Archive for February, 2009

The state of GPGPU

Friday, February 27th, 2009

Today Aaron Plattner from NVidia stated on the X.org mailing list when asked:

if there will be XVideo support for G90 cards to the nvidia free 2D driver

just:

XV on those GPUs requires the 3D engine, and setting that up is too complicated to be within the scope of that driver.

This let me to make some harsh notes what is “in the scope” of a driver and what’s not. Whether the closed source NVidia driver for U*ix-like systems is a favorable way to support (this) class of operating systems, or whether it has problems.

For me the Intel and ATi driver (or for that matter even the superb Matrox driver 10 years ago) always worked well. Of course the latest greatest drivers might have uncovered some compositing performance (or the like) regressions. But that’s the fast lane of the open source way of life. However, at least the availability of source allows to choose a suitable, working one, and also to fix such issues as they appear and invent in this area.

The quality is influenced by many factors, for me the source availability is one of the most important ones - without it many other factors (such as security among others) are hard to determine anyway. And on another side the NVidia driver rarely worked at all. Mostly because due to the sheer lack of x86 hardware on my side in the past (I stuck to PowerPC and SPARC for multiple reasons), and for the few AMD/Intel boxes other ABI incompatibility with the latest X.org server and/or OS kernel on the other side often came in between.

For me this boils down to a near zero usability of the binary-only nvidia drivers.

Let’s compare the GPU situation with classic CPUs:

If the vendors in the CPU-land would act like some of the vendors in the GPU-world still behave today, we would still not now how to enable the i386 protected mode, nor how-to use vector instructions or other new features of modern CPUs. If all programming material for CPUs would be hidden by the manufacturers, we also would only have very few operating systems, like Windows and Mac OS X, for example. Nothing like Linux or the BSDs - simply because few would know how to utilize the real power. And all programs would be limited to just utilitze what the vendor devices to expose to some form of driver API. Like the PC BIOS. Oh yeah - the VGA / VEAS BIOS. What a lovely dream.

It is no wonder GPGPU did not yet really take off and is not in wider use today: When utilizing it means calling into a VESA BIOS sort-of-thing and no-one knows what’s going on behind, how to debug or if it works tomorrow or the next computer nearby. And when innovators in operating system design have no access to proper register level documentation.

Just my 2 € cents on the matter, any why I’ll continue to strictly purchase only, what has open specs available. For drivers I can write, debug, review and improve. For hardware I can utilize in my OS and tomorrow, even when the vendor wants to phase it out.

A companion for your Sub-Notebook (and yourself)

Friday, February 27th, 2009

So I eventually got myself one of those cheap sub-notebooks (yet those kind a NetB..ks) and had to look for something suitable to carry it along.


Turned out most of the gerneral purpose bags out there are pretty ugly - well, in my opinion at least anyway. So it took some time to actually find something I would want to be seen with out in the world at all. Yeah, yeah - I know the result of (too many?) years on Apple hardware kind of design obsesssion …

Anyway, the choice I ended up with is a black Bree Punch 52 - made out of trunk awning. A little like a messenger bag on diet.

VIA Nano U2250 without VT?

Sunday, February 22nd, 2009

Just yesterday I took the first look on a Samsung NC20, the first commercial device spotting a VIA Nano CPU - finally. However, I was surprised to find the CPU not indicating the VMX/VT extension to accelerate virtualization in hardware. Let’s hope the early BIOS version just forgets to enable it and there are still hopes to have a tiny device capable to run virtual machines containers for professional work (running virtual machines for security or work on projects) abroad.

How AppKit’s NSAttributedString(NSStringDrawing) draws the string’s glyphs

Friday, February 20th, 2009

Today I catched some neat exception / assertion trace of an [NSString drawAtPoint: withAttributes:] while drawing with an invalid context. So this is is the calling sequence how Mac OS X 10.5(.6) draws an NSString internally:

CGContextSetFillColorWithColor
CGContextSetStrokeColorWithColor
CGContextSetTextMatrix
CGContextSetFont
CGContextSetFontSize
CGContextGetShouldSmoothFonts
CGContextSetFontRenderingStyle
CGContextSetTextPosition
CGContextShowGlyphsWithAdvances
CGContextSetTextPosition
CGContextShowGlyphsWithAdvances

Exactracting photos from iPhone backups

Tuesday, February 17th, 2009

Well, yesterday night I finally updated my 3G iPhone to the 2.2.1 firmware - mostly because I have bug reports pending over at Apple, where some are said to be fixed in the latest 2.2.1 FW update.

So after the usually too lengthy backup, update and restore cycle (why can’t they just send delta updates over-the-air as they do for regular, real Macs?) I was left with my iPhone not wanting to lock on a provider in Germany, and no - I’m not running it unlocked or jail-broken but with the regular T-Mobile contract here in Germany (sigh!).

Anyway, as the “reset network preferences” and googling was not of help I decided to restore without replaying the backup, and voila: it worked, locked to the all expensive T-Mobile cellular network again. Though this left me with all of my preferences and particularly my not yet downloaded photos missing. Thanks god others already created some nifty iPhone backup decoder!

In retrospect: the MSI S270 MegaBook the first NetB..k I owned

Sunday, February 15th, 2009

Even years before the recent NetB..k ^W sub-sub-notebook storm - kind of initiated by unfortunate Asus EeePC - I already used and enjoyed one. No, not the initial, real Psion NetBook (although I had a Psion Revo at that time, …), but the MSI MegaBook S270.

It was about the size of the 10″ modern flavors and compared to the most ugly Asus EeePC even had a slick and slim design, including a quite thin metallic display cover.

It’s just now - comparing all the many variants today for family and friends asking for suggestions - for me to notice what I was already enjoying back in the days.

HP Mini 2140 and Sony Vaio P - a possible Apple MacBook ultra-portable substitute

Monday, February 2nd, 2009

I’m waiting for many years now for Apple to release a sub-notebook replacement for the formerly quite handy 12″ PowerBook that is EOL from Apple’s side.

However, for some reason Apple is just not getting one pushed out. The only thing they want to sell to us is this awful MacBook Air that is just thin, but not sub-sized and due it’s outside dimensions not too useful in a plane or otherwise abroad (at least for my use cases, anyway).

Thanks to some PC vendors sub-notebooks finally start to reach a decent quality to fill this gap:

Booth the Sony Vaio VGN-P and the HP Mini 2140 are reasonably sized (or tiny), with still a decently sized keyboard to type and of good build quality. Booth also worth a second look.

While the Sony Vaio achives an amazingly tiny outer dimension it even comes with an ultra low voltage Aton Z5xx that additionally features the VT extension and results in a quite reasonable battery life for this pocket PC (nearly 4h on the standard, and up to about 8h with an optional higher capacity battery pack). Despite it’s super tiny size it even brings a WWAN - 3G - chip and GPS! The only downsides are the exorbitant price (in Germany anyway) and Intel’s Pulsbo chipset with PowerVR graphic for which there exists no open source accelerated graphic driver, yet.

The HP Mini 2140 on the other hand is a less expensive choice, but less compact. The regular Atom used, however, features neither 64bit nor VT. While it comes without 3G and GPS, it is of pretty reasonable built quality and design. Over its preceedor (the HP Mini 2133) it also increases the display size to a reasonable level (up from 8 inch to 10) and replaced the aging and hot VIA C7 CPU with the Intel Atom that has slightly more power and runs cooler. While keeping the superb keyboard and compact touchpad (with buttons on the side).

Though I got hands on some pre-sales models of booth I will still have to wait for regular sales to kick in in Germany, as well as the HD screen option for the HP Mini 2140 (1366×768, instead of the SD 1024×600 model) to become available at all.

Well - and while I continue to wait for those to become available, who knows what else is happening on the market till then: VIA Nano anyone? Endurable case design from another vendor, too?