I've passed on the upgrade to x64 distros, despite every box in my inventory other than my Eee PC 900 supporting them. The trouble with binary packages etc. seemed to be substantially worse than the minor benefits at the time to making the switch. Even today, various 32-bit OSes (Linux & 'doze alike) seem to dominate both at my current and last job, and only seem to be swapped out for 64-bit OSes when dealing with obscene amounts of RAM or specialized scientific routines.
Is there something I'm missing, though, that has a dramatic impact on the desktop-user's experience? Are there substantial speed gains to be had, either on AMD or Intel Core * architectures? Is there any practical benefit at all to making the switch other than the specialized workstations and servers mentioned above?
Thanks in advance, Sean Crago Kathmandu
1. There is no such thing as "x64" 2. For most architectures, 32 vs 64 bit mostly just allows more more memory 3. For x86, 64-bit adds a number of new registers, which can improve speed. 4. For x86, you can only mix 32-bit and 64-bit apps by having two copies of your OS (32-bit and 64-bit).
So basically, to summarize: x86_32 vs x86_64 is weighing RAM vs speed. Many people prefer the RAM savings to the speed, and some have so much RAM that they opt for 64-bit. For non-x86 architectures, it is most logical to run a 64-bit kernel with 32-bit userland, and have only a few unique applications 64-bit.
On Wed, Nov 26, 2008 at 9:13 PM, Luke Dashjr luke@dashjr.org wrote:
- There is no such thing as "x64"
Nor is there an official "x86" but we don't point this out in polite company either.
- For most architectures, 32 vs 64 bit mostly just allows more more memory
- For x86, 64-bit adds a number of new registers, which can improve speed.
It also doubles the size of existing registers and memory addresses, which gets important later.
- For x86, you can only mix 32-bit and 64-bit apps by having two copies of your OS (32-bit and 64-bit).
This is hardly the only RAM problem. It turns out pointers now need twice as much RAM to store, even on uni-arch 64bit systems. This has a number of bad effects: not only do applications need more RAM, cache fills up faster, wider byte padding is necessary and the memory pipeline takes longer (no shotgunning requests as in 32bit compatibility mode). All of this has bad effects on speed.
There's also a number of handy features aside from a larger address space you've left unmentioned. NX, IP relative addressing, fast 64bit int math and standardized floating point. There's also the recalcitrance of binary software providers to rewrite their code for 64bitness. It's something of a milestone that Adobe is working on 64bit Flash.
Justin
On Thursday 27 November 2008 01:05:50 am Justin Dugger wrote:
NX,
Pretty sure NX predates, or is at least independent of, 64-bit mode.
There's also the recalcitrance of binary software providers to rewrite their code for 64bitness. It's something of a milestone that Adobe is working on 64bit Flash.
This is completely irrelevant to anyone who doesn't run proprietary software, which is the entire point of GNU/Linux.
On Thu, Nov 27, 2008 at 3:38 AM, Luke Dashjr luke@dashjr.org wrote:
On Thursday 27 November 2008 01:05:50 am Justin Dugger wrote:
NX,
Pretty sure NX predates, or is at least independent of, 64-bit mode.
There's also the recalcitrance of binary software providers to rewrite their code for 64bitness. It's something of a milestone that Adobe is working on 64bit Flash.
This is completely irrelevant to anyone who doesn't run proprietary software, which is the entire point of GNU/Linux.
No, the point is to write and run quality software; open source is the necessary condition to achieve those goals.
Justin Dugger
On Thu, Nov 27, 2008 at 3:36 PM, Justin Dugger jldugger@gmail.com wrote:
On Thu, Nov 27, 2008 at 3:38 AM, Luke Dashjr luke@dashjr.org wrote:
This is completely irrelevant to anyone who doesn't run proprietary
software,
which is the entire point of GNU/Linux.
No, the point is to write and run quality software; open source is the necessary condition to achieve those goals.
The entire point of "GNU" and "Free Software" is to be ABLE to run only non-proprietary software if one chooses. For those of a different ideological orientation, the above paragraph applies. It is a similar disagreement to the one Objectivists have with Libertarians.
On Fri, Nov 28, 2008 at 7:34 AM, Monty J. Harder mjharder@gmail.com wrote:
On Thu, Nov 27, 2008 at 3:36 PM, Justin Dugger jldugger@gmail.com wrote:
On Thu, Nov 27, 2008 at 3:38 AM, Luke Dashjr luke@dashjr.org wrote:
This is completely irrelevant to anyone who doesn't run proprietary software, which is the entire point of GNU/Linux.
No, the point is to write and run quality software; open source is the necessary condition to achieve those goals.
The entire point of "GNU" and "Free Software" is to be ABLE to run only non-proprietary software if one chooses. For those of a different ideological orientation, the above paragraph applies. It is a similar disagreement to the one Objectivists have with Libertarians.
You are neglecting the "Linux" portion of "GNU/Linux" quoted from the grandparent. Torvalds has consistently been at odds with the FSF, who view presently view liberated software as a moral objective itself rather than a means to an end -- fixing bugs. Even Stallman's original cause was to fix a bug in a printer, and he still fights to preserve the right for people to fix bugs in products they own. In fact, Stallman had to go in and add "freedom of use" later to his free software definition, because his goal is a place where programmers can Write and Fix free software, not merely Use it.
Justin Dugger
On Fri, Nov 28, 2008 at 12:14 PM, Justin Dugger jldugger@gmail.com
The entire point of "GNU" and "Free Software" is to be ABLE to run only non-proprietary software if one chooses. For those of a different ideological orientation, the above paragraph applies. It is a similar disagreement to the one Objectivists have with Libertarians.
You are neglecting the "Linux" portion of "GNU/Linux" quoted from the grandparent. Torvalds has consistently been at odds with the FSF, who
I'm not neglecting it. I guess I didn't connect the dots well enough for you.
view presently view liberated software as a moral objective itself rather than a means to an end -- fixing bugs. Even Stallman's original cause was to fix a bug in a printer, and he still fights to preserve the right for people to fix bugs in products they own. In fact, Stallman had to go in and add "freedom of use" later to his free software definition, because his goal is a place where programmers can Write and Fix free software, not merely Use it.
No, I don't have to connect the dots, because you've shown that you already have them connected. Torvalds' goal has always been good software, not the freedoms that RMS and the FSF hold so dear. He just thinks that GPL2 is a license that facilitates making good software.
It really rankles ideologial purists when someone agrees with them for the wrong reasons; they worry about those wrong reasons leading to what they consider wrong answers on other points. But occasionally, those disagreements can be used as object lessons. I think RMS really enjoyed the BitKeeper debacle, because it gave him a chance to say "I told you so."
-
Justin Dugger -
Luke Dashjr -
Monty J. Harder -
Sean Crago