one of those inherits from IRootFunctions and the other doesnt, i dont remember which (and the one t

one of those inherits from IRootFunctions and the other doesnt, i dont remember which (and the one that does inherits from the other)

Maxine•10/20/23, 9:36 PM

Kkhitiara one of those inherits from IRootFunctions and the other doesnt, i dont remember ...

Maxine•10/20/23, 9:36 PM

hmm interesting

Kkhitiara you might need to use IFloatingPointIEEE754 instead of IFloatingPoint btw

Aqua•10/20/23, 9:37 PM

yep you're right only the ieee754 implements IRootFunctions

khitiaraOP•10/20/23, 9:37 PM

yeah theres a whole hierarchy of the interfaces with varying granularity

khitiaraOP•10/20/23, 9:37 PM

for instance ITrigonometricFunctions, IHyperbolicFunctions, etc

khitiaraOP•10/20/23, 9:37 PM

ok cool i was remembering right

khitiaraOP•10/20/23, 9:39 PM

the root functions etc are only guaranteed for binary floats where it can be done fast

khitiaraOP•10/20/23, 9:40 PM

    public interface IFloatingPointIeee754<TSelf>
        : IExponentialFunctions<TSelf>,
          IFloatingPoint<TSelf>,
          IHyperbolicFunctions<TSelf>,
          ILogarithmicFunctions<TSelf>,
          IPowerFunctions<TSelf>,
          IRootFunctions<TSelf>,
          ITrigonometricFunctions<TSelf>
        where TSelf : IFloatingPointIeee754<TSelf>?

    public interface IFloatingPointIeee754<TSelf>
        : IExponentialFunctions<TSelf>,
          IFloatingPoint<TSelf>,
          IHyperbolicFunctions<TSelf>,
          ILogarithmicFunctions<TSelf>,
          IPowerFunctions<TSelf>,
          IRootFunctions<TSelf>,
          ITrigonometricFunctions<TSelf>
        where TSelf : IFloatingPointIeee754<TSelf>?

khitiaraOP•10/20/23, 9:41 PM

and then theres IBinaryFloatingPointIeee754 which adds bitwise, ispow2, allbitsset, and log2

Perksey•10/20/23, 9:41 PM

who's working on what atm

Perksey•10/20/23, 9:41 PM

this channel seems exciting all of a sudden

khitiaraOP•10/20/23, 9:41 PM

idk

khitiaraOP•10/20/23, 9:41 PM

a long discussion is happening about generic math things

khitiaraOP•10/20/23, 9:41 PM

in which petre grifon has seemingly solved our issue with length

Aqua•10/20/23, 9:42 PM

i can recommend not requiring the

IFloatingPointXYZ<>

IFloatingPointXYZ<>

interface directly, instead only requiring

IRootFunctions<T>

IRootFunctions<T>

ITrigonometricFunctions<T>

ITrigonometricFunctions<T>

etc

Aqua•10/20/23, 9:42 PM

it will make your life easier

khitiaraOP•10/20/23, 9:42 PM

mhm

khitiaraOP•10/20/23, 9:43 PM

so theres one last concern and thats the problem of converting

khitiaraOP•10/20/23, 9:43 PM

which i dont think we have an easy generic constraint for

Aqua•10/20/23, 9:43 PM

as in casting from one

VectorX<T>

VectorX<T>

type to another

VectorX<T>

VectorX<T>

type?

Aqua•10/20/23, 9:43 PM

with a different type for

Perksey•10/20/23, 9:44 PM

we need to manually specialize those anyway seeing as you can't have variable Ts on implicit operators

khitiaraOP•10/20/23, 9:44 PM

wait hang on

PPerksey we need to manually specialize those anyway seeing as you can't have variable Ts...

Aqua•10/20/23, 9:44 PM

Perksey•10/20/23, 9:44 PM

tbh i'm really quite keen on the extension method approach, it probably makes sense to only require on the type itself the interface containing +-*/

Aqua•10/20/23, 9:44 PM

problem solved

khitiaraOP•10/20/23, 9:45 PM

yeah i had forgotten about CreateChecked

Aqua•10/20/23, 9:45 PM

CreateChecked

CreateChecked

is your best friend

Aqua•10/20/23, 9:45 PM

i forgot stuff like

float.CreateChecked

float.CreateChecked

existed in my older generic math implementations

Aqua•10/20/23, 9:45 PM

so i used

Convert.ChangeType

Convert.ChangeType

or whatever the similar function is called

khitiaraOP•10/20/23, 9:46 PM

and by forgotten i mean i didnt even know it existed until just now

Aqua•10/20/23, 9:46 PM

but

CreateChecked

CreateChecked

is much better, i use it when providing cast operators between system.numerics and my generic types

Maxine•10/20/23, 9:46 PM

i'm using CreateTruncating for speed

Maxine•10/20/23, 9:46 PM

but i can use checked if desired

Aqua•10/20/23, 9:47 PM

personally i'd recommend

CreateChecked

CreateChecked

simply because it reduces the risk of undefined behaviour

Aqua•10/20/23, 9:47 PM

so if someone tries to cast a

VectorX<int>

VectorX<int>

with an X value of 30000 into a

VectorX<byte>

VectorX<byte>

it will error, just like how that would normally happen

Perksey•10/20/23, 9:47 PM

you do realise we have checked operators now

Perksey•10/20/23, 9:47 PM

we can do both

Aqua•10/20/23, 9:47 PM

oh cool

Maxine•10/20/23, 9:47 PM

oh yeah cool

Maxine•10/20/23, 9:56 PM

GenericVector3.cs34.15KB

Maxine•10/20/23, 9:57 PM

initial impl, no extension methods

Maxine•10/20/23, 9:57 PM

rough around the edges

Maxine•10/20/23, 9:57 PM

lemme polish it...

Aqua•10/20/23, 10:00 PM

remember that

T.Zero

T.Zero

and

T.One

T.One

exists

Maxine•10/20/23, 10:00 PM

yeah that is what i meant by polishing

Aqua•10/20/23, 10:00 PM

ah ok

Maxine•10/20/23, 10:00 PM

how tf you meant to implement ISpanFormattable

Maxine•10/20/23, 10:10 PM

hacked this together

Maxine•10/20/23, 10:11 PM

if you don't like goto, i can rewrite it to use breakbreak instead :)

public interface IFloatingPointIeee754<TSelf> : IExponentialFunctions<TSelf>, IFloatingPoint<TSelf>, IHyperbolicFunctions<TSelf>, ILogarithmicFunctions<TSelf>, IPowerFunctions<TSelf>, IRootFunctions<TSelf>, ITrigonometricFunctions<TSelf> where TSelf : IFloatingPointIeee754<TSelf>?

one of those inherits from IRootFunctions and the other doesnt, i dont remember which (and the one t

Similar Threads