Random Chitchat 2012-2016

[ThuGie]

Couldn't you just upgrade your cpu instead ? if there was a usb like that it would probably be quite expensive, but funny to see.

[xoft]

Why expensive? The RasPi Zero retails for around 150 CZK, I don't think the CPU-in-USB-stick would have to be much more costly. Someone even already did a RaspiZero USB stick. The only problem is in the software - providing a real compiler and a management tool for multiple such sticks.

Upgrading my CPU would mean upgrading the entire computer, since I'd have to upgrade the mobo and move from DDR3 to DDR4. Hence the cost would be quite large (50 000 CZK for a CPU, 10 000 CZK for a mobo, 8 000 CZK for the RAMs). And still I'd get a CPU that has no more than 10 cores, so 10 parallel compilations anyway. The USB sticks would be a viable option, even if costing 1 000 CZK for one, I could get 68 "cores" for the same price, or less cores for a (much) smaller price.

Just a random thought

[ShadowCone]

You mean stuff like PKCS#11 in cryptography?

That standard actually describes that you can pass data to a smartcard and you get decrypted data back. Keeping the key in-house.

What you mean is a compiling stick similar to the yubikey with some predefined compilers, where you can pass code and get a binary, right?

Optionally available for multiple architectures supporting native cross-platform builds


"oh you have the newest clang stick! Is it nice?" - "yes, much faster than the old gcc one"

[xoft]

Something like that. But the stick could be upgradeable, the compilers wouldn't have to be predefined. If it runs Linux (like the RasPi Zero), it could just upgrade the compiler when requested.

[ShadowCone]

Well, you would want to start a business, right?

I've digged a bit into this area. Running a general purpose processor and a linux operating system can get you the same speed as a Raspberry zero at 10$ cost. It's low cost, but your old cpu would do it faster than 10 sticks. Apart from that, you got 70 usb ports? Usb also has an exact limit somewhere at 140 where no more devices can be attached cause the host controller can't control them all. But that's a different story.

You would want to either go for ASIC's or FPGA's or similar stuff. Ultimately, ASIC's would be the best way, but that means the highest effort and the highest risk cause of relatively high cost unless you get really big sales numbers.

OK, FPGA's then. They are basically ASIC's but programmable, which makes it easier for semiconductor companies to produce and sell them (they wouldn't need to produce a line for one customer thus dumping the price). You can get relatively potent FPGA's for like 10$ per chip or dev kits for like 20-30$ as a USB pen with some ports on it.

Telling you that this 10$ chip runs with 500MHz and 128kbit of total memory might underwhelm you, considering the 1GHz processor of the pi zero and the 500mb memory, but these FPGA's can be highly optimized and put in some really small casings.

The main problem is: How the fuck am I going to run a compiler written in c++ on native hardware. Basically dropping the OS or reducing it to a minimum.

After all dropping the OS and rewriting the pi zero's firmware to run a compiler directly and solely might actually be the best go, as it involves the least amount of work. Gcc's pretty sophisticated after all. After all I don't like the idea of using a pi 0 because it's relatively big in size. It contains a lot more than just the needed stuff and makes so much overhead.

Last but not least I want to name the idea that just came to my mind. An FPGA alone won't solve the problem. Targeting around 50$ would give you enough money to think about a "quad core" array of 4 FPGA chips (two on the up and two on the down of the stick) as well as 2 ram chips (again one up and one down). But I have no idea if it's ever possible to adjust gcc or rewrite a valid c++ compiler for that thing. I mean, if you even could use standard gcc code as it's released, it wouldn't help you as you would need to optimize it in order to get it faster. After all, FPGA's aren't just make-it-faster devices. You just drop the OS and build the stuff that's crucially needed yourself.

For God's sake, please leave the cmake on your computer.

[xoft]

After a long time, I decided to update to the latest Minecraft and actually play the game for a while. Suddenly there was a skeleton horse in my Overworld, and I was like "what? are those supposed to spawn?" I went to have a closer look and was taken by surprise when that horse transformed into three skeleton "riders of doom". Wtf did they implement???

[ShadowCone]

Loool didn't play standard minecraft for ages..... @xoft you read the post above? The main work would be programming a compiler for an fpga

[PureTryOut]

Lol have you been living under a rock?

They spawn when there is lightning. Sometimes if you get close, they get struck by lightning and transform into those "riders of doom". This does not happen always however
It would be great to first get regular horses and horse riding working properly before you start implementing skeleton horses lol.

[xoft]

Yeah, it's kinda shame that one of the main Cuberite devs hasn't actually started Minecraft for almost a year

I'm now reading up on the woodland mansion, the generator behind it will be a nice treat to solve

[PureTryOut]

If only my C++ was better, I would try to implement all the missing features

The End has all kind of new stuff that hasn't been implemented yet as well.