I'm pretty sure that these days, at least on x86, data is aligned to 4-byte boundaries. Some architectures (I'm pretty sure that x86 is this way) require a 4-byte aligned address as the parameter to the 4-byte memory load instruction).
Given that memory access times are bound by latency far more than bandwidth, the effect of loading another four bytes into the register file is most likely insignificantly small. I'm certain that 8-byte register-to-register operations *are* insignificantly small, and it's likely that pointers, given that they are not large but often accessed would be kept in registers. It would depend highly on the particular architecture.
Sure there's a benefit: it's cool!
After I found out how/insanely/ easy it is to port things to IPv6, I added IPv6 addresses on all the machines in my home network. It's really not that much work, you just add the IP with ifconfig like anything else. Now, home networks are easy because there's no complicated switches needed(we have a hub, which doesn't really know about IP as far as I can tell given that it works fine with IPv6 even though the hub dates from when 10baseT hubs were expensive and cool).
The key difference is that the opteron is/faster/ in 64-bit mode. There are more registers. The same is not true of a G5. OSX doesn't "break up" instructions - instructions are instructions. OSes don't interpret each one. 32-bit windows on opteron just uses 32-bit instructions, just like 32-bit darwin on a G5. Photoshop is not a 64-bit app in the case of OS X IIRC - it uses a 64-bit math library as its G5 optimization. This is fine and works. As I mentioned in another post, darwin can't run "64-bit apps" yet, because there are no 64-bit interfaces to system calls (think about it - if the kernel expects a 32-bit FILE * and you send it a 64-bit one, you're going to have trouble doing I/O). I think I missed something with what you said there. The main 64-bit part of darwin is the math library since they can throw some 64-bit ASM in there plus code to convert back-and-forth to the 32-bit bindings.
I'm geniunely curious - how? A 64-bit application can't do much without 64-bit system calls. Does mach-o let you dynamically load 64-bit code in a 32-bit program (all the 64-bit code would be able to do is computation, given the lack of system calls). Or is there a windows64-on-windows-like 64-bit wrapper over libc?
I'm sorry, but doesn't everyone know by now that the opteron is tangibly faster in x86-64 mode due to a doubling of GPRs in a register-starved ISA? Besides, I suggested running 64-bit linux on both, just to be supremely fair, but I think that it wouldn't be much different from 64-bit amd64 and 32-bit darwin.
That since they are running the Opteron in 32-bit mode, it's not taking advantage of it's full potential. Guess we'll wait until "round 2" like he says, but it still looks bad that he kind of dodges this. If it were me I'd be running the benchmarks on 64-bit linux versus 64-bit linux.(gentoo?)
I don't understand. Why buy G5 xserves when you can get the G5 in blades from IBM? Rack space is expensive, and Apple's only real renderfarm asset (the CPU) is actually IBM's asset...
The card's performance, drivers etc sucked, but one time I put it into the AGP slot and sparks flew, literally ( a bolt of electricity jumped from end to end of the slot ). Smoke rose. Powered the thing up and everything worked fine.
Actually, the amount of cache coherency checking needed in a large shared memory machine drives the latency right up. Besides, myrinet is ~15 usec latency. That's pretty goot for most things.
Actually the protocol wouldn't have to be changed much at all. I've investigated this - the alpha value can be passed in the word-alignment byte. If you load up transluxent and throw real alpha values into your XImage, then it will be real alpha. It's just that regular XFree86 ignores the word alignment byte, and transluxent just passes it on to OpenGL so it actually works fine. Unfortunately transluxent isn't very stable.
5.1 works with SB Live! cards under linux perfectly. Even the value ones if you make an RCA connector and attach it to the pins on the card (or wimp out and buy the $60 attachment).
The only problem I've had is in finding applications that do AC3 passthrough right - I've had success with VLC.
Actually I always send the info, because then IE pops up with current tracking information on the bug, and things like "this problem is fixed with x patch. would you like to download x patch now?" It's really a nice system, to tell you the truth.
Exactly. The coolest part about TK (well, at least tkinter for python) is that you can "connect" a widget and a variable - someone edits the widget, and the variable changes value, automatically. Saves a LOT of code that would otherwise listen for event -> get widget value -> set variable value.
Isn't it a property of these kinds of things that you can seperate two electrons (or some subatomic particle, can't remember) and change one's spin, and the other, no matter how far away, will instantly change? I recall an experiment in which this worked over a distance of six miles. Wouldn't this be the perfect interconnect? No wires at all?
Flawed business model? Relative to what - SCO's ingenious strategy of using rediculous claims of IP infringement to pump and dump their shares while refusing to publicly disclose what the IP infringement actually is? Yup - SCO knows all about flawed business models.
By definition, this isn't a flawed business model. SCO is making incredible amounts of cash. It's unethical, but since when did big businessmen care about ethics when they have a money printing press like this?
I think that he was referring to rpm -q. You need rpm -qp to query a package file, but only rpm -q to query an installed package.
Also, for those of us to shun nautilus and heavy GUIs in general, his suggestions to tar make sense. Are those who don't use KDE-GNOME not entitled to be annoyed by some of the GNU tools?
Slashdot Mirror
You could also leave your DNS static and use EC2 Elastic IPs to shift things around on the backend (you did mention you were using EC2).
I'm pretty sure that these days, at least on x86, data is aligned to 4-byte boundaries. Some architectures (I'm pretty sure that x86 is this way) require a 4-byte aligned address as the parameter to the 4-byte memory load instruction).
Given that memory access times are bound by latency far more than bandwidth, the effect of loading another four bytes into the register file is most likely insignificantly small. I'm certain that 8-byte register-to-register operations *are* insignificantly small, and it's likely that pointers, given that they are not large but often accessed would be kept in registers. It would depend highly on the particular architecture.
Oh, and I'll also note that the IPv6 addresses I use are /shorter/ than IPv4 addresses - fec0::1, fec0::2, etc.
Sure there's a benefit: it's cool! After I found out how /insanely/ easy it is to port things to IPv6, I added IPv6 addresses on all the machines in my home network. It's really not that much work, you just add the IP with ifconfig like anything else. Now, home networks are easy because there's no complicated switches needed(we have a hub, which doesn't really know about IP as far as I can tell given that it works fine with IPv6 even though the hub dates from when 10baseT hubs were expensive and cool).
Right. My bad. I submit read/write/send/recv/open as examples of places where applications pass pointers to kernel space.
The key difference is that the opteron is /faster/ in 64-bit mode. There are more registers. The same is not true of a G5. OSX doesn't "break up" instructions - instructions are instructions. OSes don't interpret each one. 32-bit windows on opteron just uses 32-bit instructions, just like 32-bit darwin on a G5. Photoshop is not a 64-bit app in the case of OS X IIRC - it uses a 64-bit math library as its G5 optimization. This is fine and works. As I mentioned in another post, darwin can't run "64-bit apps" yet, because there are no 64-bit interfaces to system calls (think about it - if the kernel expects a 32-bit FILE * and you send it a 64-bit one, you're going to have trouble doing I/O). I think I missed something with what you said there. The main 64-bit part of darwin is the math library since they can throw some 64-bit ASM in there plus code to convert back-and-forth to the 32-bit bindings.
I'm geniunely curious - how? A 64-bit application can't do much without 64-bit system calls. Does mach-o let you dynamically load 64-bit code in a 32-bit program (all the 64-bit code would be able to do is computation, given the lack of system calls). Or is there a windows64-on-windows-like 64-bit wrapper over libc?
I'm sorry, but doesn't everyone know by now that the opteron is tangibly faster in x86-64 mode due to a doubling of GPRs in a register-starved ISA? Besides, I suggested running 64-bit linux on both, just to be supremely fair, but I think that it wouldn't be much different from 64-bit amd64 and 32-bit darwin.
That since they are running the Opteron in 32-bit mode, it's not taking advantage of it's full potential. Guess we'll wait until "round 2" like he says, but it still looks bad that he kind of dodges this. If it were me I'd be running the benchmarks on 64-bit linux versus 64-bit linux.(gentoo?)
I don't understand. Why buy G5 xserves when you can get the G5 in blades from IBM? Rack space is expensive, and Apple's only real renderfarm asset (the CPU) is actually IBM's asset...
The card's performance, drivers etc sucked, but one time I put it into the AGP slot and sparks flew, literally ( a bolt of electricity jumped from end to end of the slot ). Smoke rose. Powered the thing up and everything worked fine.
IBM hard drives don't exist anymore. The HD in my thinkpad is a Hitachi. IBM sold their drive division.
Actually, the amount of cache coherency checking needed in a large shared memory machine drives the latency right up. Besides, myrinet is ~15 usec latency. That's pretty goot for most things.
Actually the protocol wouldn't have to be changed much at all. I've investigated this - the alpha value can be passed in the word-alignment byte. If you load up transluxent and throw real alpha values into your XImage, then it will be real alpha. It's just that regular XFree86 ignores the word alignment byte, and transluxent just passes it on to OpenGL so it actually works fine. Unfortunately transluxent isn't very stable.
Actually it is. He said he just bought a laptop.
Yes.
5.1 works with SB Live! cards under linux perfectly. Even the value ones if you make an RCA connector and attach it to the pins on the card (or wimp out and buy the $60 attachment). The only problem I've had is in finding applications that do AC3 passthrough right - I've had success with VLC.
Actually I always send the info, because then IE pops up with current tracking information on the bug, and things like "this problem is fixed with x patch. would you like to download x patch now?" It's really a nice system, to tell you the truth.
Actually OS crashes do get sent. When you boot up, xp will recognize that it had just crashed and will offer to send the info.
Exactly. The coolest part about TK (well, at least tkinter for python) is that you can "connect" a widget and a variable - someone edits the widget, and the variable changes value, automatically. Saves a LOT of code that would otherwise listen for event -> get widget value -> set variable value.
OK, I RTFA, and it wasn't what I was expecting.
Isn't it a property of these kinds of things that you can seperate two electrons (or some subatomic particle, can't remember) and change one's spin, and the other, no matter how far away, will instantly change? I recall an experiment in which this worked over a distance of six miles. Wouldn't this be the perfect interconnect? No wires at all?
I think that he was referring to rpm -q. You need rpm -qp to query a package file, but only rpm -q to query an installed package.
Also, for those of us to shun nautilus and heavy GUIs in general, his suggestions to tar make sense. Are those who don't use KDE-GNOME not entitled to be annoyed by some of the GNU tools?