freenode/#clasp - IRC Chatlog
Search
5:55:25
beach
Bike: You can almost always avoid the test for an &optional argument if you write one or two functions without &optional, and have the compiler macro expand to a call to one of them.
5:56:18
beach
Then, even if inlining works only for functions without &optional, you can benefit from inlining anyway.
13:50:58
Bike
the function in question is part of someone else's library, is the thing... i mean we could define one anyway
13:59:18
beach
I suggested it because you hinted that inlining created a big LLVM code. If it works fine, then forget what I said. I must have misunderstood something.
13:59:47
Bike
Oh, if you mean when I was talking about dot files yesterday that was just because it's a big lisp function
14:00:34
beach
I see. I seem to remember that it had to do with inlining functions with &optional parameters. I must be misremembering.
14:07:56
Bike
sure, but i thought the whole philosophy was we'd inline things and then cut out irrelevant parts, rather than e.g. acting like sbcl where it replaces known function calls with computed specialized bodies