| 1 |
On Mon, 20 Sep 2004 14:25:53 -0400 "David F. Newman" <dnewman@×××××.com> |
| 2 |
wrote: |
| 3 |
| No, I have no reason aside from pure geek value. I was under the, |
| 4 |
| probably false, impression that sparc64 executables could move data |
| 5 |
| around faster. i.e. doubles could be loaded in fewer clock cycles. |
| 6 |
| So I have 2 questions that maybe you can answer, |
| 7 |
| |
| 8 |
| 1) Is the only advantage to a sparc64 executable access to memory |
| 9 |
| larger than 2G per process? |
| 10 |
|
| 11 |
Pretty much. |
| 12 |
|
| 13 |
| 2) Assuming yes to question 1, then what is with the hype with 64bit |
| 14 |
| processors? |
| 15 |
|
| 16 |
x86/amd64 isn't like sparc/sparc64. amd64 CPUs are fast because they |
| 17 |
have larger cache and a whole bunch of extra registers, not because |
| 18 |
they're 64bit. On sparc, we can gain pretty much all the advantages of a |
| 19 |
v9 CPU whilst still running a 32bit userland. |
| 20 |
|
| 21 |
But then, "four times as many general purpose registers!" just doesn't |
| 22 |
sound as cool as "64 bit!". |
| 23 |
|
| 24 |
-- |
| 25 |
Ciaran McCreesh : Gentoo Developer (Sparc, MIPS, Vim, Fluxbox) |
| 26 |
Mail : ciaranm at gentoo.org |
| 27 |
Web : http://dev.gentoo.org/~ciaranm |
Archives