1000 CPUs on a Chip

Often, CPUs that work together operate on SIMD (Single Instruction Multiple Data) or MISD (Multiple Instruction Single Data), part of Flynn’s taxonomy. For example, your video card probably has the ability to apply a single operation (an instruction) to lots of pixels simultaneously (multiple data). Researchers at the University of California–Davis recently constructed a single chip with 1,000 independently programmable processors onboard. The device is energy efficient and can compute up to 1.78 trillion instructions per second.

The KiloCore chip (not to be confused with the 2006 Rapport chip of the same name) has 621 million transistors and uses special techniques to be energy efficient, an important design feature when dealing with so many CPUs. Each processor operates at 1.78 GHz or less and can shut itself down when not needed. The team reports that even when computing 115 billion instructions per second, the device only consumes about 700 milliwatts.

Unlike some multicore designs that use a shared memory area to communicate between processors, the KiloCore allows processors to directly communicate. If you are just a diehard Arduino user, maybe you could scale up this design. Or, if you want to make use of the unused power in your video card under Linux, you can always try to bring KGPU up to date.

Hackaday 10th Anniversary: Jon McPhalen and the Propeller

[Jon] came out to our 10th anniversary mini-con to talk about the Propeller, and judging from his short introduction, his hacker cred is through the roof. He has a page on IMDb, and his first computer was a COSMAC. Around 1993, he heard of a small company introducing the BASIC Stamp, and like us with most new technology was incredulous this device could perform as advertised. He tried it, though, and for a few years after that, he was programming the BASIC Stamp every single day.

Having a lot of blinky light project under his belt, [Jon] was always struggling with interrupts, figuring out a way to blink an LED exactly when he wanted it to blink. A lot has changed over at Parallax since 1993, and now they’re spending time with the Propeller, an 8-core microcontroller where interrupts are a thing of the past. He showed off a huge, 10-foot tall bear from League of Legends, all controlled with a single Propeller, using 1000 LEDs to look like fire and flames.

[Jon] shared the architecture of the Propeller, and the inside of this tiny plastic-encapsulated piece of silicon is wild; it’s eight 32-bit microcontrollers, all sharing some ROM and RAM, controlled by something called a Cog that gives each micro access to the address, data, and IO pins.

When the Propeller was first released, there were a few questions of how the chip would be programmed. C isn’t great for multicore work, so Parallax came up with a language called Spin. It’s written for multicore microcontrollers, and from [Jon]’s little session in demo hell, it’s not that much harder to pick up than Python. Remember that hour or two where you learned the syntax of Python? Yeah, learning Spin isn’t a huge time investment.

Even though you can program the Propeller in C and C++, there’s a reason for Spin being the official language of the Propeller. It isn’t even that hard, and if you want to dip your toes in multicore microcontroller programming, the Propeller is the way to do it. It’s an open source chip as well so you can give it a try with an FPGA board.

16 core computer made of ATMegas

Your desktop has two, four, or even eight cores, but when’s the last time you’ve seen a multicore homebrew computer? [Jack] did just that, constructing the DUO Mega, a 16 core computer out of a handful of ATMega microcontrollers.

From [Jack]’s description, there are 15 ‘worker’ cores, each with their own 16MHz crystal and connection to an 8-bit data bus. When the machine is turned on, the  single ‘manager’ core – also an ATMega328 – polls all the workers and loads a program written in a custom bytecode onto each core. The cores themselves have access to a shared pool of RAM (32k), a bit of Flash, a VGA out port, and an Ethernet controller attached to the the master core.

Since [Jack]’s DUO Mega computer has multiple cores, it excels at multitasking. In the video below, you can see the computer moving between a calculator app, a weird Tetris-like game, and a notepad app. The 16 cores in the DUO Mega also makes difficult calculations a lot faster; he can generate Mandelbrot patterns faster than any 8-bit microcontroller can alone, and also generates prime numbers at a good click.

Continue reading “16 core computer made of ATMegas”