2008-06-24
| 08:50 | cemerick | Is there a function or var available that contains the path to the current file? |
| 08:50 | cemerick | (current file being loaded, I should say) |
| 09:18 | rhickey | cemerick: not officially - (.get clojure.lang.Compiler/SOURCE_PATH) will work |
| 09:20 | arbscht | cgrand: thanks for the correction :) |
| 09:22 | cemerick | rhickey: Hrm. I think I'll skip that for now. FYI, I'm setting up a build process so that gen-and-save-class forms can always use a consistent output path, regardless of where the .clj files are. For now, I'll just always use "." for the output path, and run that process at the package root. |
| 09:22 | cemerick | Having a clojure/*file* var that files can reference as they're being loaded would be super-handy. :-) |
| 09:22 | rhickey | yes |
| 11:15 | cgrand | rhickey: about reduce, what I still doesn't understand is that, sometimes, the seq wasn't retained -- hotspot JIT voodoo? |
| 11:16 | cgrand | for 10M lines: |
| 11:16 | cgrand | real 3m33.048s |
| 11:16 | cgrand | user 33m52.490s |
| 11:16 | cgrand | sys 0m53.788s |
| 11:17 | rhickey | cgrand: yes, I think hotspot sometimes optimized it away |
| 11:18 | rhickey | are those good 10M line times? |
| 11:20 | cgrand | slightly worse than the imperative version I think (let me check) |
| 11:20 | rhickey | how many procs? |
| 11:22 | cgrand | proc usage is never precise -- only someone with root access could tell. (8 procs with 4 hardware threads) |
| 11:31 | cgrand | imperative on 10M: |
| 11:31 | cgrand | real 3m11.918s |
| 11:31 | cgrand | user 32m35.173s |
| 11:31 | cgrand | sys 0m44.201s |
| 11:41 | cgrand | one lesson so far is to always run with -XX:+UseParallelOldGC on multiproc hardware without that full GCs runs only on one CPU and the others are sleeping |
| 12:09 | blackdog | what does this mean? Caused by: java.lang.Exception: Can't let qualified name |
| 12:11 | blackdog | nevermind, it's in a macro, i was confoozed |
| 16:12 | Lau_of_DK | Hey guys |