In this case, the stupid programmer is me :-(
For the last week or so I've been working (off and on) on tracking down the difference in query optimization between jSuneido and cSuneido that I mentioned earlier.
Along the way I found and fixed a few minor bugs in cSuneido, which was good. One of them was caused by re-using an instance variable for two different purposes. Of course, this eventually caused a problem. I can't remember when I did this, but it seems incredibly short sighted. It was probably just the "easiest" thing at the time. Sigh.
You'd think by now we would have run into any bugs in cSuneido. But when the only affect is a slight difference in optimization, it's not at all obvious.
To find the optimization difference between cSuneido and jSuneido I ended up having to insert debugging into both cSuneido and jSuneido and compare internal intermediate results until I found where the difference was coming from.
As usual, it turned out to be something stupid. When I was porting the code from C to Java I had made a "minor" improvement. I don't even remember doing it. But that minor change was enough to throw things off. It wouldn't have been so bad if I'd made the same change in the C code, but I didn't. When will I learn that I shouldn't change things during porting!
Oh well, one more down, a bunch more to go.
No comments:
Post a Comment