Видео

The Emotion Engine was fun to code on :) and thanks!

It's hardcoded in the FPGA's Verilog: initial begin $readmemh("rom.txt", fixed_memory); end The Makefile has some recipes for building specific test programs and generating the rom.txt from it. I could add a bootloader and read from UART, but meh... :)

Thanks!!

@@wirelessdirk i was lucky to even still have that amiga 1000 laying around to do that project.

This is an iceFUN board .. in North America you can get it from www.robotshop.com/ .. I have more info on my website.

xaos is a realtime Mandelbrot generator while this was 5000+ frames generated into .png files and turned into a video with ffmpeg. The reason I did this project is because I have interest in CPU instruction sets and I had been wanting to mess around with ARM64 vector instructions for years and finally found the time to do this. The Kubernetes part (other than clusters are just cool) was an excuse to mess with Kubernetes again incase I need it at my job or want to talk about it in a job interview. All the source code for this project is posted online, if you know of ways to improve the vector code, it would be cool to hear it. Please feel free to shoot me an email to discuss.

Short answer: Yeah, this is 32 bit floats. I kind of wanted to show the limit of the 32 bit, but looking back I probably shouldn't have done it with the flicker. It kind of looked interesting at some point though I think. Long answer: The coordinates for all frames to draw were picked with a Python script. The script starts with a r0, r1, i0, i1 that is known to have a good image. There were functions to zoom and shift (left, right, up, down) that took a delta value to know how much to add to r and i. So I would give it X iterations of a zoom and look at the picture at the point and make it shift up, down, etc for Y iterations to move to an interesting point.. then zoom again, shift again, etc. There's actually a point further in where the image started warping from the zoom but it destroyed the values of r and i so I didn't go in that far.

Thank you for the positive words :).

No, this is Patrick!

Hey, I recognise you...

I expected massive zooming in. What's in this video doesn't need a cluster to calculate I had no problem with the music. I assume the sound was intentional.

wydm? the music is fire ngl. leaving a like on this vid

Thanks!!! If you're interested in microcontrollers, MSP430 is one of the easiest assembly language to learn.

No chickens were eaten in the making of this video.

Thanks!! I think the chip used here to detect the tones (LM567) was actually used in very early modems (300 baud?), so something like this really could be done.

@@mikeakohn Maybe I can try it myself with arduinos. I don't have that level electronic knowledge. But i'll try file transferring.

Go for it.. If you need help feel free to ping me. I'll tell you though, I had a really hard time with the receiver side of the circuit... I myself am a software guy... the issue being making the microphone sensitive enough to pick up such a soft sound without adding noise to the input of the LM567. The LM567 also can be tuned for frequency and I believe also how fast it will respond to that frequency...

@@mikeakohn Thank you for your support! I will start a week later, I can't access my tools right now. Maybe I will share it to internet too.

@@mikeakohn And I have a recommendation for your project if you have it now. You can change signal length to frequency so you can make communicaton like 2x faster

Nah... no GPU. The reason for the project was to learn Kubernetes, so it was more of an exercise in that rather than to try to figure out how to generate the images very fast. Could be interesting though if one processes per Pi did have a GPU based generator.

@@mikeakohn Ah okay, fractal generation is definitely a handy CPU workload for this sort of thing. I've used it myself. Nice benefit that the demo looks pretty too!

That would be a pretty long video I think :(. I have a somewhat long explanation of the project on my website (there's a link in the description). If you do software and have a basic understanding of logic gates, you'd have no trouble picking up Verilog. In some ways I think it's a little easier than it looks. I can send you a link to a Verilog tutorial of you want?

Send a link, please.

This one was really nice: www.asic-world.com/verilog/veritut.html

@@mikeakohn Thank you very much!

Yep! This is my 2nd or 3rd CPU... still learning it.

@@mikeakohn oh very cool. I also was thinking to write a risc-v core myself but you know, we all have not much time. Btw did you write the core by just reading the opcodes?

@@randypeeters3931 Yeah, I wrote an assembler for it a while back. RISC-V was for the most part was pretty easy, the hardest part is some of bits in some of the opcodes are permutated. So not only are the bits broken up in different areas of the 32 bit opcode, they aren't in order.

​@@mikeakohn that is so cool! I wonder if you could get Linux running on it

@@whatamievendoing I only implemented the bare minimum instructions (no MMU, no FPU) .. around 87% of the FPGA is full. Also, the FPGA has some built in SRAM.. this design is only using 8k of it for RAM and some for other stuff. The chip only has 16k bytes of RAM I think. Can do other fun stuff with it though :).

It's an iceFUN board (sold by Robot Electronics in Europe and The Robot Shop in North America). This is a really great board for projects like this... the grid of 4x8 LEDs really helped with debugging.

That would actually be super easy to write firmware for. Hardest part is taking out the old circuit board.

I put "License: BSD" on the top of the source code. Feel free to grab it from my github repo.

This is actually not a VM, Java Grinder compiles Java byte code into MIPS assembly language..

Thanks! Do you mean move the 7 segment displays to the start circuit instead of at the end?

@@mikeakohn yes. Would be easier to see the winners time at the start line. I’m getting old my eyes stink.

With this particular circuit.. it's using an MSP430G2553 microcontroller and every single pin on the microcontroller is being used. It would take 4 extra pins to add a display like that.. and to have 2 displays, they would have be kept in sync. I have a new design in mind with a chip that has more pins, but I have other ideas for that one that might be better than adding a display on the front circuit...