okay, I'm officially stuck

Forum: LXer Meta ForumTotal Replies: 7
Author Content

Jun 07, 2014
11:52 AM EST
You know that post-win letdown that Olympic gold medalists feel? Well, I wrote a Linux kernel driver, and now I'm left with that feeling that any article I write from here on out will be "fluff" by comparison.

The one interesting discovery I got, thanks to the driver, is that moving from register to register in ARM is 400 times slower than moving an immediate to register. The amount of operand shift has no impact on the timing. I've been over the code several times, run the code in GDB to make sure the countdown loops are all working as intended, and have found no coding error that would account for such a wide discrepancy. But such an arcane performance question would hardly be an interesting article.

So, does anyone have any interesting ideas, that don't involve distro reviews?

Jun 07, 2014
3:03 PM EST

Solve, if you please, the insidious lack of .mov file recognition, playability, and edit-ability so prevalent in Linux. I'd do it, but can barely decipher a bash file. ;)

Jun 07, 2014
5:49 PM EST
Expand outward from there into the hardware. Are there finite-state-driven device chains (I2C, SPI, OneWire, debugWire, etc) for which no kernel drivers exist?

I'm currently looking towards pulling together a lightweight single-master RS485 protocol for MCUs ('51,AVR,..?) using the Intel 9-bit convention (top bit raised = address, all devices check to see if it's for them, the addressed one responds to all bytes while the others mask off serial ISRs until the next address byte comes round), but already I know that ARM devices such the one in the rPi lack 9-bit mode. It can be faked at driver level by bit-fiddling the UART's parity bit...

Oh, and CONGRATULATIONS on your achievement. I have yet to find spare time or high priority in which to undertake that.

Jun 09, 2014
11:39 AM EST
@notbob, I can play .mov files just fine with MPlayer, at least the ones that my brother sends me of my nephew being cute. As for editing a multimedia file of any kind, I'd be out of my depth.

@cr, according to the peripheral specs (p. 10), the mini-UART in the BCM2835 doesn't support parity.

I had an idea last night for a generic ARM article: an examination of multimedia instructions in ARM v6. It supports a rough equivalent of MMX, with a few (*ahem*) interesting additional instructions that support complex math.

Jun 09, 2014
12:35 PM EST
@gus3, I'm running Slack 14.1 on an ancient P4 platform and am having bizarre issues with X. xine plays the vid portion of .mov files, but audio keeps dropping out. vlc plays aud/vid pretty well, but keeps having short vid scrambles every 10 secs. mplayer jes flat out crashes X clear back to a tty.

I'm pretty much resigned to buying new i5/7 guts fer my desktop box and quit trying to keep this old hardware wheezing along. Maybe make it a headless server. I think it's about time I quit trying to run protocols/formats on hardware a decade older than the software. But, thnx fer taking the time to reply. ;)

Jun 09, 2014
1:18 PM EST
Okay... /me puts on tech support hat

1. Do you have /etc/X11/xorg.conf on your system?

2. Have you customized it for your video card?

(I don't need to see it.)

Jun 09, 2014
2:38 PM EST
Yes to both questions, but probably couldn't hurt to refine my tweaks. I'll do some additional research.


Jun 09, 2014
6:18 PM EST
My netbook's is very simple:

Section "ServerFlags"
Option "AutoAddDevices" "False"
Option "AutoEnableDevices" "False"
Option "AllowEmptyInput" "False"

This leaves the video driver auto-detected, but I make no guarantees about functionality or democracy for anyone else.

Posting in this forum is limited to members of the group: [Editors, MEMBERS, SITEADMINS.]

Becoming a member of LXer is easy and free. Join Us!