AXI 1-Wire Driver Is AMD's Latest Upstreaming Effort To The Linux Kernel

[phoronix]

Phoronix: AXI 1-Wire Driver Is AMD's Latest Upstreaming Effort To The Linux Kernel

Since AMD's acquisition of Xilinx and working to broaden the portfolio of offerings for the data center, more AMD-Xilinx drivers have been working their way toward the mainline Linux kernel. There's been upstreaming efforts such as the Versal EDAC driver, generating DeviceTree nodes for PCI devices, Versal watchdog driver, QDMA driver, CDX bus support, and more. The latest driver working its way toward the mainline kernel from AMD is the AXI 1-wire driver...

https://www.phoronix.com/news/AMD-AXI-1-Wire-Linux-Driver

[timofonic]

Isn't an FPGA for sensors a bit overkill?

What about AMD making Xilinx (now AMD) FPGA toolchains open source?

I hope AMD makes someday an economic and better competitor to DE-10 NANO and somebody make a Mister-like project of it.

[Spacefish]

Isn't an FPGA for sensors a bit overkill?

Yes probably, but depends on the application i guess.. If you already have an FPGA in your design, and have a lot of 1-Wire sensors, why not use it for reading these?

Some applictions don´t even need an MCU and only use an FPGA + "PROM" (typically Flash), but you still need some minimal memory to store a serial number, some calibration values or something like that.

[osw89]

Isn't an FPGA for sensors a bit overkill?

This isn't added to the kernel so that you can use an FPGA just to read sensors. They sell many ARM + FPGA chips, all of them capable of running linux. You can instantiate the processing system(CPUs) in your block design, instantiate whatever AXI peripheral IP comes with Vivado, connect them to the CPU via AXI, and boom you have an ARM core running Linux with whatever peripherals you added. The purpose is that you can use their IPs in your projects even if you want to run Linux instead of bare metal C code.

What about AMD making Xilinx (now AMD) FPGA toolchains open source?

That would be amazing but highly unlikely, as painful to use as vivado can be, it still produces decent results. There were some people trying to add support for some xilinx parts to yosys and nextpnr but support it doesn't support the cheaper zynq 7000 devices unless something changed. If you want an FPGA with an opensource toolchain you should get an ice40/ECP50 instead. I only tried the ice40 personally, it's quite small and slow but yosys+nextpnr works very well with it.

I hope AMD makes someday an economic and better competitor to DE-10 NANO and somebody make a Mister-like project of it.

FPGA vendors don't create those boards, the DE-10 is made by terasic for instance. You can get Zynq 7020 boards like the Pynq-Z2 for around $160 (no idea if it got more expensive since I bought it) and there are other ones in the $200 to $250 range. The DE10 costs $225 without an academic discount so it's pretty much in the same price range as the Zynq boards.

[pong]

Relative to what you seem to want at a high level (board that does X project) I'm sure you're right.
Relative to FPGA devkits themselves & their prices / merits, well, it's usually a sad story in some market segments and a happy story for some use cases / market segments.
Usually FPGA vendors "push" what technology they consider is new / capable of rising in market share by more marketing & user test / adoption starts.

So they make their newest things they want to establish / boost their market share of into a greater number of new devkits / boards, and they can price those agressively while they're relatively current / new. The down side is they're quickly discontinued or the price might rise a lot when the vendors aren't promoting those chips so much after a couple of years or whatever.

So it has been a long time since one has seen Spartan 6 come out so not earth shaking new boards for it but 3rd parties make more and more options for a "mature / legacy" product that is long available.

When artix was newer there was a flurry of good capable and low cost devkits for that.

Where spartan 6 / 7 / artix had no solution for some capabilities even some of the medium / low size kintex chips were put on devkits for really good overall prices / capabilities, things one would not expect to see if they had available bigger size / function parts in spartan 7 / artix or whatever.

Now they're pushing their AI/ML accelerator / SBC / SOC stuff to markets like module based robotics, computer vision, IP camera, et. al. designs that would not normally have the engineering capability / time / money / skills to create such designs from bare ultrascale BGA parts and a low level IP core library so they're pushing all the vitis stuff, modular SOMs accelerators / solutions, etc. along with HLS and more RAD / drag & drop / ML library oriented offerings so they can sell to markets who instead of a FPGA SOC would choose to buy some ARM SOC iMX8 or whatever from their competitors to solve some IP-camera / robotics / machine control problem etc.

So among other places that's whre the current "best value" development kits are from them & their board / kit vendors.

Probably overkill / too expensive for your needs if DE10-nano capability is all you want / need (DE10-nano IS a nice board at a price that was quite attractive for when it came out in the past years), but for xilinx the offerings now are pretty amazing / cool considering that there are unprecedented capabilities / functions / tools available and HW that is generationally more integrated / powerful & easy to use for higher level real world applications.

I hope AMD makes someday an economic and better competitor to DE-10 NANO and somebody make a Mister-like project of it.

[dimko]

does this allow for hardware sync I wonder?(gnome could benefit from this)