Root Access

@DavidPostill Yea, pretty shady. Security Now covered that issue last week. The correct way to dismiss that box is to click the "Click here to...." text link.

@MichaelFrank God forbid you leave a computer running for a week without interaction.

And this is why I have auto updates disabled.

And also why if Win10 doesn't stop pissing me off with daily prompts, I'm going to make good on my promise and rip out the update infrastructure.

Hm... or null-route the update servers.

@Bob My least favourite prompt is the "We installed updates. Click here for more info." that doesn't bloody do anything.

12:51 AM

lol

I hate the lack of feedback on what an update is

1:09 AM

er

never mind. I'm not awake yet

Ben N

Speaking of updates, I'm going down for some now!

!!;p

ChatBot John Cavil

Q: How can i find internal IP of Router

1

I have Ruckus MM2211 router. I wan to access router login page so i can customize to my needs. The simplest way is the Default Gateway is the ip address of router. But in this case its different. I received this router from my ISP and with modified settings (which remain default even after re...

2:08 AM

@JourneymanGeek That device looks like he pulled it out of a bin... in the 90s.

@Bob more like it was in the bin, since the 90s

Q: Belkin do not work on my internet

2:28 AM

0

Hi I have a question about my belkin. I plugged my belkin in, but I don't see no lights. What does that mean? Do I have to purchase a belkin again? Thanks

google-chrome

._.

4:25 AM

> I recently investigated problems with code that changed the OS time. Very confusing. It took -20 hours btw.

3

@Bob I think I did that a while ago

My belkin give me stomach cramps

!! Caat

ChatBot John Cavil

@qwertyuiop you should buy a new belkin

4:44 AM

Victory or Sovngarde! #Skyrim

Tweeted by bwDraco on May 31, 2016 at 4:41 AM

steamcommunity.com/app/281990/discussions/0/405692224242832513/…

@allquixotic Nitpick: Are you sure shared memory IPC is significantly slower on Windows vs other contemporary OSes?

If you're using named pipes or (shudder) window messages, sure...

AFAIK shared memory is largely bottlenecked by hardware.

You'd effectively be left with some weird kind of pseudo-segmented memory :P

In fact, you don't even need IPC... could just map files directly and rely on the page cache.

Optimal? Not really. Works? ...yea.

@Bob what files? we can assume that game state is transitory and is very frequently updated by both the AI and the player(s); we're not talking about static data here...

Nicolas Raoul

I decoded an email's attachment with Linux `munpack`, the content is fine but the attachment's filename is like below. Any idea how to decode it? (probably Chinese characters)
=XGBKXBX1cK5scf4s8e53L7WudjT2s34wufT38fp0MXPoteoX=XXX=XGBKXBXsai1xLTwuLQoz8K6vszBMbrFKS5kb2M=X=

if you're creating separate "game state files" (whether it's some kind of well-known DB or a custom binary format), and they reside on disk, eventually it's going to sync them to disk, imposing overhead.

4:58 AM

@allquixotic The concept of "shared memory" in Windows (not 100% sure about Linux) is analogous to mmap on Linux. You map views of files.

All virtual memory in Windows is supposed to be backed by a file, whether the automatic page files or a custom named file.

@Bob Wouldn't mapping stuff into the virtual address space be limited in size by the size of the virtual address space to begin with?

There's techniques that allow you to map files without letting them flush to disk, iirc

@allquixotic Views of files. Chunks.

The thing I'm not so sure about... I don't know if re-mapping forces a flush to occur.

Might need another process to hold the same view to keep it in RAM.

I mean, there are only 2^32 (0 to 2^32-1) possible integers you can use to address memory segments within a 32-bit process. If you're going to mmap() some files into your process, the bits of those files you map in, take up some of your memory addresses within your VAS - no?

@allquixotic Yes. Hence pseudo-segmented memory.

you end up having a sliding window or "aperture" into the larger file, and constantly having to move that around - that sounds like context switches to me

5:01 AM

Segment 1: 0-1 GB. Segment 2: 2-3 GB. etc.

@allquixotic It's not ideal, no. Obviously having direct access is better.

But as far as IPC goes this is probably as efficient as you can get.

There's no way transferring data through TCP or Unix sockets would be faster.

I'm not sure how many calls to the shared memory architecture and mapping/remapping functions you'd have to make to bottleneck that subsystem on a typical desktop system, but the ceiling is necessarily going to be a lot lower than having those regions of memory directly addressable through the hardware virtual memory manager in a native 64-bit process.

@allquixotic Yes, definitely worse than just native 64-bit.

but yeah, I wasn't really thinking of shared memory when I wrote that post, and it is probably a good ways faster than other alternatives, short of having the data in-process

My nitpick was the claim that an OS is significantly slower for IPC :P

there's some non-zero cost associated with effectively having a software virtual memory manager

5:03 AM

Which is probably true for things like named pipes.

But they aren't designed for high throughput/low latency in the first place.

@allquixotic I wonder how it's implemented.

In theory, the mapping that occurs wouldn't have to be be much more expensive than having to refill the TLB with 64-bit anyway.

With 64-bit you can use huge pages, but I don't think Windows uses them and I know Linux only uses the mid-sized one and even then only in specific scenarios.

There's 4k, 2M and 1G pages in modern x86.

Linux uses 2M at most. I don't think it ever uses 1G.

I recall reading somewhere, and maybe it was really outdated info from the NT days, that the Windows kernel's time slices for thread running time were quite coarse-grained compared to, say, a 1000 Hz Linux scheduler

Ah, Windows has large page support too but only for privileged users and only when specifically requested.

And probably for 2M only too.

unless new Windows OSes have adjusted that, that could be a source of overhead any time you do something involving a context switch; and shmem almost certainly involves two (user -> kernel and kernel -> user)

Linux was also on specific request only but at some point a patch came in to do it automatically.

@allquixotic Hm. I'm not too familiar with the details there.

@allquixotic "adaptive" - as of Windows 2000:

> A context switching rate of 300 per second per processor is a moderate amount; a rate of 1000 per second or more is high. Values at this high level may be a problem.

So it can switch very often, as required.

technet.microsoft.com/en-us/library/cc938613.aspx

kinda like dynticks with 1000Hz then O.o

might be interesting to measure the context switches per second during a heavily loaded game of Stellaris :P

5:11 AM

Assuming it does the dumb multi-process/shmem thing!

64-bit would be far easier, faster and more reliable.

@allquixotic There's also the question of whether the syscall requires a context switch at all...

Apparently Linux at least will avoid that in certain contexts, pun intended

Linux avoids context switches when doing shmem stuff? or just some syscalls?

@allquixotic "some" syscalls. I don't know which.

shmem really, really seems like the kind of thing that a userspace process couldn't just do on its own without kernel help -- I mean, it involves taking a region of memory that's literally not within the process's memory segment and doing stuff to it to allocate it and make it available within the process's virtual address space

when that kind of thing has to happen within, say, an Nvidia graphics driver, to map stuff into the process so the userspace driver stack can beat on it, it has to do a context switch into the driver to do the kernel-side allocations to support it.

> Typically a system call only leads to a context switch if the call will block, or if the caller has used up its timeslice; otherwise, it's just a couple of relatively cheap protection domain crossings.

a userspace process really shouldn't have the kind of permissions necessary to do that without kernel help :P

5:17 AM

@allquixotic Well, apparently modern CPUs can avoid flushing the TLB on a 'context switch' (speaking of which, the MIPS I played with could do that too!)

you're pretty much preserving registers and changing stacks (changing the stack pointer)

so it's not really expensive at all

The most expensive part traditionally would probably be the TLB flush

I feel almost compelled to write a C program exercising this functionality on Linux and Windows and see if there's any performance difference, and see how many times it can do it per second

lol

I suspect that you'll find the Linux one faster, if only because there's significantly more indirection in the Windows one (unless you want to use NT calls directly)

general idea would be to compile a 32-bit program that allocates a large shmem region and does something like... write out a series of increasing unsigned 64-bit integers into the address space and read them back as fast as possible up until a pre-defined limit, then stop and report the time taken

But it shouldn't be that much faster than Windows - end of the day, most of your time is still spent in reading the data from RAM.

really the question I have is how many round trips are possible per second

because the number of round trips per second when directly accessing RAM within a userspace virtual address space probably numbers in the billions at this point (?)

5:23 AM

@allquixotic You can probably speed up Linux a lot with MAP_HUGETLB and MAP_HUGE_1GB

For Windows, it's MEM_LARGE_PAGES

DMI advertises 5 or 10 Gigatransactions per second (can't remember the exact number); wonder how many DMI transaction slices it takes to do a memory read on Skylake

ideal would be 1 :P

@allquixotic 7-cpu.com/cpu/Skylake.html

Using 1 GB huge pages will save you some ~256x the page table walks :P

oooo

This is probably how other people feel when they walk in on me geeking out over hardware ;p

(I have no idea what's going on)

lol

I'm relying heavily on a single OS-dev course -_-

But it was very good info on how *nixy OSes handle virtual memory, and I have a habit of getting stuck in big branching links with dozens of tabs of info.

@allquixotic Another thing to keep in mind, assuming context switches are primarily expensive due to having to discard the TLB: you can only really hold some 4 GB of data TLB entries at once, so... you'd have to do a fair bit of the work of a context switch when reading such a massive amount of RAM, even in a 64-bit process :P

And if we're talking not using huge pages (which Linux only sometimes does and Windows doesn't do at all without specific request), well. That's now down to some 250 kB. Total.

5:34 AM

@Bob little math from there and you can hit RAM about 615 million times per second at 4 GHz

@allquixotic Remember, though, that's only for direct access or addresses already in the TLB.

e.g. repeated access to one address.

If you're scanning memory with 4kB pages, the time spent mapping those pages would be significant.

256 kB in data TLB, ~6 MB in second-level (slower) TLB.

Every 256 kB you're doing that mini-context-switch.

A context switch is saving registers (very fast), changing stacks (should be fast?), flushing TLB (slow-ish). And the sysenter/sysexit/interrupt (couple cycles, dropped pipeline?).

Oy vey. Not into system programming.

Dropped pipeline is way more than a few cycles.

It's more like 10-15 cycles.

@bwDraco While we're talking reloading the TLB mappings with some 42 cycles (possibly many more if caching doesn't kick in considering most page tables are 3-4 levels nested)... 10-15 once per switch is a fairly negligible part :P

But of course I don't have the real numbers.

But my "?" was more not knowing if sysenter actually drops the pipeline.

Oh, @allquixotic blog.tsunanet.net/2010/11/…

I'm going to sleep now, but I'll catch that when I wake up... night :)

'night

ugh

I had my adblocker off... giant animated ad literally every paragraph on tomshardware

the ads were bigger than the paragraphs

the text was unreadable

Psycogeek

5:54 AM

Thats like me, fire up a "normal" setup or just installed, and start surfing the web and it's like F---This , why would anybody even bother with the web.

I guess some people will have grown up with it, have built in imunities :-)

7:32 AM

hey

HackToHell

hey

i've had the same review challenge twice now

7:48 AM

i really need to get some headphones

not sure if i can survive another week of terrible local radio

9:08 AM

9:31 AM

morning all

;)

yay

HackToHell

@DavidPostill congy

9:58 AM

congrats

streamable.com/efv0

10:46 AM

anandtech.com/show/10337/…

Broadwell-E is here. The top-of-the-line part now costs upwards of $1700 o_O

As expected, it is based on the Broadwell-EP LCC die used for Xeon E5 v4 processors up to 10 cores.

11:15 AM

3dmark.com/3dm11/11290643

T___T

I wanna die... and reborn somewhere with cheaper tech.

Not unusual for a mainstream system.

Well, it costed me 1.5x my monthly salary back in 2014

11:31 AM

Oy vey. My computer has processed 200+ images and it took more than 45 minutes to do it.

Jan 3 at 2:25, by bwDraco

No wonder why creative professionals often use workstations with insane core counts...

Pixel counts are getting higher and higher. This machine is barely keeping up.

@bwDraco heh. 3d rendering is embarassingly parallel

Well, my post-processing software is fully multithreaded.

heh

@bwDraco we had someone bring a machine with twenty cores to a crawl

All cores running full tilt for more than 45 minutes.

@JourneymanGeek Then again, I'm doing this only because my trade requires it.

I'm just using the workflow I've always used on the hardware I've had for nearly 2-1/2 years.

By the time I start dealing with 36MP images, this is going to become untenable.

Then again, I'm not even sure if I'm still going to be shooting for the college since I'm graduating...

Camera frame rates are higher than ever, too. I've already cut down on loads of duplicates and unusable images. If I do it the way some pros do and shoot hundreds of images a game, this laptop is clearly not going to be adequate.

Jan 8 at 22:54, by bwDraco

...especially if they are from today's breed of super-high-resolution cameras (*cough* Nikon D810 *cough* Sony A7R II *cough* Canon EOS 5DS *cough*).

Interesting new feature on these new HEDT processors: Turbo Boost Max Technology 3.0.

Special software can identify which cores can run the fastest (due to normal variation, just as some chips can overclock better than others) and put the most CPU-intensive tasks onto those cores first.

That can prove handy for overclockers, too, as they can identify which core is the slowest and make the very most of every core on the chip by adjusting each core's clocks individually.

It'll push beyond the processor's specified Turbo speed, yet still be considered within spec.

11:56 AM

@bwDraco That means nothing. What kind of images, and what processing, makes all the difference

24 MP RAW images with noise reduction and several other steps through RawTherapee.

hmm

Anyone's done a system refresh on windows 10 yet? Anything I should watch out for?

walking a friend through it

@JourneymanGeek done it on 8 and lost the licence...

I warned her, and have a backup

A: what kind of m.2 ssd does MSI GE62 Apache Pro-004 support?

@bwDraco You might want to respond to this answer "That's probably the worst answer someone could give - bwDraco."

-1

That's probably the worst answer someone could give - bwDraco. What if you already have the device but want to upgrade, and what if there are a lot of people (including me) who want to find more about the correct SSD to use? There are numerous forums for deciding which device to buy, but a lot o...

12:06 PM

meh

Don't bother

...gone.

I just voted to delete.

No reason to feed the troll.

and/or someone who dosen't quite get the site

Good that it got sorted so quickly :p

lol

I was waiting on my friend to test her rescue disk ;p

@ThatBrazilianGuy lol

@ThatBrazilianGuy Huh I did think that looked a bit weird

12034 = minimum spec? But my <10,000 machine meets minimum spec :-/

12:18 PM

I'm currently a rubber duck

6

I'd rather be a teddy bear :p
"Another effective technique is to explain your code to someone else. This will often cause you to explain the bug to yourself. Sometimes it takes no more than a few sentences, followed by an embarrassed "Never mind, I see what's wrong. Sorry to bother you." This works remarkably well; you can even use non-programmers as listeners. One university computer center kept a teddy bear near the help desk. Students with mysterious bugs were required to explain them to the bear before they could speak to a human counselor."

May 22 at 22:08, by qwertyuiop

I am a noodle

I'm a poodle

2

@Bob Yup. Welcome to... most older tech sites

@allquixotic I think they were.

wiki.kerbalspaceprogram.com/wiki/…

12:26 PM

@qwertyuiop ... :( they used to be good! I think?

I'm not entirely sure if I'm thinking along the correct lines but the kernel tick in Windows was a few milliseconds

@Bob Yeah. They used to.

I believe anywhere from 2-15ms was the Windows 7.x kernel tick timer

IIRC Windows 8.x moved to a tickless kernel

I'm a human and I request minority quotas.

@qwertyuiop where are you getting these numbers from?

@ThatBrazilianGuy 3dmark.com/results

3dmark.com/fs/7195909

(From "My results" section)

No, I mean, where does it say 10k = minimum?

@allquixotic you're a cat in a tube

2

12:31 PM

Well it says 9271 = minimum on my screenshot

heh. apparently refresh failed for my friend

it just booted back in

And I know my <10k PC meets the minimum because I ran the test in Steam's minimum checker

anomaly intro

►

^-- Agreed the end of the video the font ruins this?

@GuitarShoeDave Could you try saying that again, in English this time? :P

@GuitarShoeDave Seems fine to me

12:40 PM

the font is very very very dull

At least it's not Comic Saaaaaaaaaaaaaaaaaaaaaaaaannnnnnnnnsssssssss

@qwertyuiop The performance test is 4.7 GB! O_O

yup

Well, 1.9 GB download.

@qwertyuiop minimum for oculus rift.

12:44 PM

@Burgi: The font is disgusting, It should be more fluid id think. (friends asking for feedback)

needs to be more computery

or a gothic style

with the really thin lines

I told him that font is soo bad to may as well use "times new roman" if he wanted to use a bad font.

US govt uses times new roman on everything -- STILL to this day.

google.com/fonts/specimen/Exo+2

@Bob that will be amazing to download when all ISPs here instate limited broadband nationwide. Now all citizens will be able to live the mobile experience!

@Burgi: that font's wayyyyy cooler

1:09 PM

Someone stole my .tk domain

Or I forgot to renew it.

@ThatBrazilianGuy What is the full domain name?

lol

Eh, it was at ~85 for a bit. Mostly 100-110. Not all that terrible, for a mid-range old card :P

1:43 PM

I need to run that test tonight on my new box

shouldn't have much problem on a 6700K with a 980, though it'll be insane once I get the 1080

supply isn't remotely close to meeting demand on the 1080 at this point, probably won't be for a month or four

2:02 PM

@allquixotic I seem to recall the same issue with the 980s when they were new

ow

took 1 min video

200 MB

HackToHell

2:20 PM

@JourneymanGeek its a 2 gig download

:O

@Bob 4k?

Q: Want a VIRUS for Testing or Virtual Machine? You come to the right place

@JourneymanGeek Please nuke

-2

Want a VIRUS for Testing or Virtual Machine? You come to the right place! This is a real 60+ Virus and really mess up your PC. Before download: Please disable Antivirus, DO NOT open any of these viruses on your actual pc. I won't be responsible if your PC is infected. File size: 123 MB File for...

virus

yeah

that is seriously lulzy

@HackToHell 1080p

30fps I think

@JourneymanGeek thanks

@JourneymanGeek ^ one pair came like that; the other was fine :P

2:49 PM

@DavidPostill reminds of the time when I downloaded a zipfile called.. UltimateVirusArchive.zip and antivirus alerts practically killed my computer the moment the archive was scanned years ago.

@GuitarShoeDave Oh yea. EICAR proved just how worthless Norton was.

In that it refused to give me the option to keep the file.

Put that together with many actual false positives and I stopped using that pile of crap.

lol jeeze

3:20 PM

ugh

refreshing fails, resetting fails, trying to do a in place install fails

I think we'll need to try reinstalling off a disk ._.

all this because of a printer?

yup

3:41 PM

superuser.com/review/low-quality-posts/539247

lol

@allquixotic going to try reinstalling

or talk her through it

Boris_yo

4:03 PM

Is aspect ratio of 2:1 is not similar in shape to 4:3?

Just want to make sure 4:3 photo will be suitable for 2:1 aspect ratio without resolving to cropping again.

@Boris_yo well, no, they're not similar because they're different

Is it considered abuse if you basically use a trainee as a rubber duck?

What if it's unintentional? Because I did just that.

1000x500 is not equal to 1024x768

@ThatBrazilianGuy only if it involves a bath tub.

2

if your horizontal dimension is the same, you'd get something like 1024 x 768 vs. 1024 x 512 -- the 2:1 will have fewer vertical pixels than the 4:3

4:10 PM

@JourneymanGeek Hey @qwertyuiop, you're looking different today!

I look zee same!

@JourneymanGeek yeah, you're still Ash the wauzer

@Boris_yo These aspect ratios are not similar, because, well, they are different! It's like asking if circles and ellipses are similar in shape.

@allquixotic the same ava I've had a while too

@JourneymanGeek You should try again, but with a Groucho disguise this time!

So I had this issue bugging me since Friday. Simple, but yet annoying and I couldn't find a solution. The intern just asked for a new assignement, I started explaining it to him and them... bam, the solution is obvious.

Also, Drupal can be quite confusing sometimes.

5:02 PM

3

@DavidPostill Sonic is slim and fast, Drupal not quite so...

@ThatBrazilianGuy oh yeah, you're one of Those Content Management System Guys, aren't you?

@allquixotic I am forced to use Drupal at work, but I rebel wearing a different WordPress shirt each day of the week ;p

ThatCMSGuy...

sure it's not your hardware or your web frontend making it slow?

Let me save you some time: if the question about my workplace involves "is this problem caused by infrastructure", the answer is always "yes".

A coworker from other department was just telling me how it takes 7hrs to send 45k emails, and someone somewhere else sendes 800k+ emails in less than that time.

We also have blackouts every two weeks and near-daily internet outages.

And the situation of the datacenters... just don't ask.

@allquixotic Although I wasn't calling Drupal slow, I mean it's not slim because to just start using it you need to (1) create user profiles, (2) create a content type, (3) create an input text type, (4) something I'm probably forgetting

Compared with other popular CMS's, Drupal is much more flexible, but with the cost of being much more complex.

5:23 PM

@ThatBrazilianGuy yey rubber ducking

Sooooooooooooo

we have a machine that only the local admin can do a windows update on :/

How? No idea :/

as anyone on the domain you're hit with what looks like group policy, cept when you check the group policy, it doesn't set anything :/

so i tried using gpedit to override it - no help

yey giant ducky

Why is giant ducky dead

did a small child fall in it's pond?

@djsmiley2k No, the pond is fine...

See? Nothing to worry, the duck got better and even went to a festival!

5:28 PM

yey

Giant rubber duck in Taiwan: Crowds welcome giant yellow duck

►

Rubber duck artist's team alleges plagiarism by Brazil protesters

D:

|:

5:32 PM

@DavidPostill NOOOOOOOOOOOOO (＋Θ＋)

lmfao

@ThatBrazilianGuy: We were thinking the same

6:03 PM

Anyone have any experience with PXE network booting? Trying to figure out if I can install windows 10 with it.

But as far as I can tell, I need enterprise edition to do so.

@WilliamMariager you can install any operating system at all by loading an image over PXE, but if you want to actually invoke the official windows installer/bootloader, you may indeed need enterprise edition for that

if you can get away with pushing an image you might want to do that instead

How would I do that?

PXE load Linux and use ftp + dd in a startup script, for example

could hook up an automated script that bootstraps networking, downloads the image, dds it to the disk, then reboots

rlemon

o/

6:31 PM

oh no!

runs you need help with ssh or something like that, right? :P

jk

6:47 PM

PXE w linux + nfs / cifs works too +dd

its a bit easier to implement imo

nfs probably the best option although not necessarily the most secure.

Ben N

7:00 PM

If the Home edition of Windows 10 has WIM files in the "sources" folder on the installation media, you can WDS it

I'm fairly certain all editions use WIMs

7:15 PM

CaptionBot thinks an NVidia GPU is a cellphone, lol

7:33 PM

imgur.com/5S34X73

Felipe Falanghe, original creator and developer of KSP, is leaving the KSP project :'(

8:12 PM

blogs.sciencemag.org/pipeline/archives/2013/05/08/…

Tom Wijsman

8:25 PM

8:46 PM

What makes an SSD industrial?

Better tolerances, like temperature, voltage, etc?

Developer

is windows 10 using nt2 hashes in sam ?

Ben N

9:45 PM

@Developer I thought it used the same hash as before, NTLM

Huh... Was about to ask for suggestions on a good File Manager for Android that wasn't ES File Explorer. Then I realised I was sitting in front of a computer with a USB cable handy.

10:08 PM

@WilliamMariager never heard of an "industrial" SSD....

however, there's enterprise-grade SSDs; the term basically refers to SSDs that have been more rigorously tested and engineered to last for more write cycles before dying

@WilliamMariager Yes

@allquixotic google.co.uk/search?q=industrial+ssd

Usually wider temperature and humidity tolerances. Sometimes power failure protection (battery/supercapacitor backup), more robust or proprietary connectors, dust/liquid ingress protection, and pretty much always high-grade shock and vibration tolerance.

10:50 PM

Broadwell-E once again proves Intel's dominant position allowing them to rip off and milk clueless "enthusiast" "gamers" with ridiculous overpriced products

isn't it $1750?

heh... This guy made his "Me too!" answer a community wiki to avoid rep loss.

Oh... Unregistered user. Perhaps he just didn't know.

11:13 PM

@allquixotic newegg.com/Product/Product.aspx?Item=N82E16820208346

Just to show what I was talking about :)