PDA

View Full Version : Problems negating a number

reppig
10-18-2011, 07:05 PM
When I run the following code cos_lon evaluates correctly to 0.1618761.
However, neg_cos_lon evaluates to -27.27986 not -0.1618761 which is what I want??
I also tried what was in the manual -neg_cos_lon and I got the same wrong answer.

neg_cos_lon := -cos_lon

DBG.Str(FS.FloatToString(cos_lon))
DBG.Str(STRING(" "))
DBG.Str(FS.FloatToString(neg_cos_lon))

Phil Pilgrim (PhiPi)
10-18-2011, 07:10 PM
You're negating cos_lon, a floating-point number, in the integer domain. That does not work. You need to negate it (or subtract it from 0.0) using a floating-point method.

-Phil

Bobb Fwed
10-18-2011, 07:44 PM
Or just flip bit 31?

Like:
neg_cos_lon := cos_lon ^ \$8000_0000

Much faster than a floating point operation, and I can't think of a scenario where it would give you incorrect results (though I'm no expert on floating point).

reppig
10-18-2011, 08:09 PM
I understand what you are saying and it makes sense. The manual implied it just changed the sign of a number. So, should I assume everything the manual says only apply to integers??

reppig
10-18-2011, 08:10 PM
I will give it a try tommorrow. Thanks.

Or just flip bit 31?

Like:
neg_cos_lon := cos_lon ^ \$8000_0000

Much faster than a floating point operation, and I can't think of a scenario where it would give you incorrect results (though I'm no expert on floating point).

reppig
10-18-2011, 08:11 PM
I understand what you are saying and it makes sense. The manual implied it just changed the sign of a number. So, should I assume everything the manual says only apply to integers??

You're negating cos_lon, a floating-point number, in the integer domain. That does not work. You need to negate it (or subtract it from 0.0) using a floating-point method.

-Phil

Phil Pilgrim (PhiPi)
10-18-2011, 08:14 PM
The manual implied it just changed the sign of a number. So, should I assume everything the manual says only apply to integers??
As it pertains to native Spin operations, yes.

-Phil

Bobb Fwed
10-18-2011, 08:33 PM
The manual implied it just changed the sign of a number. So, should I assume everything the manual says only apply to integers??
As already said, yes, everything is assumed as integers.

The problem with negating a float, when it thinks it is a integer is because the bit structure for an integer of negative and positive 5 (for example) is completely different from each other (%11111111111111111111111111111011 vs %00000000000000000000000000000101), most of the bits get changed. Where as a floating point number, the difference is very simple (%11000000101100011100001010010000 vs %01000000101100011100001010010000), just bit 31 changes.