currently editing the binder json
currently editing the binder json
unknown.png94.67KB
D48 minutes 
PI’m working on it haha
DBtw, when you make the preview 4 packages make sure to move the unshipped api tracking to the shipped.
Pwill do
P...........
Pwe forgot Rectangle.FromLTRB
Pand probably wouldn't hurt to add Rectangle.X and Y
Pwouldn't do that for the boxes though
Ponce again i have fallen victim to dylan time
Palways underestimating how long things take
Pit'll be tomorrow now
Pi still need to
a) add OpenGL manual overloads for Silk.NET.Maths
b) look at input
c) rework the SDL event pump
a) add OpenGL manual overloads for Silk.NET.Maths
b) look at input
c) rework the SDL event pump
Phowever, windowing is fully ported over
Pand so are the bindings (bar the manual overloads)
DI'm guessing LTRB is
Pnah it’s (l, t), (r - l, b - t)
P@Deleted User
DOh I fucked second vector... Can you just change it quickly? My sister is moving right now so I'm not at home
Pwill do in a bit
Poh no
Pi did not think the ops files through
P@Deleted User our vector ops conflict with system.numerics
Pconsidering that i still want to use system.numerics wherever possible because they have the best perf, this could be a problem
Pshould we scrap the ops idea and move the static methods back into the generic classes?
Palso @tannergooding how much of a backlog is there for .net 6?
Phow many years are we talking?
Pmaybe we can propose and contribute some stuff from Silk.NET.Maths back
DNo - just use our maths. Perf will be fixed by preview5
Pi'm sorry but you will never be able to match system.numerics' perf
Pand one of our larger users was very adamant about not wanting to use custom maths types
Pso if we're going the maths library route we must not force it upon people
TAre you asking for (which includes things like , , etc) in particular?
Pyeah
Pi'm just trying to get a gage of "if i were to open well-constructed, proper proposals for the maths stuff akin to Silk.NET.Maths's current surface api, how long will they be sitting around for?"
TSo, basically, there exists two types of "work"
top-down work: which is work management tells me to do because it's important to the company's interests
bottom-up work: which is work I tell management is important because I think its important for the continued improvement of an area
top-down work almost always takes priority; but I always have time to work on and try to push through some bottom-up work as well
in .NET 5, ARM64 Intrinsics were top-down and bottom-up. Most of the other improvements were bottom-up work exclusively
Starting with .NET 6, we are trying to do more planning in the open: https://themesof.net/
We currently have 106 issues in the System.Numerics area, several of which are just things I need to triage/close: https://github.com/dotnet/runtime/issues?q=is%3Aopen+is%3Aissue+label%3Aarea-System.Numerics+
I now have a second co-owner for Numerics that is interested in picking up and helping various things so that will help with some items
So what gets through for bottom-up work is mostly up to me and is likely limited to about the same amount of bottom-up work that happened for .NET 5
top-down work: which is work management tells me to do because it's important to the company's interests
bottom-up work: which is work I tell management is important because I think its important for the continued improvement of an area
top-down work almost always takes priority; but I always have time to work on and try to push through some bottom-up work as well
in .NET 5, ARM64 Intrinsics were top-down and bottom-up. Most of the other improvements were bottom-up work exclusively
Starting with .NET 6, we are trying to do more planning in the open: https://themesof.net/
We currently have 106 issues in the System.Numerics area, several of which are just things I need to triage/close: https://github.com/dotnet/runtime/issues?q=is%3Aopen+is%3Aissue+label%3Aarea-System.Numerics+
I now have a second co-owner for Numerics that is interested in picking up and helping various things so that will help with some items
So what gets through for bottom-up work is mostly up to me and is likely limited to about the same amount of bottom-up work that happened for .NET 5
TFor .NET 6, my current primary focus for bottom-up work is:
- Finalizing a proposal up for math vs math
- Finalizing a proposal for consistent across the various OS/Architecture combinations
- Implement
Vector2/3/4<T>
TBut, those are mostly "big" work tickets and I'll be pushing through a range of smaller work items as possible; like the approved and will be commited in Jan when I get back from vacation
TI'd also currently like to get what exists in up in as well, so certain simple ops can be cross-platform and reduce the amount of trivial code that gets duplicated
Pthis sounds awesome
TIf you have proposals for other things, I can certainly prioritize them as appropriate; but that somewhat depends on the proposal 
Pok cool, basically i'm thinking of opening a series of proposals for System.Numerics.Vectors so that we can contribute back some of the stuff introduced in our new maths library
Pbecause the fragmentation in the c# maths space is ridiculous, i basically just want to make system.numerics.vectors a little more complete
Pseeing as the generic vectors originally took a half slot, i'm not sure how quickly they'll be able to be reviewed though
TVectors took a half slot namely because it was a like 10 types covering several concepts and required some design explanation, etc
Individual or small groups of proposed APIs tend to go much quicker
Individual or small groups of proposed APIs tend to go much quicker
TIn particular, I'd like to extend the numerics types to be basically as complete as /
some of that, however, requires designing how to expose the LHS/RHS variants in the world where we only expose one today (RHS iirc)
some of that, however, requires designing how to expose the LHS/RHS variants in the world where we only expose one today (RHS iirc)
Pyeah well that's one of the long-term goals of our maths library and what i want in sys.num as well
TI also need to optimize , as while Vector2/3/4 are accelerated/fast.
The other types, like tend to not be
The other types, like tend to not be
Tand of course perf improvements to existing APIs require no API review
they just require a benchmark and me signing off
they just require a benchmark and me signing off
Pyeah, the problem here is that our maths library will never be able to match system.numerics
Pso having a good api surface becomes quite the sacrafice