Archive for September, 2007

Colorado Autumn

A few days ago we went up to Estes with my folks to hear the elk bugle. On the way home we got to see two bull elk fighting over a harem. We were close enough to hear their antlers clash.

Then yesterday Elyn and I went up to Brainard Lake and finally made it to the summit of Mt Audubon. Awesome views up there. On the way home we stopped near Nederland to see the aspen — they are turning and this weeked is the peak; if you live in Colorado you should go up now.

No photos, but maybe it is better this way.

Fan-out

Most of my big projects seem to go through a “fan out” phase, where the to-do list grows very quickly. This is a scary phase of the project — one where it feels as though things are accelerating out of control, and where success seems as though it could be in doubt. The compiler seems to be entering this phase now.

While this phase feels dangerous, I think the more dangerous phases come earlier, when I’m more confident.

Do you experience fan-out? I’m curious whether this is an artifact of how I approach projects, or whether it is a more universal experience.

First Success

I just managed to get through an ordinary make of a real program (zenity, which for some reason is my standard test) using gcc as a server. Yay! It is, as predicted, about twice as fast as an ordinary make.

Now pile on the caveats.

Compiler Progress

I’ve been working steadily on the incremental compiler project these last couple of months, but not blogging about it. So, time for a status report.

I made a branch in the GCC repository, cleaned up my existing patches, and checked it all in. There’s a short web page describing the project on the GCC wiki (branch details there).

I’ve also written a rough draft of the support code for “decl smashing”, and also changed the parser to consider “per-declaration hunks”.

The idea behind per-declaration hunks is that, for code written by the user, we get better incremental behavior if we consider each hunk to hold a single declaration or definition, rather than the previous rule of “everything between two file change events”. The former is better because it means that a change is isolated to its containing declaration and uses of that declaration (and we can do even better here, perhaps, with a bit more analysis); the latter is better for system headers and the like because it has lower overhead.

Finally, I’ve started server-izing GCC. This part should be ready for real testing pretty soon.

Dessert Sushi

dessert sushi

We’ve been making (vegetarian) sushi lately and a month ago or so I had the idea to make dessert sushi. “I’m a top-chef-worthy genius,” I thought. “Gordon Ramsay will never call me a donkey!”

Of course then I googled and found that only a couple million people did this first. Then when he was visiting, Manu bought us a veggie sushi book and it has a whole section on dessert sushi. On my birthday we made some; the above photo shows a few — some with mango and kiwi, some with chocolate shavings and raspberry, and some with chocolate sauce.

Death at a Funeral

The title of this movie brought to mind the title of Chris Shera’s PhD thesis, “listening to the ear”.

There were a couple moments where this film appeared ready to veer into the dark and disturbing. But, luckily (and I hope this doesn’t spoil it for you — perhaps someone enjoys that feeling), it did not. Also I did not like the use (I counted twice) of the phrase “a gay”. Somehow that sounds mildly homophobic to me, a fact I find somewhat curious and interesting, but that thought nevertheless doesn’t render the phrase more acceptable.

Anyway, overall it was pretty funny, and one scene at least was hilarious. I laughed out loud a few times.