apr_pool_t

While browsing the Subversion doc.s they had a code example of using the apr_pool_t. apr_pool_t is part of the the Apache Portable Runtime, which Subversion uses. The thing that struck me was how similar the mechanics of apr_pool_t are to NSAutoreleasePool.

The difference between these two mechanisms is that NSAutoreleasePool also allows for reference counting, whereas the apr_pool_t is a strict alloc/free allocator. The apr_pool_t interface is more explicit, whereas NSAutoreleasePool does a lot of things for you, such as global tracking of pools instead of constantly passing around the pool as an argument.

apr_status_t  apr_pool_create (apr_pool_t **newpool, apr_pool_t *parent)

This creates a new pool and links it onto the parent pool. +[NSAutoreleasePool new] does the same thing, except that it implicitly links it to the parent instead of the programmer needing to explicitly specify a parent.

void *apr_palloc (apr_pool_t *p, apr_size_t size)

This allocates memory from a pool. NSObjects are not allocated from a pool, rather they are placed into an NSAutoreleasePool via -[NSObject autorelease].

void apr_pool_destroy(apr_pool_t *p) 

This frees all memory allocated in a pool, all of its subpools and the pool itself. -[NSAutoreleasePool release] (or dealloc to be specific) behaves the same way.

APR pool function reference

Given how similar these two mechanisms are, I have to wonder if it could be subjected to the NSAutoreleasePool patent?

1924 trip to Mt. Everest and beyond

In 1924, my grandfather went to Tibet, India and Sri Lanka, he accompanied the Mallory & Irvine expedition to Mt. Everest, taking what are perhaps some of the only photos of the expedition which were not taken by John Noel, the expeditions photographer. My understanding is that he was accompanying the expedition to investigate the possibility of a hydroelectric dam in Tibet. He then went on to India and Sri Lanka for what appears to be leisure.

I inherited the album and scanned it in for my family, here it is for my droves of readers to enjoy:

W.F Lloyd trip to Mt. Everest, India and Sri Lanka

From Windows to OS X, what is it about dual processors.

Somewhere along the way in the transition from a NeXT Cube to a modern Mac I got stuck on Windows. I am a longtime Objective-C developer and have been using the old YellowBox ProjectBuilder on NT/2000 for way too long. The people who use my software are on Windows, so my stuff needs to run there and I need to develop there. For a bunch of reasons I was unable to use a Mac as my main machine, but things have slowly changed and I can start the transition.

The most significant development recently has been Xcode and target custom build rules. For a long time I have hated makefiles and the pre-Xcode PB using jam files was just as bad. It always annoyed me that a project has all the information it needs to be compiled from the IDE, and yet got compiled in this hugely indirect way via Makefiles. Xcode is now at the point that it can do what I want, sans makefiles, and I am pretty happy about it. So I have started to use Xcode in earnest on a Mac mini in an attempt to transition my work life to a Mac.

At some point in the past I switched from a single processor PC to a dual processor capable PC with one processor, and upgraded my version of WebObjects, I am pretty sure it was 3.51 to 4.0, at the same time. The odd thing I noticed at the time though was that despite the PC being faster, using PB was an exercise in patience. It had become horrible unresponsive while compiling, infact, all the YellowBox apps were really unresponsive when one was being used heavily. This was really only apparent with PB because PB was one of the few apps which was actually under load when the user wasn't using it, compiler and error output mainly, or GDB dumping a long backtrace. I soon added the second processor to my PC and suddenly PB and all the YellowBox apps were so much more responsive. The difference was tremendously dramatic, the apps went from locking up a lot to being extremely usable. I have been using this same machine and WOF 4.5 since without incident.

Jump to 2006. Using the Mini with Xcode is giving me flashbacks to that PC with the single processor. I've had Xcode go to lunch a few times, and when it does the whole system is just horribly unresponsive. Force Quit takes forever to come up, the system event queue seems to drop events like flies as I try to shut some windows and the whole situation just takes way too long to fix. In general, when one app is busy, the system is really bogged down in general. I have a dual G4 (450) which I use for personal stuff and it is just so much more responsive than the Mini despite being a much lower clock speed.

Something about Apple's UI demands a second processor. Granted having two processors is almost a doubling in performance, but so should a leap from 450Mhz to 1.4Ghz and it just doesn't happen when it comes down to responsiveness. Scheduling, priorities, something is not tuned enough to user interaction.

So until I have a dual processor machine, or all Mac's are dual processor. I am wishing for some sort of high priority key strokes. Hate to say it, but ctrl-alt-del is really nice to have on a very busy Windows machine, the task manager comes right up and I can terminate an offending program.

And I want the sound volume/mute keys high priority too. When Corey Taylor is screaming his brains out at volume 10, I'm compiling a huge framework and my wife walks in, I need to turn the volume down.

Right Now.

First Night Ice Sculpture