I was wondering if anybody has tried Magic Lantern's free software, and how well does it work.

I took a look at the details and what's happening is that different lines of the sensor array are read out with different gain amplification. Basically you read out two lines via an ISO 100 gain amp and the next two lines via an ISO 1600 gain amp. You can do this with the 7D and 5D3 because of the hardware involved (8 channel readout and two gain amplifiers). You couldn't do it with a 70D for example.

So you get one image (with 50% resolution) that has limited shadow detail (ISO 100) combined with one image one image (with 50% resolution) with good shadow detail, but highlights blown out by 4 stops (ISO 1600). If you cleverly seperate, process and recombine these partial images you can blend them into a single image with good highlight detail (from the ISO 100 shot) and good shadow detail (from the ISO 1600 shot). The downside is that you only have 1/2 the resolution in the areas that depend on information from only the ISO 100 or only the ISO 1600 shot. In the mid range where both shots have information you can pretty much get back to standard resolution. However there are also potential banding and moire problems with this technique.

I guess that when you actually NEED high dynamic range and you have a moving subject so you can only get one shot, this technique should be useful. If you have a static subject and use a tripod, standard HDR techniques will work better. If you don't need 12+ stops of dynamic range (and most of the time you don't), then using this technique doesn't sound like a good idea.

I can see why Canon didn't use it, but it seems like a potentially useful technique under specific circumstances that require expanded dynamic range.