lua-users home
lua-l archive

[Date Prev][Date Next][Thread Prev][Thread Next] [Date Index] [Thread Index]

Hi all,


I'm having trouble making accurate measurements of the impact of OOBIT
patch because in many cases it is so small that any fluctuation in
execution time can produce unexpected results - like the patched
version being slightly faster than original version (although it is
theoretically possible). I spent almost three weeks on it. Timeout!

My  platforms are Intel T7250 or Intel Q6600 both running Windows XP
32bits. I already tried in normal mode, safe mode and safe mode with
command prompt. In every mode, some variation did occur.

I would appreciate it if a kind soul would offer to run the benchmark
on a different platform. Of course, it will be necessary to patch Lua.
Please, send emails directly to me to not pollute the list.

In time:

1) The patch optionally may:
 add unquoted strings of form \string, so "test" == \test
 make separator optional inside tables "a = { 1 2 3 4 }" or "countries
= { \US \BR \FR }"
 disable the debug test applied on each LVM instruction interpreted
(performance improvement)
 add "!=" as alias of "~="

2) The method-token is defined in luaconf.h, so it is possible to
rename it as you wish. The default string for the method-token is now
"__method" to avoid collisions.

3) I think that the patch could be classified as beta quality.

4) All modifications that implemented OOBIT are indicated in code with
an OOBIT word on the line changed.

Thank you for your attention,