 |
lua-l archive |
|
|
||
|
On 9/21/2010 8:29 AM, Ryota Hirose wrote:
Hello Lua Hackers,
2010/9/21 KHMan wrote:
So, source line 1 was compiled to 7 lines binary, line 3 was 8
lines, line 5 was 9 lines.
Look closely, why are you counting the RETURN?
I'm sorry, it's my mistake. Last RETURN becomes from whole chunk,
not a single source line. Exactly, source line 1 was compiled to
7 lines, line 3 and 5 were 8 lines.
So this is a 'win'? Colour me unimpressed.
The difference between line 1 and 3, a ternary operator and
and/or, is little. and/or needs one more TEST line (And, this
TEST line is what I want to eliminate by a ternary operator...).
There would not be any noticeable performance difference in normal programs. I think it's better to stick to a discussion on syntax rather than tout a case of one or two instructions saved.
I doubt adding syntax here and there for these reasons is good programming language design.
But the difference between line 1 and 5, a ternary operator and if-statement, is big. With a ternary operator, a compiler calculated left operand of assignment only once. But with if-statement, the left operand must be compiled twice. When the left operand is complex, this difference became bigger.
If by 'left operand' you mean the LHS of 'z'? I can't remember any complex variations for assignment other than a list of identifiers.
But look, only one path is executed -- nothing is 'saved' during execution. There would be negligible performance difference.
16 [5] GETGLOBAL 0 -2 ; a 17 [5] TEST 0 0 0 18 [5] JMP 3 ; to 22 19 [5] GETGLOBAL 0 -3 ; b 20 [5] SETGLOBAL 0 -6 ; z 21 [5] JMP 2 ; to 24 22 [5] GETGLOBAL 0 -4 ; c 23 [5] SETGLOBAL 0 -6 ; zIf I were using my VIC-20, then I would applaud such measures of saving an instruction here and there. But with Lua, I don't think these can be used as very good reasons in support of your proposed syntax.
-- Cheers, Kein-Hong Man (esq.) Kuala Lumpur, Malaysia