Floating point calculations

[pascalc]

Hi,

I am working on the Akai S2000 panel and it turned out that the tune parameter is coded in a signed floating point value of four nibbles…

For example:

-50.00 => 00 00 0E 0C

-10.00 => 00 00 06 0F

-0.10 => 07 0E 0F 0F

-0.01 => 0E 0F 0F 0F

0 => 00 00 00 00

0.01 => 02 00 00 00

0.10 => 09 01 00 00

10.00 => 00 00 0A 00

50.00 => 00 00 02 03

I have been searching quite a lot on the Internet for an algorithm in Lua that would translate this for me but could not find anything except to implement it by myself!

Now finally I found this topic on SO http://stackoverflow.com/questions/29592951/lua-hex-to-float that suggests using Lua 5.3.

My question is if there are any plans to upgrade Ctrlr to Lua 5.3 soon or if there is any other support for floating point calculations available that could save me the work of implementing the translation algorithm from scratch.

Thanks

/Pascal

[goodweather]

Hi, why not just copying the functions mentioned in http://stackoverflow.com/questions/18886447/convert-signed-ieee-754-float-to-hexadecimal-representation?

In my panels, I have implemented a series of functions (like trim and co…) in a method I called Miscellaneous.
You can call any function in Miscellaneous from any other method and the only thing to do is copy/paste…

I suspect that going to Lua 5.3 will mean a lot of work and tests for atom…

Anyway, very good topic (I mean the conversion hex-float)!

[atom]

i can have a look at 5.3 but i remember there were issues with luabind, the framework i’m using for c++/lua integration, but i’ll have a look. i wanted to do that and supoort LuaJIT also to get better performance. it’s on my todolist.

[atom]

could you also post that on github so i don’t forget ?

[pascalc]

Atom, Thanks! I have submitted an issue on Github now.

Goodweather, I had tried your suggested SO code but obviously not thoroughly enough. I’ll take a new look at it now. BTW, which panels have you made? I would like to have a look at them as reference.

BR
/Pascal

[pascalc]

Hi again,

After some more help from Stackoverflow I now have the two following algorithms:

function float2nibbles(value)
	local nibbles = MemoryBlock(4, true)
	local n = math.floor(math.abs(value) * 256 + 0.13)
	n = value < 0 and 0x10000 - n or n
	for pos = 0, 3 do
		nibbles:setByte(pos, n % 16)
		n = math.floor(n / 16)
	end
	return nibbles
end

function nibbles2float(memBlock, offset)
	local bi = BigInteger(0)
	bi:setBitRangeAsInt(0, 4, memBlock:getByte(offset))
	bi:setBitRangeAsInt(4, 4, memBlock:getByte(offset + 1))
	bi:setBitRangeAsInt(8, 4, memBlock:getByte(offset + 2))
	bi:setBitRangeAsInt(12, 4, memBlock:getByte(offset + 3))
	local n = 0
	for i = 0, 15 do
		local factor = math.pow(2, i - 8)
		n = n + bi:getBitRangeAsInt(i, 1) * factor
	end
	return memBlock:getByte(offset + 3) >= 0x8 and n - 256 or n
end

For more information you can read this thread on SO: http://stackoverflow.com/questions/38912827/convert-floating-point-values-to-hexadecimal

• This reply was modified 7 years, 8 months ago by pascalc.
• This reply was modified 7 years, 8 months ago by pascalc.

[Author]

Posts