Marc, would it be easy to make it so that negative constants in the AST (eg: -1) show up as negative number constants, instead of showing up as the negation operator applied to a positive constant? It's posing me a minor issue at the IIR level.
- Maxime
Afficher les réponses par date
On 2010-12-17, at 5:13 PM, Maxime Chevalier-Boisvert wrote:
Marc, would it be easy to make it so that negative constants in the AST (eg: -1) show up as negative number constants, instead of showing up as the negation operator applied to a positive constant? It's posing me a minor issue at the IIR level.
I would think the constant folder would easily take care of that. JavaScript (like many languages) does not have negative constants. The parser should not make such a transformation because we might use the parser for other things (for example program analysis where we care that "-" is actually an executed operation).
Here's an interesting related issue. In a programming language with fixed precision integers, and signed integer constants, how can the most negative integer be parsed? For example on a 32 bit machine, the most negative integer is -2147483648. But that integer can't be generated as -(2147483648) because 2147483648 does not fit in a signed 32 bit integer. The trick is to accumulate the number as a negative number, i.e. for each digit d, the computation is n = n*10 - d. So we get:
n = 0 n = -2 n = -21 n = -214 n = -2147 n = -21474 n = -214748 n = -2147483 n = -21474836 n = -214748364 n = -2147483648
So the number always fits in a 32 bit signed integer. To share the code for the positive and negative cases, the negative value is always computed, and if the sign was *not* "-" then the negation is computed (an exception is raised if the number was -2147483648, because the number +2147483648 does not fit in a 32 bit signed integer).
This trick can be used in other contexts where a greater range of integers is required.
Marc
The constant folder takes care of it, it's just my IIR hacks operating directly on the AST that don't at the moment. In one place (iir.constant) I'm expecting to see an integer constant, but I get the result of an operation instead. I'll have to either add a special case to those IIR special forms, or transform negative integer constants in the AST->IR translation.
- Maxime
On 10-12-17 05:44 PM, Marc Feeley wrote:
On 2010-12-17, at 5:13 PM, Maxime Chevalier-Boisvert wrote:
Marc, would it be easy to make it so that negative constants in the AST (eg: -1) show up as negative number constants, instead of showing up as the negation operator applied to a positive constant? It's posing me a minor issue at the IIR level.
I would think the constant folder would easily take care of that. JavaScript (like many languages) does not have negative constants. The parser should not make such a transformation because we might use the parser for other things (for example program analysis where we care that "-" is actually an executed operation).
Here's an interesting related issue. In a programming language with fixed precision integers, and signed integer constants, how can the most negative integer be parsed? For example on a 32 bit machine, the most negative integer is -2147483648. But that integer can't be generated as -(2147483648) because 2147483648 does not fit in a signed 32 bit integer. The trick is to accumulate the number as a negative number, i.e. for each digit d, the computation is n = n*10 - d. So we get:
n = 0 n = -2 n = -21 n = -214 n = -2147 n = -21474 n = -214748 n = -2147483 n = -21474836 n = -214748364 n = -2147483648
So the number always fits in a 32 bit signed integer. To share the code for the positive and negative cases, the negative value is always computed, and if the sign was *not* "-" then the negation is computed (an exception is raised if the number was -2147483648, because the number +2147483648 does not fit in a 32 bit signed integer).
This trick can be used in other contexts where a greater range of integers is required.
Marc
Tachyon-list mailing list Tachyon-list@iro.umontreal.ca https://webmail.iro.umontreal.ca/mailman/listinfo/tachyon-list
-
Marc Feeley -
Maxime Chevalier-Boisvert