Mailing List Archive
tlug.jp Mailing List tlug archive tlug Mailing List Archive
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]Re: [tlug] Only 32bits for Linux on UltraSparc II :(
- Date: Sun, 21 Aug 2005 17:58:50 +0900
- From: "Stephen J. Turnbull" <stephen@example.com>
- Subject: Re: [tlug] Only 32bits for Linux on UltraSparc II :(
- References: <875a4213050820074033cc3cda@example.com><20050820204231.5a0dc716.plate@example.com><875a421305082016131e3b299b@example.com>
- Organization: The XEmacs Project
- User-agent: Gnus/5.1007 (Gnus v5.10.7) XEmacs/21.5-b21 (corn, linux)
>>>>> "Zuco" == Zuco Pietro <zucomailinglist@example.com> writes: Zuco> There is something I don't understand. If the benefit of use Zuco> 64 bits for userland applications is so poor, why spend time Zuco> and money, compiling an entire operating system to 64 bits Zuco> as Solaris? One of the elements of the race to improve Zuco> performance is incrementing from 16, 32, 64, and so Zuco> forth. UltraSparc is useful only for large databases or Zuco> complex mathematical operations? If you think of graphics or national accounting (this definition of accounting includes Toyota and GM and Microsoft as nations :-) as complex, yes. For most purposes, it's clock speed and how much you can get into RAM (somebody else covered that, though). I don't know anybody who has more than 65535 fingers, and for most applications that's what you're doing: math that's about as complex as counting on your fingers. Think about it: how often do you use numbers bigger than 255? As we used to say in junior high, "more than a mouthful is a waste!" In fact there's a theory of the "natural distribution" of integers, and it's basically negative exponential: the probability of encountering the integer n is proportional to exp(-n) or something like that. Wide words can be used in a lot of situations. For example, if you know that a string is word-aligned, there are algorithms that allow you to detect whether there are any zero bytes in the word in two machine instructions. So these algorithms buy nothing with 16 bit words, a 2:1 speed up for 32 bit words, and 4:1 for 64 bit words, plus some overhead for handling memory alignment. That's a lot if you're doing a lot of strlen() or memcpy(). However, these situations are fairly specialized. In most cases, except for the big speedup that comes if the CPU (and your wallet) can handle enough RAM to never use swap, most advanced applications spend most of their CPU chasing pointers! Ie, the size of the address doesn't matter. Zuco> For example the X server will not be faster if it is 64 bits Zuco> compiled? With the Render extension (heavily used in Xft), the device independent portion will get faster, I suspect, and the device dependent portions will get faster if and only the graphics bus for the card can handle 64 bits. However, there's lots and lots of code that is not 64-bit safe. -- School of Systems and Information Engineering http://turnbull.sk.tsukuba.ac.jp University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN Ask not how you can "do" free software business; ask what your business can "do for" free software.
- References:
- [tlug] Only 32bits for Linux on UltraSparc II :(
- From: Zuco Pietro
- Re: [tlug] Only 32bits for Linux on UltraSparc II :(
- From: Ulrich Plate
- Re: [tlug] Only 32bits for Linux on UltraSparc II :(
- From: Zuco Pietro
Home | Main Index | Thread Index
- Prev by Date: Re: [tlug] Only 32bits for Linux on UltraSparc II :(
- Next by Date: [tlug] Yet another followup, this one on Vector Linux
- Previous by thread: Re: [tlug] Only 32bits for Linux on UltraSparc II :(
- Next by thread: [tlug] Yet another followup, this one on Vector Linux
- Index(es):
Home Page Mailing List Linux and Japan TLUG Members Links