switch to room list switch to menu My folders
Go to page: [1] 2 3 4 5 ... Last
↑↑↑ Old messages ↑↑↑            ↓↓↓ New messages ↓↓↓
[#] Sun May 13 2012 21:23:04 EDT from LoanShark @ Uncensored

Subject: Re: EC2 FF11 and EC2.....

[Reply] [ReplyQuoted] [Headers] [Print]

I know that at the bottom tier, they can choke off performance to the

point where things really start to be baaaaaaaad...

This is true of Micro instances: if one of your processes starts spinning, and consuming the CPU, they will allow you to burst the CPU for a few seconds, but after that, you're throttled pretty severely. You get what you pay for; that is the lowest-end instance type. Micros might have a few other stability issues but that's the main one, and it manifests from time to time as a briefly unresponsive instance with a high "cpu steal" count.

[#] Sun May 13 2012 21:30:53 EDT from LoanShark @ Uncensored

Subject: Re: EC2 FF11 and EC2.....

[Reply] [ReplyQuoted] [Headers] [Print]

The same configuration as used in EC2 appears on my laptop, where I
am so far unsuccessful in getting Firefox to die. I have to manually
kill it to reproduce the symptoms as I'm seeing them in EC2.

My default assumption is that this is not EC2 specific; Xen is pretty good technology, and from the point of view of a userland process that isn't doing anything fancy with the kernel, it's usually identical to a bare-metal Linux instance. That said, there have been issues with specific Intel processor models within the last year. I am losing the details, but the one I am thinking of was I believe fixed in an Ubuntu -318 patchlevel kernel for Ubuntu 10.04 LTS. I don't know of the applicability to other distributions, but this issue was specific to a particular model of Nehalem-based Xeons. So the standard advice in the Linux world applies: make sure you're running the latest kernel. I recommend Ubuntu over Fedora for the EC2 use case. Also, if you are running into trouble, take note of /proc/cpuinfo for the CPU model information, and it may be worth relaunching your instance repeatedly until you wind up with different hardware. But of course this is may not really be specific to EC2 at all.

Standard stupid questions apply: why is your Selenium/FF process dying? Could it be something simple like the OOM killer? Enable core dumps (ulimit -c unlimited) and capture the result.

[#] Sun May 13 2012 23:18:01 EDT from IGnatius T Foobar @ Uncensored

Subject: Re: EC2 FF11 and EC2.....

[Reply] [ReplyQuoted] [Headers] [Print]

So the Ubuntu kernel runs better than Fedora in EC2? Any idea what the secret sauce is?

[#] Mon May 14 2012 10:45:41 EDT from kc5tja @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

oom not occuring as far as sysslogs indicate.
I've updated the instances to mediums, but
the problem persists.

[#] Mon May 14 2012 12:47:21 EDT from LoanShark @ Uncensored

Subject: Re: EC2 FF11 and EC2.....

[Reply] [ReplyQuoted] [Headers] [Print]

So the Ubuntu kernel runs better than Fedora in EC2? Any idea what the

secret sauce is?

No, because it's a secret--literally. For a long time, EC2 kernels were somewhat customized, and technical details of EC2's Xen customizations are under NDA. This is changing, however; although Ubuntu 10.04 ships an ec2-specific Xen kernel, the 12.04 is using the mainline linux-virtual kernel.

All that said, Amazon has their own Linux distribution which is CentOS based, which might also be a good bet. Haven't tried it because it's not RightScale integrated.

[#] Thu May 17 2012 10:07:10 EDT from fleeb @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

Has anyone here tried to create a set of color bars (as in, for television) in code, but not for RGB... for YUV instead?

(That is, not RGB as in a standard bitmap, but YUV where you have chroma and luma values instead of straight red/green/blue).

I want to generate a color bar pattern in raw v210 color scheme, 1080p resolution.
But I'm not very up on setting up those values, and wondered if anyone here might have a little more insight.

[#] Thu May 17 2012 13:29:13 EDT from Spell Binder @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

A quick google and I found this site describing how to convert color values between RGB and YUV, also called Y/Cr/Cb:

It seems like it's a pretty straightforward calculation.
Color Binder

[#] Thu May 17 2012 15:16:29 EDT from fleeb @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

Yep, that's the site I found as well.

Oh well. It takes a while to create the 1920x1080 Y/Cr/Cb image for the v210 color scheme, but I think I got it. That's a damned weird packing order.

I hope I got it, anyway. I'd love to do this with the proper color bar test, but what I have should do the trick instead. It's close, at least, just missing the bottom colors, since I felt a tad too lazy to figure that out.

Now, to create a 1k Khz tone in the right format, and see if my filters work.
Trying to work around not having the hardware I need to do a proper test.
It also makes it easier to do testing on my dev box (which can't install the hardware).

[#] Thu May 17 2012 16:29:41 EDT from Spell Binder @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

Now that I think about it, I would imagine that there must be pre-canned color bar images available somewhere in the right format. Why not just grab the one you need in the right resolution and play it back when requested?

Are you sure you meant 1k KHz tone? That would be equivalent to a 1 MHz tone, which is beyond the range of human hearing. Though that would be a good frequency if you're testing a software low-pass or band filter.
Tone Binder

[#] Sun May 20 2012 08:21:13 EDT from fleeb @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

You're right, of course, I mean 1k Hz, not 1k Khz. Stilly me.

I am not sure where to find the pre-canned color bar image for the right format, although I think I can find it for RGB, then convert it (although I am not sure I can find it as a true 1920x1080 image).


I didn't finish the project, but I oughta. I'd be nice to create a kind of source filter that just generates this image.

[#] Fri May 25 2012 11:53:17 EDT from IGnatius T Foobar @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]


Whenever I see the PHP directive "include_once" I want to change it to "include_now_once" to make it more Pennsylvania Dutch.

[#] Fri Jun 01 2012 21:15:38 EDT from IGnatius T Foobar @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

Wow. Just ... wow.

$ php

function foo() {

Segmentation fault

If you accidentally perform an infinite recursion in PHP, it doesn't throw
an exception, it doesn't tell you it ran out of stack space, it just SEGFAULTS

I've been banging my head on the keyboard for a week trying to figure out why
my Z-Push backend isn't allowing attachments to be downloaded, only to discover
tonight that I've been crashing the interpreter.

So much for the brave new world of managed code. At least I can open my C
programs in a debugger and immediately find out what went wrong.

[#] Fri Jun 01 2012 23:04:26 EDT from LoanShark @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

LoanSharks-MacBook-Pro:~ ls$ php

function foo() {

Fatal error: Allowed memory size of 134217728 bytes exhausted (tried to allocate 523800 bytes) in - on line 4
LoanSharks-MacBook-Pro:~ ls$ php --version
PHP 5.3.8 (cli) (built: Dec 5 2011 21:24:09)
Copyright (c) 1997-2011 The PHP Group
Zend Engine v2.3.0, Copyright (c) 1998-2011 Zend Technologies

All php.ini at default settings for OS X 10.6

[#] Sun Jun 03 2012 21:56:11 EDT from IGnatius T Foobar @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

I've heard about it failing that way too. What I really need is for it to tell me what I screwed up, and print a stack trace as is commonly done on computers that are not older than I am. The reason I'm at this point in the first place is because I did some Googling on PHP segfaults and eventually landed on a web page that mentions "when the PHP interpreter segfaults it's usually because of infinite recursion."

There's something called "xdebug" that I can supposedly add to my PHP to help out with this. I'm going to give it a try.

[#] Mon Jun 04 2012 09:00:10 EDT from LoanShark @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

RTOL as far as I'm concerned. PHP was not designed; it's actually an amalgamation of warts. It has warts growing on other warts.


[#] Mon Jun 04 2012 13:20:08 EDT from IGnatius T Foobar @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

I'm not using PHP by choice; it's an existing mobile sync framework to which I'm adding my own backend. I would not have chosen PHP. (No surprise there, I write everything in C, but even if managed code were a given I probably would have chosen some other language.)

[#] Fri Jun 08 2012 15:51:37 EDT from IGnatius T Foobar @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

One sign that you're a good programmer is that you don't put double dots into hostnames. :)

[#] Fri Jun 08 2012 16:10:59 EDT from dothebart @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

yea, thats what happenes when you click codeproject links ;-)

fancily some proxies/dnses like it, some don't...

[#] Mon Jun 11 2012 12:59:08 EDT from IGnatius T Foobar @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

Decades later, I still find the expression "x = 1 - x" to flip the value of a boolean represented as an integer to be wonderfully clever. Call me a simpleton but there's just something attractively elegant about it.

Go to page: [1] 2 3 4 5 ... Last