Re: In place modification/assignment - A Patch
[
Date Prev][
Date Next][
Thread Prev][
Thread Next]
[
Date Index]
[
Thread Index]
- Subject: Re: In place modification/assignment - A Patch
- From: Joshua Jensen <josh.jjensen@...>
- Date: 2011年6月27日 20:09:04 -0600
----- Original Message -----
From: Robert G. Jakabosky
Date: 6/27/2011 7:22 PM
On Monday 27, Joshua Jensen wrote:
It isn't just about typing. It's also about performance. When written
like:
player.ammo.shells = player.ammo.shells - 1
Lua must:
* Look up player.
* Look up ammo.
* Look up shells.
* Subtract 1.
* Look up player.
* Look up ammo.
* Set via hash the shells key to be the result of the subtraction.
That is almost right, the last two lookups come before the 'substract'
operation.
Opcodes:
1 GETGLOBAL 0 -1 ; player
2 GETTABLE 0 0 -2 ; "ammo"
3 GETGLOBAL 1 -1 ; player
4 GETTABLE 1 1 -2 ; "ammo"
5 GETTABLE 1 1 -3 ; "shells"
6 SUB 1 1 -4 ; - 1
7 SETTABLE 0 -3 1 ; "shells" -
Ah, right. I wrote the majority of the email without a build of Lua
available to me. I kind of knew I'd end up getting it wrong.
When written as "player.ammo.shells -= 1", Lua must:
* Look up player.
* Look up ammo.
* Look up shells.
* Subtract 1.
* Directly store the result.
Hooray! (I hope this is how the patch works, anyway...)
The new syntax saves two lookups(one global, one table) in this example.
Opcodes:
1 GETGLOBAL 0 -1 ; player
2 GETTABLE 0 0 -2 ; "ammo"
3 GETTABLE 1 0 -3 ; "shells"
4 SUB_EQ 1 -4
5 SETTABLE 0 -3 1 ; "shells" -
The last operation is still a normal "SETTABLE". No special support for
modifying a table slot directly, doing so wouldn't work with tables/userdata
that have __index/__newindex metamethods.
Okay, I understand why the SETTABLE would still be used, although I
certainly like the idea of the performance gained by not needing the
SETTABLE.
Thank you for clarifying. I really should have checked first.
-Josh