64-bit Geocaching.com Servers

[Lt. Sniper]

By the looks of it they are now running 64-bit servers, probably on an AMD Opteron platform (not the reason for the error and stack trace, a missing picture caused it)

Part of a stack trace related to a missing picture:

Code: Select all

[ImageNotFoundException: This image does not exist in the system.]
Groundspeak.Web.Gallery.GalleryImage..ctor(Int64 ID, ImageDataSources DataSource) +276
Groundspeak.Web.CustomWpt.TravelBug.set_ImageID(Int64 Value) +77

My next machine to replace this one will be a dual 244 Opteron system
(Only true hardware nuts would understand this...)

[Aushiker]

(Only true hardware nuts would understand this...)

<p>
Yep!
<p>
Andrew

[The Garner Family]

Some information on the platform they are running on <a href="http://uptime.netcraft.com/up/graph/?ho ... m">here</a>.

[ideology]

time since reboot: 46 days?! yikes!
maybe we should reboot ours?

(touch wood!)

[Mind Socket]

I can see references to 64-bit integers, which as I understand is just a data type, but what's the indication that they have 64 bit servers?

- R

[EcoTeam]

I can see references to 64-bit integers, which as I understand is just a data type, but what's the indication that they have 64 bit servers?
- R

You can run 64 bit (or 128bit, or 256bit, or...) words on an 8 bit processor if you want to. Data types and processor word sizes are mutually exclusive.
Data types greater than 32bits are only useful for encryption and other very specialised tasks. Most applications and number crunching are performed with 32bit or less words, so a "64 bit" processor won't execute those instructions any quicker.

EcoDave

[Mind Socket]

Yes, therefore my question ... how did the honourable lieutenant infer the presence of a 64-bit server based on unrelated information about an unsigned variable that goes up to 18446744073709551615?

As I understand, A 64-bit instruction set could potentially process < 64-bit data faster. It is the instructional processing capacity of the hardware that's increased, not so much related to the data it receives. For example, on an 8 bit machine, it might take several instructions to do a particular operation on an 8 bit number (for simplicity's sake). On a 64-bit architecture, that operation may be covered by a single instruction (potentially using fewer clock cycles).

- R

[CraigRat]

time since reboot: 46 days?! yikes!
maybe we should reboot ours?

(touch wood!)

I think the OS's might have something to do with that....

46days is good for a Win/IIS based system

/linux zealot

[Mind Socket]

I'd agree, and add that the generic humanoid carbon units involved are probably a big factor also.

... also a linux zealot, but with terrible uptime due to a dodgy HDD (I think).

- R

[EcoTeam]

Yes, therefore my question ... how did the honourable lieutenant infer the presence of a 64-bit server based on unrelated information about an unsigned variable that goes up to 18446744073709551615?

As I understand, A 64-bit instruction set could potentially process < 64-bit data faster. It is the instructional processing capacity of the hardware that's increased, not so much related to the data it receives. For example, on an 8 bit machine, it might take several instructions to do a particular operation on an 8 bit number (for simplicity's sake). On a 64-bit architecture, that operation may be covered by a single instruction (potentially using fewer clock cycles).

- R

Integers are signed by definition, so that's +n/2 and -(n/2)-1)

A 64bit processor will not nessesarily process 8 x 8 bit instructions at once, that requires parrallel processing and pipelining architectures for every combination up to 64 bits, eg. 8 x 8bit, 4 x 16 bits, 2 x 32bits
That is an incredible amount of ALU complexity, and I am not aware of a procesor which does this for smaller data word sizes. Modern processor since the Pentium however include "parrallel" procesing for the maximum word size. But that won't speed up the smaller words size instructions and data.
Other technologies like pipelining and branch prediction allow modern processors to process several instructions per clock cycle

EcoDave

[EcoTeam]

time since reboot: 46 days?! yikes!
maybe we should reboot ours?

(touch wood!)

I think the OS's might have something to do with that....

46days is good for a Win/IIS based system

/linux zealot

Can't beat DOS. I know systems that have stayed up for almost a decade. They only die when the power supply or hard drive eventually fails

EcoDave

[Lt. Sniper]

Now whos the Uber geek? Not me says Snipie!!

Anyway , I dont know for sure, What I do know is W2K2S runs on 32bit (x86, IA-32) and 64bit systems(x86, IA-64 and Alpha and maybe an Xeon system with 64-bit addressing) and they would really be wasting there processing resources to be calling 64 bit data types on a 32bit system which would have to go through an emulation of somekind to work (You would know more about that then me!!!).

The main reason for moving a server to a 64-bit platform is memory, 64-bit memory addressing allows access to Terabytes of memory. A any large database would gain significantly if it had more RAM installed as more data could be cached in memory which is accessed alot faster then directly from disk.

[Mind Socket]

I think there's still some confusion. A 32/64-bit architecture refers to the size of the instruction set on the CPU. It is independent of the memory address space and the data sizes supported. That is, I believe a 32-bit Wintel box can support 64-bit memory addressing, but as you say, there is probably some trickery involved, as with large data types. The int64 data type is a .net (software level) idea. How it is implemented on different platforms can differ.

I'm no expert on this stuff, and hopefully my understanding is correct.

Regarding Dave's signed int comment, my bad, I misread the data type definition. Even google can't compensate for my skim-reading

Whew, beer o'clock!

- Rog

[Lt. Sniper]

But 32-bit systems have physical memory addressing limits, which are 4 gigs, more can be accessed but you go into some heavy-duty manipulation.

64-bit systems allow access to something like ridiculously large amount of memory, which can be accessed easily without any special manipulations.