Digital design is still too locked behind licenses and closed tools.
Modern open-source flows are worth another look.
We wrote about the hardware design stack we use:
https://t.co/EIJlWzJ6yT
#DigitalDesign#OpenSource#ASIC#FPGA
You're probably overcomplicating your testbench.
Model the behavior, enforce standard interfaces, and use lightweight verification libraries. That's it. And it scales.
Check our latest blogpost:
https://t.co/TQNuQy9JAH
#FPGA#DigitalDesign#ASIC#Verification
CRC, Reed-Solomon encoding, PRBS all use the same serial feedback loop underneath.
The trick to running k inputs per cycle isn't pipelining, it's matrix exponentiation.
Check our latest blogpost:
https://t.co/EcjpE1KHI2
#FPGA#DigitalDesign#ASIC
I’m glad to share that we started @hardmatrixdev
I get to do it alongside people who are not only friends, but also great professionals and engineers I deeply admire
More coming soon
We started HardMatrix to build serious digital hardware: IP cores, ASIC/FPGA systems, and to help create a more open and capable semiconductor ecosystem.
Small team. Deep technical focus. Long-term mission.
@JustinLDallas you do both. Your testbenches should always check de results so you build a regression suit. When the tests don't pass, you open the waveform and analyze the signals to debug the problem
@redixhumayun@kai_fall@RazorSharpFang Is not obvious at all, in hardware is really common to use msb first reprentation so a concatenated with b could easily be interpreted as [a, b] where the bits of a are in the high part
@splinedrive I'm glad you liked it! Sutherland has a couple of nice papers, I would also recommend the one about X propagation and the one about assertions embedded in the design.
@OlofKindgren I've been using fusesoc generators to integrate IP cores from Xilinx to my desings and to run tests with svunit, its a really useful feature
@mikeev@OlofKindgren Haven't tried it yet, but I really liked the concept of generators in Chisel, where you have this clear boundary between a more powerful abstract part that is executed in compilation time and a more constrained part that represents actual hardware
@mikeev@OlofKindgren Also the language have some limitations, functions are more ergonomic than modules for small hw, but cant use them since arguments need a fixed width, you can wrap them as static methods of parameterized abstract classes but that has poor tool support and its rather hacky.
Let's start with an introduction. World: this is the new cocotb Twitter account. @cocotbnews: this is the Twitter world.
Hello everyone! Just to let you know: I now also have a brand new logo and a website at https://t.co/kodSswjxEv. Thanks to Lena Alm for the logo design!
Here's @ylecun in 1993(!) demoing LeNet-1.
This was the world's first convolutional neural network for handwriting recognition. Today—29 years later—this same approach powers the majority of computer vision.
Such a historical moment, and breakthrough for machine learning.