Since posting The Great Agile Hardware Myth last week, I tried to think of some obvious myth that exists in the mainstream; some claim that we’ve all made that, without fail, turns out to be absolutely and entirely false. Took a while but I think I found it. We could have called it The RTL Done Myth, but I chose to call it the The 90% Done Myth.
|We’re 90% done. If everything goes well, all that’s left is testing and debugging this last 10%.
||The gant chart that says we’re 90% done is lying to us. We know it’s lying to us because it’s lied to us before, many times. We’d rather not use this gant chart but because tracking development progress is so important and we haven’t figured out a better way to do so, we feel like we have no other option than to keep trying to believe it.
What’s completely clear is that agile was created by software developers for other software developers that write software. Agile was not created for hardware developers that build hardware so there can be no such thing as agile hardware, right?
Nope. Wrong. I’ve heard a few comments to that effect over the years from people that are just mildly familiar with agile. I’ve heard similar comments from others honestly trying to wrap their heads around it while tripping over preconceived notions and losing themselves in hearsay. Truth is, what it means to be agile is not cut-and-dried, even within software nevermind hardware. I figured it’s finally time to take a step back to clear a few things up so that people new to the discussion can form a healthy definition for “agile hardware”. Continue reading
With the arrival of our board last week, I figured now would be a good time for a quick update on how our Agile2014 software/hardware co-development demo is coming along…
Our starting point was a reference design that is similar to what we’re trying to do in that it sends video frames to an HDMI port. We’ve relied pretty heavily on that reference design to build our hardware abstraction layer that connects our application to the device drivers. We threw away the application from the reference design – which is basically just a loop that sends coloured bars to the HDMI.
So far, Soheil and I have focused all our time on the software side. We’ve written our software application code and we’re almost finished the hardware abstraction layer. When the software is DONE, our Conway’s Game-of-Life application will send a grid of characters to the hardware abstraction layer; the hardware abstraction layer will convert the grid into pixels; and the pixels will be sent as frames to the hardware.
Because the reference design includes everything on the hardware side necessary to fetch frames and push them out to pins on an HDMI connector, we’ve been able to ignore the hardware for now. Our board was on back-order through april and may led us to do a lot of mocking. That let us model the interaction with the hardware without the hardware.
All our software code is C++ and we unit tested it using GoogleTest and GoogleMock. As I said, some of our code comes directly from the reference design (i.e. we imported and unit tested existing features) while other parts are newly written for the demo (i.e. TDD of new features).
That takes us to about 2am last Wednesday… which is when I got home for the university. It was cold and rainy.
With the postman delivering our Zedboard, we’ve changed focus to take a few unknowns off the table, the primary unknown being the quality and reliability of the reference design as a starting point. So from late last week to last night, Soheil has been installing tools on an old laptop I borrowed from work so we can build and run the reference design; it’s not the goal but we’re calling it a proof of concept that helps us learn about the board and validates our starting point.
The good news, as of about an hour ago, is that Soheil has the reference design building and running. That’s the picture with the board and the multi-coloured bars on the monitor. It doesn’t look like much, but we’re pumped to get over this hurdle!
Now we’re off to polish and deploy the real demo!
I’ve had people asking about the SVUnit new scripting proposal I posted a few months back. I forgot about it for a while but now I’m back. I’ve taken some of the feedback I’ve received and folded it into a new version. Slightly different in that I’ve added switches for compile and runtime options.
Here’s the help. Take a look and let me know what you think…
This is ready to go. I have a couple people trying it out. If you’d be willing to test-drive the new scripts before an official release, I’d sure appreciate it. Just let me know at firstname.lastname@example.org and I’ll ship you a copy. If all goes well, I’d expect to release this within a month or so as version 3.0.
Thanks in advance for you comments!