[jeffreyH]

I think our friend needs to use software because he is unable to comprehend the concepts enough to program the maths by hand.

[mxplxxx (OP)]

- My impression is that UML is intended to "simulate software". Software can be modelled as a discrete event system (and many obscure bugs happen when those discrete events are no longer atomic!)

UML StateCharts can simulate anything that is object-oriented (see https://www.embedded.com/design/uml/4219602/UML-Statecharts )

So it is for embedded control systems then. The example given was traffic lights.

They are used for many types of systems including simulations.

[Bored chemist]

I think our friend needs to use software because he is unable to comprehend the concepts enough to program the maths by hand.

That's' not a problem.
There are plenty of people who drive cars but couldn't hope to design the engine for one.

The problem is when they claim that the designer is lying, and there is no engine.

[Bored chemist]

- My impression is that UML is intended to "simulate software". Software can be modelled as a discrete event system (and many obscure bugs happen when those discrete events are no longer atomic!)

UML StateCharts can simulate anything that is object-oriented (see https://www.embedded.com/design/uml/4219602/UML-Statecharts )

So it is for embedded control systems then. The example given was traffic lights.

They are used for many types of systems including simulations.

That's nice to know.
But not a great shock.
Most computer programing systems(be they languages or s/w packages) are Turing complete.
You can pretty much do anything with any of them.

[jeffreyH]

They might even be used to measure the energy of a photon someday.

[evan_au]

most simulations I am aware of are based on UML StateCharts

When a man has only a hammer, all the world looks like a nail...
      ....and when he meets something which is definitely not a nail, it looks like an impossible task.
      ....like simulating the entire life of the universe, or the last few minutes before a supernova

Differential equations and matrix arithmetic were originally used to describe concepts in physical systems because computers were not available at the time

For an object-oriented view of the universe you could simulate every atom (or every star-sized mass), and it would have an "event" when that atom (or star-sized mass) bumps into another one
- But that requires a computer with as many memory words as the universe has atoms (or stars - still not practical)
- And ignores the fact that atoms interact via gravity continuously, not just when they bump into each other
- That is why continuous-time models (like differential equations) are still in use for simulating the motions of matter under gravity (like the universe) or in fluids (like the weather, an airplane or an exploding star)
- And even with the most powerful computers available, turbulence in these systems (and chaotic behavior in general) causes immense problems, since the behavior on the smallest scales impacts the behavior on the largest scales.

Because (in all conceivable probability) UML uses (matrices) anyway.

Actually, the core data structure of discrete event simulation software is a scheduling queue.

This queue is often hidden from the programmer, but every time a future event is triggered (perhaps after a random delay), a future event is posted on the queue. The simulation environment takes the closest approaching event off the queue to process events in sequence.

[mxplxxx (OP)]

But that requires a computer with as many memory words as the universe has atoms (or stars - still not practical)

Like I have mentioned before, this fact implies that the universe IS a simulation of something much bigger and much more complicated!

Simulation is a computer are rarely done on base data. A computer is real good at processing abstractions. The less abstract the abstraction the  closer the simulation is to reality.

The only way to simulate the universe is via abstractions. It just so happens this is possible because the universe is a abstraction hierarchy. I have previous described the universal system object which can be used to simulate universal system such as a solar system (or an Atom!). An abstraction connects to black boxes. It processes events from its parent and its children but needs no knowledge of how the parent or children work. To simulate the universe you only need to know the events its galactic clusters are sending the universe system object and what events to send the the galactic clusters in response to changes in the universe system object .Easy peesy lemon squeezy.

A solar system object will consist of a central sun object and peripheral planetary systems. The central sun object has state variables relating to the overall state of the object, a parent system that is the solar system object and child shell systems that correspond to the sun's core, photosphere etc.. The solar system object's parent will be a galaxy system. The galaxy system is more abstract than the solar system (or the solar system is more concrete than the galaxy system).

Each of these systems/objects  can be dealt with totally abstractly or can contain more concrete child objects all the way down to quarks or up to the level of the universe itself. The more this system of objects is abstract, the less accurate the final simulation and vice versa for concreteness. But the point of this is that simulations of part or all of the universe are possible with current computers. My 3dAbstractions software can deal with these sorts of simulation with ease. Hopeful it will be on open source soon.

This obviously is an extremely precied version of the technology required to run simulations on a computer and I would be very surprised if it gave you more than an inkling of the power of abstractions. 

BTW, quarks seem to be the equivalent of computer digits (0/1). They may be the only part of reality that can have state.

[mxplxxx (OP)]

- And ignores the fact that atoms interact via gravity continuously, not just when they bump into each other

I am investigating gravity as a revolving sequence of photons between the central object of a system and its peripheral (child) systems. i.e. how everything is PUSHED towards the centre of gravity of the system. A photon reaching the central object has nowhere to go but back where it came from. And therefore? ... well that is what I am considering at present. Can a photon change the position of an object without breaking the conservation of energy law?

Just recently? in physics, Newtonian gravity suddenly seems to be affecting the whole universe at the same time. Makes sense given the chaos that would like result form gravity of millions of years ago affecting the present. But seems to defy current physics re the graviton. But there is a lot of this sort of thing happening in physics today:) 

On the other hand it is very likely that the universe is a continuous entity and we exist in a simulation of part of it. That being the case, changes in the universe because of gravity can be reflected simultaneously over the total universe. BTW if the universe is a continuous entity then abstractions like a galaxy, solar system etc. probably don't actually exist. A quark exists with a single motion. All of the structures in the universe will be the result of quarks motions in relationship to other quarks. Is gravity involved in this? Possibly but not necessarily if Einstein's theory that matter warps spacetime is correct.

Theoretically, it may be that we can simulate the universe using only the motion of quarks and how aggregations of quarks warp spacetime.

[mxplxxx (OP)]

Can't give you examples of my code. But here is a picture of my personal (not for sale) gambling software as used today.

MOD EDIT: Link removed. Please do not post any sort of advertising.

[Bored chemist]

Can't give you examples of my code.

Why not?

[evan_au]

this fact (the complexity of the universe) implies that the universe IS a simulation of something much bigger and much more complicated!

No, they were just trying to simulate our universe, and it still takes a lot of time on the largest supercomputers.
They were not trying to simulate anything bigger.

The only way to simulate the universe is via abstractions.

I agree. Usual ways of abstracting in astronomy are:
- Model a small part of the universe (but still big enough to be "typical")
- Don't model individual atoms, but model galaxy-mass "blobs"
- Model the Dark Matter separately from the visible matter (with both interacting via gravity)
- Don't model individual stars, but use an "average" behavior of a galaxy

And the conclusion was that statistically, it is similar to the universe we see.

An abstraction where you don't model the individual cards in a deck, but just whole decks would fail in a game of chance!

I would be very surprised if it gave you more than an inkling of the power of abstractions.

I am aware of the usefulness of abstraction for software developers.

But when it comes to turbulent events (like a supernova), where tiny details are tightly coupled to the high-level outcome, abstraction fails. For these tightly-coupled problems, you need a numerical solution - with today's technology, the optimum answer seems to be a continuous-time simulation in FORTRAN, running on a large supercomputer.

Once the astronomical community understands what fraction of stars go supernova at certain stages vs those that turn into black holes (and publish it in a peer-reviewed journal), then other astronomers and cosmologists can use that encapsulated knowledge in predicting the life course of a whole galaxy. This is a level where there is loose coupling between stars and their galaxy and abstraction works.

The central sun object has state variables relating to the overall state of the object

And if you want to simulate the state of this sun object, I suggest that you read up on continuous-time numerical methods.
See, for example: https://en.wikipedia.org/wiki/Runge%E2%80%93Kutta_methods

[mxplxxx (OP)]

An abstraction where you don't model the individual cards in a deck, but just whole decks would fail in a game of chance!

So, so True. My gambling (racing) software is based on historical averages of odds/dividends over 4 meetings, which is approximately the same on any particular day. Unfortunately, in an individual betting pool (usually a set of 4 meetings) spikes often occur. There will be a run of favourites losing (or winning which is almost as bad) which breaks the bank. To date, no amount of smoothing can handle this situation.

[mxplxxx (OP)]

I agree. Usual ways of abstracting in astronomy are:
- Model a small part of the universe (but still big enough to be "typical")
- Don't model individual atoms, but model galaxy-mass "blobs"
- Model the Dark Matter separately from the visible matter (with both interacting via gravity)
- Don't model individual stars, but use an "average" behavior of a galaxy

And the conclusion was that statistically, it is similar to the universe we see.

Not surprised. Software that simulates the universe will be helped by the fact that all systems in the universe seem to derive (inherit in computer science) from the same underlying system object as I have explained in previous posts. Also the knowledge that all systems in the universe are scaled-down versions of their parent can be used in the simulation. Additionally, parent systems are abstractions of their children. A galaxy system is an abstraction of the solar systems that comprise it.

[mxplxxx (OP)]

But when it comes to turbulent events (like a supernova), where tiny details are tightly coupled to the high-level outcome, abstraction fails. For these tightly-coupled problems, you need a numerical solution - with today's technology, the optimum answer seems to be a continuous-time simulation in FORTRAN, running on a large supercomputer.

Related to Chaos theory? Actually I am not sure a supercomputer is required for heavy computational problems other than the fact that they are quick. You can simulate any system an event at a time as happens in computers without "core" processors. It is quite possible the universe is running an event at a time but doing it extremely quickly.

[Bored chemist]

Also the knowledge that all objects in the universe are scaled-down versions of their parent can be used

Of what is an electron the "scaled down" version?
(Or, if you prefer, what do you get when you scale up an electron?)
If there's no clear answer to that, then your central idea is wrong.

To date, no amount of smoothing can handle this situation.

And yet, the bookkeepers all drive expensive cars.

Actually I am not sure a supercomputer is required for heavy computational problems other than the fact that they are quick.

Well, duh!
Yes the advantage of fast  computers is that they are fast.
That's very important if you want to get an answer before the grant runs out..
So, in practical terms, yes, they are required.

[evan_au]

Actually I am not sure a supercomputer is required for heavy computational problems other than the fact that they are quick.

For tasks like weather forecasting, there is actually a very specific speed requirement: To be useful, you have to produce a 1 day weather forecast in less than 24 hours.
- To be really useful, you need to produce around ten 7-day forecasts in under 24 hours, with different starting points.  This then allows you to provide a probabilistic weather forecast, eg "60% chance of rain"
- This requires the latest, most powerful supercomputers.

Another task that requires supercomputer power from perhaps 5 years ago is for autonomous vehicles:
- These have a strict real-time requirement - you must process images, extract the details, correlate different sensors and react - before you hit a pedestrian, a car or a pole.
- It has additional constraints on physical size, power consumption, operation in temperature extremes and limited cooling.

[Bored chemist]

Actually I am not sure a supercomputer is required for heavy computational problems other than the fact that they are quick.

Come to think of it, isn't that a bit like saying "I don't think food is required for eating- except that it needs to be edible"?

[mxplxxx (OP)]

Actually I am not sure a supercomputer is required for heavy computational problems other than the fact that they are quick.

For tasks like weather forecasting, there is actually a very specific speed requirement: To be useful, you have to produce a 1 day weather forecast in less than 24 hours.
- To be really useful, you need to produce around ten 7-day forecasts in under 24 hours, with different starting points.  This then allows you to provide a probabilistic weather forecast, eg "60% chance of rain"
- This requires the latest, most powerful supercomputers.

Another task that requires supercomputer power from perhaps 5 years ago is for autonomous vehicles:
- These have a strict real-time requirement - you must process images, extract the details, correlate different sensors and react - before you hit a pedestrian, a car or a pole.
- It has additional constraints on physical size, power consumption, operation in temperature extremes and limited cooling.

Yep!! Real-time system processing requires a computer quick enough to process whatever load is put on the system in question when running under time constraints.

[mxplxxx (OP)]

Yep!! Real-time system processing requires a computer quick enough to process whatever load is put on the system in question when running under time constraints.

But I am not sure a simulation of a super nova is a real-time system. In this case the extra processing power of a super computer may just be to speed up a computationally intensive process.

The ideal for real-time processing would be a processor per object, each object processes one event at a time, and objects to be totally independent of each other. This seems to be what reality does (the quantum nature of reality ensures that one event at a time is processed).

Real-time systems need to handle the situation when not enough time is available. Our brain is  a real-time system that has to do this. Putting you hand on a hot item on the stove gives you an immediate withdrawal reaction from your hand because your brain is not quick enough to process the event.

[Bored chemist]

In this case the extra processing power of a super computer may just be to speed up a computationally intensive process.

Come to think of it, isn't that a bit like saying "I don't think food is required for eating- except that it needs to be edible"?