Int64 to Float?

[Andy Li]

Hi,

It seems to me strange that Int64 class does not have a toFloat method.

So how should I convert a Int64 number to Float, considering it could be outside the 32-bit range?

Best regards,

Andy

[Justin L Mills]

Andy

We should add a Float64 to Haxe.

Best Justin


Andy Li:

[Andy Li]

The reason I mentioned the value of Int64 could be outside the 32-bit range, is that we cannot use toInt:

    var float:Float = Int64.toInt(int64); //toInt will fail, because Int is 32-bit only

Float is a floating point number which is already able to store the whole range of Int64, no matter using single/double precision.

And actually all the current Haxe targets use double precision floating point number for Float, which is 64-bit.

Thanks for your reply though :)

Best regards,

Andy

--
To post to this group haxe...@googlegroups.com
http://groups.google.com/group/haxelang?hl=en
---
You received this message because you are subscribed to the Google Groups "Haxe" group.
For more options, visit https://groups.google.com/d/optout.

[Nicolas Cannasse]

Le 30/04/2014 12:02, Andy Li a écrit :
> The reason I mentioned the value of Int64 could be outside the 32-bit
> range, is that we cannot use toInt:
> var float:Float = Int64.toInt(int64); //toInt will fail, because
> Int is 32-bit only
>
> Float is a floating point number which is already able to store the
> whole range of Int64, no matter using single/double precision.
> And actually all the current Haxe targets use double precision floating
> point number for Float, which is 64-bit.

That's not true, since even in double precision you will only get 52
bits of fraction, so you will lose some lower bits precision compared to
Int64.

Best,
Nicolas

[Nicolas Cannasse]

Le 30/04/2014 12:02, Andy Li a écrit :

> The reason I mentioned the value of Int64 could be outside the 32-bit
> range, is that we cannot use toInt:
> var float:Float = Int64.toInt(int64); //toInt will fail, because
> Int is 32-bit only
>
> Float is a floating point number which is already able to store the
> whole range of Int64, no matter using single/double precision.
> And actually all the current Haxe targets use double precision floating
> point number for Float, which is 64-bit.

For the same reason there can be lose of precisions when multiplying two
Int32 stored as double precision floats.

Best,
Nicolas

[Andy Li]

Thanks for the explanation :)

If I allow lose of precision, what is the best way to convert Int64 to Float?

Best regards,

Andy

[Nicolas Cannasse]

Le 30/04/2014 12:14, Andy Li a écrit :
> Thanks for the explanation :)
> If I allow lose of precision, what is the best way to convert Int64 to
> Float?

lowInt32 + highInt32 * 2^32 ?

Best,
Nicolas

[David Elahee]

There is no perfect answer, i propose some not so crappy one ...

var a :Float =  0 

a += int64.low;

a += float(int64.high) * float(1<<32) 

You have to use less multiply and use more power of 2 floats  as possible me think...

--
David Elahee

[David Elahee]

pwned by Nicolas...

--
David Elahee