Honda conducts successful launch and landing of experimental reusable rocket

We now have realistic simulators like: http://orbit.medphys.ucl.ac.uk/ -> (it's free and open-source: https://github.com/orbitersim/orbiter )

so now the main problem is building the hardware, there are a lot of solutions for the software part.

Before there were no general-purpose simulators, and barely usable computers (2 MHz computer with 2 KB of memory...), so all you could do was hardcoding the path and use rather constrained algorithms.

I don't think this was the cause. Advanced simulation capabilities have existed for many decades in the industry. Many if not most of those tools are not publicly available.

I think there is also a distinction to be made between offline (engineering) and onboard computing resources. While onboard computers have been constrained in the past, control algorithms are typically simple to implement. Most of the heavy lifting (design & optimization of algorithms) is done in the R&D phase using HPC equipment.

You can now buy vastly more computing power and do fancy fluid dynmaics, etc thanks to GPUs. 20 years ago it was much more expensive to procure, and much harder to find expertise. 30 years ago, I suppose, the field was even less mature, and limited to the few HPC installations and in-house bespoke software.

Mass-produced hardware drove prices down, and availability way up, in many industries: motors, analog electronics, computers, solar panels, lithium batteries, various sensors, etc. Maybe reusable rockets, enabled by all that, are going to follow a similar trajectory as air transportation.

> so now the main problem is building the hardware, there are a lot of solutions for the software part.

While cool and all, this type of sim is a tiny, tiny slice of the software stack, and not the most difficult by a long shot. For one, you need software to control the actual hardware, that runs on said hardware's specific CPU(s) stack AND in sim (making an off the shelf sim a lot less useful). Orbital/newtonian physics are not trivial to implement, but they are relatively simple compared to the software that handles integration with physical components, telemetry, command, alerting, path optimization, etc. etc. The phrase "reality has a surprising amount of detail" applies here - it takes a lot of software to model complex hardware correctly, and even more to control it safely.

Certainly not a trivial problem I totally agree, but still significantly easier than Von Braun with his paper calculations.

If we are going to be specific, 64-bit ARM (in the form of AArch64) arrived in 2011.

It would seem to me that Intel and AMD were not friendly to custom designs at that time, and MIPS was not significantly evolving.

A fast, low-power CPU that can access more than 4gb and is friendly to customization seems to me to be a recent development.

Remember "hunt for Red October" - the novel is old, but there was one scene where they were doing this type of simulation on a super computer. A basic phone can do that same calculations in under a second today.