PropBasic - Multiply

camelot2

hi, I am trying to calculate frqa for a pulse generator project. The frequency I am trying to program is 2222Hz, the sysfreq is 100_000_000Hz. The formula I am using is:
frqa = myfreq X (2^32/sysfreq) = 2222 X 42.94967296 = 95434.17331712 and rounded frqa = 95434. But the program gives 93321 !!!!!
Attached is the PropBasic program. WHAT AM I DOING WRONG ?
thanks for your help - Dave

Martin_H

Please post your code as a compiler/interpreter can either truncate the float or promote the integer depending upon how the code is written. I would bet you are mixing integer and floating point operations and getting bit by a truncation error in an intermediate step of the calculation. It's happened to me more times than I can count over the years and has been a source of numerous obscure bugs.

Update: Sorry didn't see the attachment. After looking at it, this is the math you are trying to do:

Result = 42 * 2222 + 0.94967296 * 2222
Result = 93324 + 2110 (Actually 2110.17331712 but truncated to an integer)
Result = 95434

You are computing the whole quantity correctly, but the fractional compution is likely wrong and is coming out negative. Likely because of this step:

value_dec = value ** 4078814305

In reading the prop basic manual I think that code looks good and I'm not sure what is wrong with it. If you just do that operation what value do you get?

Bean

In the code you have

 'multiply 2222 X 42.94967296 from Win calculator
 '0.94967296 x 65536 x 65536 = 4078814305.09551616
  value = 42
  value_dec = value ** 4078814305
  value_whole = 2222 * value
  value_total = value_whole + value_dec

I think you want

 'multiply 2222 X 42.94967296 from Win calculator
 '0.94967296 x 65536 x 65536 = 4078814305.09551616
  value = 2222
  value_dec = value ** 4078814305 ' 2110
  value_whole = value * 42 ' 93324
  value_total = value_whole + value_dec ' 95434

Bean

camelot2

thank you both for the replies.
Bean, I replaced my code with your code, see attached
results:
value_whole = 93324
value_dec = 4294967184
value_total = 93212

It still is not giving the correct result.
Any thoughts as to why ? thanks for helping

Bean

Oh, I see the problem. The value 4078814305 is seen as a negative value (since bit 32 is set) by the multiply routine.

You'll have to use half the value, then shift the result.

  value = 2222
  value_dec = value ** 2039407152 ' 1055
  value_dec = value_dec * 2 ' 2110
  value_whole = value * 42 ' 93324
  value_total = value_whole + value_dec ' 95434

Bean

Bean

You could also shift the original value up, then do the ** with the constant shifted down the same amount.

  value = 2222
  value_total = value << 7 ' (* 128)
  value_total = value_total ** 1441151881 '  (42.94967296 / 128) * 65536 * 65536

Bean

pedward

ugh, so much easier in SPIN:

FRQA := 2222 / CLKFREQ << 32

Well, that's algebra for ya! (yeah, sorry, wrong!)

Bean

pedward wrote: »

ugh, so much easier in SPIN:

FRQA := 2222 / CLKFREQ << 32

Hmmm, I don't think it is quite that easy... Try it.

Bean

camelot2

thanks Bean, it now WORKS.
Last question - How do I know if a value is seen as a negative value ?

pedward

Bean wrote: »

Hmmm, I don't think it is quite that easy... Try it.

Bean

Doh, you read my reply before I thought about it too hard. Yeah, you caught me using Algebra!

Bean

If it has bit 32 set, it is seen as negative.
> $7FFF_FFFF or 2147483647

Bean

pedward

Bean wrote: »

If it has bit 32 set, it is seen as negative.
> $7FFF_FFFF or 2147483647

Bean

Just to clarify some fine points, if you start with a small value and want to bit shift it above $7FFF_FFFF, it will not consider sign, you will just end up with the bits shifted (which will look funny when interpreted as a signed value).

I do wish the prop had the ability to specify unsigned in calculations, when using SPIN.