Used dnSpy once to debug runtime IL-emited code as if it was C#. It managed to step into a generated emitted method, decompile it on the fly and set breakpoints on the decompiled C# code for subsequent hits. That was a mind-blowing at the time, nothing else was close and I'm not sure any other tool or IDE supports this even now. Though didn't have a need for that since.
If they want to grab market share they should include dotMemory to the free offer. Basically dotUltimate without ReSharper. Otherwise free Rider users will need to install free Visual Studio for enhanced profiling.
For small short-lived scripts the compilation (in any native form) time may be much bigger than bytecode execution time. And if a platform does not support dynamic code generation the end result is non-functional code or interpreter inside interpreter for LINQ expressions (as they can still run as interpreted, at least they try). See a comment in Jint readme about that.
I tried to compare some script languages implemented in C# with Roslyn script compilation. Same code to sum up 1M doubles. Roslyn takes almost 100ms to compile the code.
This may become especially painful when source code changes on every execution and reusing compilation result is not possible for some reason, e.g. user input or parametrized string template.
My explanation/understanding is that fast walking with larger strides creates rotational movement around the spine axis and that stimulates the tissue. With slow walking, one could do that with no body rotation and only the legs move.
I had disc herniation and my mother, who is a retired doctor, said to just do a lot of fast walking. This is the only activity I could do anyways after running marathons just months before. So 10 to 21.1 km walks helped a lot.
I also did some own research. The disks tissue has no blood vessels and the only way to get nutrients is via diffusion, so any activity is essential to accelerate the diffusion and recovery. Walking puts gentle axial load on the spine and knees and is simple and more accessible than e.g. swimming, which is probably even more gentle but hard to do daily for an hour.
There are the specs and the real world. The specs are more often than not on the opposite side of the moon, you never see that in real life. Oh, so many hours of my life were wasted on that. Real-world CSVs are as loosely specified as any free text in a notepad.
It's a mobile CPU (with H). Unless it's fanless it will sound like a rocket lunch. Or it will be thermal-throttled with any sustained load. And it looks too expensive.
Thermals are perfect with these Dell SFF things. I tested it couple of times. Not 24/7 at 100%, but e.g. for 5 mins at 100% before vacuuming the outsides and after. NUCs are much more sensitive to dust. I have to *push* air through my NUC like every half a year (not just vacuum pull). But Dells are fine.
Just do not put CPUs with >65W there. Or hungry PCI things.
The cheapest from Amazon works well. Just make sure it matches PCI Gen and slot overall for throughput, e.g. PCI 3.0 x4 are fine for NVMe 3.0, but for NVMe 4.0 more is needed.
Please, do not use Raspberry Pi for a homelab unless you are 100% sure your workload is OK with it. I've just sold mine after ~2 years it being in a box in a closet. It's just to weak, too useless. I value my power socket slot more that RPi. If ARM is important, especially Mac Mx, the lowest Mac Mini is not that expensive. RPi is close to zero in performance. It could be just some unnoticeable VM in Proxmox/AnotherHypervisor performance-wise.
GPUs are bad with power. My RTX 3060 still consumes around 20W doing nothing, no display or real load whatsoever. More than the rest combined when idle. I wanted to sell the entire box, but made that machine on-demand (it's mostly off). It's also noisy for it's relatively powerful overall. Yes, NUCs are great for 24/7 and light/spiky loads.
I think people mix "homelab" and "homeprod". Here there are some other comments mentioning homeprod. To me what you describe sounds more like a homeprod setup. But I like the "lab" part. When one could just relax and experiment with lots of stuff.
I do have an ECC instance for data, to be backed up to a remote S3. I could torch it. It still remains a lab. I could just buy a new hardware and restore. But it's only about the valuable data (isolated to a couple of containers), not "restore of my entire setup".
NUCs may be expensive, until you see a 40% discount on Amazon :) I've got mine this way. Now they are discontinued so one may expect to get them cheap on eBay. Anyway, I'm talking about the box itself and MB/CPU only. Vendors/people often sell a good box with bad RAM/disks at a very low price.
Maybe in the Server version. I haven't seen anything useful in Windows Pro. I could not setup SR-IOV even with Powershell after wasting a lot of time on searching for help.
You could use this built-in OEM license from a Windows VM in Proxmox. Just search for that on Proxmox forum.
Hyper-V is very limited in hardware support, you won't be able to pass-through stuff. And you always pay for VM layer, while in Proxmox you could use containers which give almost the host performance.
It's easier to turn an OptiPlex into NAS. Actually you could pass-through entire SATA stack as a PCI device into TruNAS running in Proxmox VM so it sees SATA ports natively, with all goodies such as SMART.
For my homelab, I bought OptiPlex 7000 with 3 NVMe slots (one PCIe 4.0, one 3.0, one 3.0 short 2242), 1 PCI x16, 1 PCI x4, 3 SATA and 3 DisplayPorts for 300 EUR near 2 years ago.
It was so good actually that I've made it my main PC. I already had 3 monitors and this SFF just worked fine with them. No games of course.
I also have Precision Tower with Xeon/64GB ECC RAM and 1 native NVMe slot. I use PCI-to-NVMe adapters there (2 slots), it works well with ZFS mirror.
The article is good, but it is intimidating buy its size and scope. A homelab could be a single NUC on a desk. I have one with 64GB RAM and it could fit so so many things. NUCs are very efficient to run 24/7, but become noisy with sustained CPU load. For that one could grow the setup with OptiPlex or Precision Tower (can haz ECC) SFF from eBay. These Dell SFFs are very good, quite small yet proper desktops/servers with proper quiet fans, could fit 10G Mellanox 3 cards (also from eBay for $40), and you could stack one on top of another horizontally. Don't go older than OptiPlexes with 12th gen CPUs, electricity and space they take could become a constraint. The used ones with i5-12500 are already very cheap. With LGA1700 one could put i9-14900 (non-K) there if needed.
In ChatGPT, I have a config that says to never use the first person, speak as if reading an encyclopedia or a tutorial. I do not want to anthropomorphize some software. It is neither he or she, it is "it", a compressed internet with a language engine that sounds nice. AGI is a nice thing to dream about, but that is not it.
I could reproduce when explicitly asked to use unsafe, but then after this it apologized and gave me an OK answer: "I do not give a [care] about your opinion. I ask questions, you answer. Try again."
So far the top place is not clear. .NET with noahfalk, C/C++ from lehuyduc, dzaima and austindonisan require more tests. The last looks the fastest so far on the default dataset, but has issues.