) which can be a bit confusing, although I understand this is how the chronolgy was discovered (I think)
 * Headers do not all follow WP:HEAD - use of "a" and "the" for example.