Xilinx

with the CPU instructions and put it onto the FPGA platform.

So this one is actually very be helped by the Xilinx platform because within the Xilinx SOC, you have the CPU there and also you have the program logic.

So we put the program logic as the DPU part and

also...

We get the money in Silicon Valley, actually, in 2016, February.

And then we start around and around.

And last year, July, we were acquired by Xilinx.

So now, actually, as I have already said, Xilinx is a great company, and he is building up all the FPGAs, the hardware, but he needs reference design and the IPs inside.

I remember a guy, his name was Eric King.

He used to always yell out, Xilinx crunching at the top of his lungs, and he'd just magically jump onto Xilinx and start trading Xilinx.

Well, I didn't know what the hell Xilinx did.

I didn't know, but I would trade it because somebody got me all excited about it.

And then I started to really analyze what I was doing on a daily basis.

So later on, you'll see a lot of effort on the circuit design in the second half of the talk.

So what percentage of a high-end Xilinx FPGA would your design take up? Do you have to minimize the amount of hardware?

I didn't try too hard to reduce the area, but the end result is it's like 7% of a 40 nanometer FPGA.

It's like 1% of one now.

At the end, This is the scheme representing the resource usage on one of the superlogic regions of our FPGA chip.

We are using this Xilinx Alveo U 250 FPGA chip, which is made of four of such SLR regions, having just weak wiring between these SLR units.

So it wouldn't be a good strategy to go for high data throughput between these regions.

Instead, we implemented four independent quantum computer simulators on each of these regions, because we can split the calculation of gradients between these regions, so it makes sense.

How do you remain a confident, effective leader when the world around you seems to be changing faster than you can keep up? My guest today is Kathy Debbie McFadden, an executive coach who has spent more than 20 years in helping leaders navigate organizational challenges and develop their leadership capabilities.

Debbie's held leadership and coaching roles at Cisco and led global people services at Xilinx.

She brings experience in human resources, change management, and leadership development.

She's also a professional certified coach and has extensive training in mindfulness and

Compiler tools for VHDL or Verilog for this, they're free.

Now I'm not saying they're open source, but the low level basic Alcatree tools are based on, it's a Xilinx RTX 7.0. You can get a free license of Vivaldo from Xilinx because they make the chip.

They don't give you all of the tools, but you have enough.

I did everything with Claude.

But just to sort of, you know, relate it to, you know, uh, familiar, more familiar territory.

Back in nineteen eighty-four, when Xilinx was making gates with about nine thousand gates on them, FPGAs with nine thousand gates on them, you know, people were going in and twiddling these bits in the logic blocks and twitching, twiddling the bits in the switch, switch blocks.

You know, you basically had to be a chip designer to use one of these, these, uh, you know, FPGA chips.

Now- You can get a chip with 4 million gates on it, and nobody goes and even knows that there's these things like lookup tables in there [laughs] and switch matrix.

It goes off and executes that.

Would you run in a programmable fabric, like with an Altera or a, I would say Xilinx AMD?

Uh, we could sit beside that.

I think we have some use cases in robotics, uh, where near real-time sensor acquisition is still happening with the FPGA fabric, and then we take over post-processing, pre-processing, all of inference, specifically in robotics, uh, doing, you know, the heavy, uh, VLA processing, if you will, uh, that is very tightly built with the FPGA fabric.