Computing desk
< June 3	<< May \| June \| Jul >>	June 5 >

Welcome to the Wikipedia Computing Reference Desk Archives
The page you are currently viewing is an archive page. While you can leave answers for any questions shown below, please ask new questions on one of the current reference desk pages.

June 4

Hypervisor for GPU Virtualization

Hello Everyone,

I'm familiar with regular hypervisors which virtualize CPUs and present them virtual machines as vCPUs, but are there any hypervisors which do the same with GPUs? That is, virtualize the cores of a GPU card (particularly Tesla) and present them to a VM as vGPUs?

Thanks as always. Rocketshiporion ♫ 08:24, 4 June 2011 (UTC)[reply]

CUDA (at the driver layer) is essentially a virtualization host environment for your GPUs. What additional level of virtualization do you need? The Tesla uses NVidia's PTX instruction set so even if virtualized, it will still be unable to seamlessly run code for other platforms. But if you just want a Tesla to be visible inside your VM guest operating system, you only need to install the CUDA driver in the guest OS (provided, of course, that the VM supports PCIe and plumbs it properly to the guest virtual machine). Perhaps you are seeking processor emulation implemented in CUDA, rather than hardware virtualization? Nimur (talk) 15:01, 4 June 2011 (UTC)[reply]

I would like to present a specific number of Tesla cores, rather than all the cores available on a Tesla card, to a virtual machine. Rocketshiporion ♫ 17:25, 4 June 2011 (UTC)[reply]

CUDA 3.0 allows you to run different programs (functions) on the same Tesla, dynamically allocating the tasks to the cores. Conceivably, you could write a program to plumb that up to the VMs, but it would be a lot of effort. Here's an implementation of pthreads in CUDA: POSIX Threads and NVIDIA's CUDA - it's a start, but the work isn't done to plumb that up to the guest OS. Nimur (talk) 03:03, 6 June 2011 (UTC)[reply]

Apple Iphone touchscreen sensitivity

Phones even with higher processing capacities do not seem to have the same touch experience that an I-phone carries.Does Apple use a dedicated processor for enhancing the touch screen sensitivity and touch response?Sagarbidri (talk) 13:41, 4 June 2011 (UTC) — Preceding unsigned comment added by Sagarbidri (talk • contribs) 11:15, 4 June 2011 (UTC)[reply]

My guess would be absolutely not, and that the majority of differences are in the software, but there's plenty of information out there to make sure yourself: http://www.google.com/search?q=iphone+4+breakdown ¦ Reisio (talk) 17:21, 6 June 2011 (UTC)[reply]

Facebook question

I wish to ask a facebooker a genealogical question. Is it a breach of Facebook's terms and conditions to messenge her when we don't know each other? Could I get struck off? Kittybrewster ☎ 15:12, 4 June 2011 (UTC)[reply]

Not at all. I have used Facebook to search for people I used to know physically and because so many people have the exact same name, its hard to narrow it down sometimes, so on occasion I have resorted to messaging people directly. The worst that has happened is my messages have been ignored. --KägeTorä - (影虎) (TALK) 15:31, 4 June 2011 (UTC)[reply]

It would be strange for Facebook to make it possible to send someone a message while at the same time disallowing it in their T&Cs. AndrewWTaylor (talk) 18:30, 4 June 2011 (UTC)[reply]

As others have hinted, people can choose whether or not they want to receive messages from people who aren't friends (or friends of friends). If they choose to accept messages then obviously it's not against FB T&C for you to message them. Of course if you message an excessive number of people, or your messages themselves are offensive or otherwise violate the T&C or you keep messaging someone when they tell you not to that's a different matter. Nil Einne (talk) 19:38, 4 June 2011 (UTC)[reply]

Can a flashable BIOS be modified to start a bootloader larger than 512 bytes?

Same as the section header. --Melab±1 ☎ 22:24, 4 June 2011 (UTC)[reply]

In theory yes, you would just read two sectors of the disk instead of one. But why would you want to do it? As you may need a custom disk format to support your boot record. Are you trying to cut a few milliseconds from boot time? Graeme Bartlett (talk) 23:23, 4 June 2011 (UTC)[reply]

How does VPN affect my Internet connection/privacy/security?

If I VPN into my university network (to access certain networked drives for my work/study) does that mean that all my traffic is directed through their servers? Would that mean they'd know if I downloaded a movie while VPN-connected? Also, does it hide my IP address from the websites I access? — Preceding unsigned comment added by 2.220.167.229 (talk) 23:30, 4 June 2011 (UTC)[reply]

Yes, it means all your traffic is directed through their switches and routers (which I think is what you may have meant by "servers"), and they would have access to complete logs of all your Internet traffic, including downloading movies, if they've configured their routers to keep such logs. Which they probably have. You can use a site like ipchicken.com to quickly ascertain what IP address you are at; websites you access will see whatever IP address the university's VPN routers are set to. Comet Tuttle (talk) 03:22, 5 June 2011 (UTC)[reply]

Actually this depends on the configuration of the VPN. It's possible to set up a VPN with "split tunneling", in which case only the traffic to the organization's own network will traverse the VPN tunnel; traffic to the open Internet will use your connection as usual. People sometimes do this to reduce the amount of unnecessary traffic going over the VPN. However, it takes extra configuration and reduces security, since a worm or attacker that got into your computer over the open Internet could then walk right in the VPN.

In any event, you can tell by using a traceroute command, such as "traceroute" on Linux or Mac, or "tracert" on Windows. Start the VPN, then traceroute to secure.wikimedia.org; if it goes through the university network, you don't have a split tunnel. —FOo (talk) 17:15, 5 June 2011 (UTC)[reply]