Talk:XOP instruction set
|WikiProject Computing / Hardware||(Rated Start-class)|
An anonymous user has removed the section about market failure with the comment ("Market failure" means it's bad for all involved, including intel. That's not obviously the case.), incidentally also removing a reference. I have undone this deletion because it is based on a misunderstanding of what market failure means. Market failure does not mean that it is bad for all involved. See the definition of market failure. Afog (talk) 11:44, 29 June 2009 (UTC)
- (Sorry, didn't see this feedback at first since you didn't mention it in the edit summary) Okay. Let's compare the two scenarios. For there to be market failure, the number of people who by no CPU because of these compatibility problems must be higher than the number of people buying a CPU because of extensions that are superior because designers are not hampered by compatibility considerations. That this is the case seems far from obvious to me. Most people don't care about these extensions at all; those that do care tyically only whether their applications run better with them, and probably don't whether some other CPU that they're not going to buy also has it. --220.127.116.11 (talk) 14:54, 4 August 2009 (UTC)
- You have a market failure if the sum of disadvantages to anybody is bigger than the sum of advantages to anybody else. The classical example is that a cheap but polluting production process is an advantage to the polluting factory but a disadvantage to his neighbors (who neither buy nor sell the product). In the case of CPUs, an incompatibility may give one company a short term advantage over its competitor, but it is a disadvantage to programmers and end users as well as to the competitor. Afog (talk) 15:11, 1 December 2009 (UTC)
There's conflicting info about this instruction.
http://developer.amd.com/wordpress/media/2012/10/26568_APM_v41.pdf states it uses odd-numbered packed 16 bits integers from each source. But http://msdn.microsoft.com/en-us/library/vstudio/gg445175%28v=vs.100%29.aspx says even-numbered. Jamrial (talk) 09:06, 4 February 2014 (UTC)
- Must be some off by one counting. I think the even indexed makes the most sense, and looking at http://support.amd.com/TechDocs/43479.pdf . The wording there is similar to the AMD tech manual, but it has a great illustration that seems to argue against it. So I guess odd numbered means odd ordered numbered, 1st, 3rd, aka index 0, 2. A classic computing confusion.Carewolf (talk) 20:36, 4 February 2014 (UTC)