|
|
Atari coding BBS
| Re: DSP frac arithmetic help |
|---|
|
Posted by: dml
|
Feb,27.2015-15:28
|
Hi,
Yes the multiplier is fractional, and assumes your sources are denormalized by 1 bit, which gets accounted for during multiplying. That's why you end up with 2x the expected result when multiplying integers.
So:
$400000 = 0.5
$200000 = 0.25
$7fffff = 0.99999
However you'll notice that it's not possible to represent the value 1.0 exactly.
There is a workaround in many situations, which is to work with *signed* sources, since it is possible to represent -1.0. Where necessary, use the special signing version of the multiplier to work with tables of -ve numbers.
e.g.
mpy -x0,y0,a
where:
x0 == -1.0 == $800000
Just be prepared to get slightly confused again when you try to square a fraction :)
Hope that helps.
|
[All messages in this thread] [Start new thread]
|
Topic
|
Posted by
|
Date
|
|
DSP frac arithmetic help
|
Laurent / Thadoss
|
Feb,27.2015-15:03
|
|
Re: DSP frac arithmetic help
|
Laurent / Thadoss
|
Feb,27.2015-15:08
|
|
Re: DSP frac arithmetic help
|
dml
|
Feb,27.2015-15:28
|
|
Re: DSP frac arithmetic help
|
Laurent / Thadoss
|
Feb,27.2015-18:32
|
|
Re: DSP frac arithmetic help
|
dml
|
Feb,27.2015-18:42
|
|
Re: DSP frac arithmetic help
|
dml
|
Feb,27.2015-18:46
|
|
Re: DSP frac arithmetic help
|
Splash/S1
|
Feb,27.2015-20:21
|
|
Re: DSP frac arithmetic help
|
dml
|
Feb,27.2015-20:38
|
|
Re: DSP frac arithmetic help
|
Laurent / Thadoss
|
Feb,27.2015-20:42
|
|
Re: DSP frac arithmetic help
|
mikro
|
Feb,28.2015-10:44
|
|
Re: DSP frac arithmetic help
|
dml
|
Feb,28.2015-12:35
|
What's the anti-troll code?
That's your personal code to be able to add comments and messages on the dhs.nu site.
Don't have a code or forgot it? Fix it here.
|
