Author Topic: Next step from AVR  (Read 2031 times)

0 Members and 1 Guest are viewing this topic.

Online rstofer

  • Super Contributor
  • ***
  • Posts: 4930
  • Country: us
Re: Next step from AVR
« Reply #50 on: October 19, 2018, 02:54:29 am »
Right now there is no useful RISC-V hardware, there is nothing to talk about. And suggesting it to newbies moving from AVR is plain cruel.

Why cruel? He's a hobbyist. He doesn't have a boss who yells at him every morning pressing for results (whatever this is in his mind). Why do you think that repeating LED blinking/LCD driving exercises with STM (after he already did exactly that with Arduino) will fill him with joy? Looking at the first ever open-source RISC-V processor, which does not yet have a fully operational ecosystem, may be something differed and much more rewarding. I think Bruce is absolutely right suggesting this. OP is free to choose what he likes.

OK, I somehow got interested in the FE310 board.  It starts out with a 48 pin chip so, by definition, IO is limited when compared to a 100 pin chip.  But it does have a UART, 3 PWMs and SPI (master only) with 3 slave selects.  Getting to the details is going to take a bit more reading.  It's also fast!

It uses an external Quad SPI flash for program storage (there is a 16 kB internal instruction cache) and it has 16 kB of data SRAM.  AFAICT, there is no analog input or DAC output (other than PWM).  There doesn't seem to be support for Ethernet (like in the LPC1768, among others).  No USB Host or Device AFAICT.

So, pretty spartan by some measures.  But, as pointed out, this is just one implementation.  Something to get started with.  But, at $59, it's kind of expensive when I can buy an Arduino Nano for about $5.  And the Arduino Nano is pretty well understood.  I'm not comparing speed here, the Nano is SLOW!  The RISC-V is FAST!

Intertia, that's the problem.  Static inertia says "Nobody is using this, prototype projects aren't available, why bother?" and dynamic inertia says "I'm using ARM, it's a RISC core, why bother?".

The ISA is irrelevant to the user in most cases.  They want to pump C in one end and having blinking and flashing come out the other end.  At the hobby level, I'd be willing to bet that most users don't know, nor do they care, how the NVIC works.

Think of how different the PIC, AVR and ARM ISAs are.  They are wildly different and, yet, the PIC 16F series is still being sold.  By the bazillions!


 

Offline Siwastaja

  • Super Contributor
  • ***
  • Posts: 1374
  • Country: fi
Re: Next step from AVR
« Reply #51 on: October 19, 2018, 03:52:25 am »
IMHO, if brucehoult's company wants to succeed on the microcontroller or embedded market, the ways to get there is:
1) Try to forget your great RISC-V core for a second
2) Try to come up with flexible and numerous peripherals, and
3) Document them clearly.

The managers who want the ARM core on Powerpoint are hard to convince anyway. They need the inertia the most.

But, there are engineer-like managers as well, and sometimes even engineers can make the component decisions. This is where good peripheral sets totally shine, since most engineers are not that interested about the core, as long as it has the basic tool support. This is a free advice for brucehoult, but IMHO they should try to concentrate on the peripherals from day one - and here I can see a huge opportunity for the configurability: Every core licensing company has the same graphical toolset you can use to "design your CPU", meaning, choose whether you have a hardware divider or not, or select the amount of cache, and so on...

No one has done the same for peripherals on a level that would allow building a modern microcontroller. This could be huge; look at high-end microcontrollers that always implement easily over 100 peripherals, of over 30-40 different types, just that the user can use 5 of them and leave the rest unused. Maybe there would be market for custom SoCs in this $5-$15 mid-to-high-end microcontroller market?
« Last Edit: October 19, 2018, 04:24:26 pm by Siwastaja »
 

Offline mark03

  • Frequent Contributor
  • **
  • Posts: 376
  • Country: us
Re: Next step from AVR
« Reply #52 on: October 19, 2018, 04:01:12 am »
If you want to start playing with an IDE for STM32 (or any other ARM Cortex-M line) that is truly plug-and-play, it's worth mentioning that the free (code-size-limited) version of IAR is pretty capable.  You're unlikely to hit the code-size limit on a hobby project, as long as you don't use the vendor-supplied peripheral drivers (HAL), which as a hobbyist, you shouldn't.

Obvious drawbacks:  Windows only, and a regular license costs serious money.  As a student you would likely be better served by learning how to roll your own gcc-based toolchain.  OTOH, IAR is very widely used in for-profit embedded development, so you might look at it as a job skill, although any engineer worth their salt should be able to pick up a new IDE quickly.

As for STM32 vs PSOC vs others, really, just pick one... they have far more in common than their differences.
 

Offline brucehoult

  • Frequent Contributor
  • **
  • Posts: 759
  • Country: nz
  • Currently at SiFive, previously Samsung R&D
Re: Next step from AVR
« Reply #53 on: October 19, 2018, 10:06:57 am »
No one has done the same for peripherals on a level that would allow building a modern microcontroller. This could be huge; look at high-end microcontrollers that always implement easily over 100 peripherals, of over 30-40 different types, just that the user can use 5 of them and leave the rest unused. Maybe there would be market for custom SoCs in this $5-$15 mid-to-high-end microcontroller market?

Funny you should say that. Already done and in testing internally, switched on for the public soon: https://www.sifive.com/chip-designer

There are already hundreds of IP blocks licensed from various DesignShare partners under a deal where you pay no license fee until you go into volume production.

Both the FE310 and FU540 have been recreated using this flow and taped out to TSMC.

Yunsup talked about this at the Embedded Linux Conference
 

Online legacy

  • Super Contributor
  • ***
  • Posts: 3288
  • Country: 00
Re: Next step from AVR
« Reply #54 on: October 19, 2018, 11:18:10 am »
But, at $59, it's kind of expensive when I can buy an Arduino Nano for about $5

Yesterday I went to the restaurant for a business meeting, and when I went out, a few colleagues were complaining they were still hungry because the dishes haven't been filled to the brim and for 60 euro per head it was rather expensive.

The cook was nearby and heard, so he answered something like "well if you compare the street food with a tasting of excellence, you can do similar complaints:D
the Bunker is open!
 

Offline brucehoult

  • Frequent Contributor
  • **
  • Posts: 759
  • Country: nz
  • Currently at SiFive, previously Samsung R&D
Re: Next step from AVR
« Reply #55 on: October 19, 2018, 12:31:21 pm »
So, pretty spartan by some measures.  But, as pointed out, this is just one implementation.  Something to get started with.  But, at $59, it's kind of expensive when I can buy an Arduino Nano for about $5.

A fair assessment (not only the part I quoted).

You can by the way buy bare FE310 chips for $25 for 5, and there is also the "LoFive" design done by a 3rd party. The board design is open source, you can make it yourself, and there have been a couple of GroupGets campaigns for assembled LoFives for $25.

https://groupgets.com/manufacturers/qwerty-embedded-design/products/lofive-risc-v
 

Online rstofer

  • Super Contributor
  • ***
  • Posts: 4930
  • Country: us
Re: Next step from AVR
« Reply #56 on: October 19, 2018, 12:54:58 pm »
So, pretty spartan by some measures.  But, as pointed out, this is just one implementation.  Something to get started with.  But, at $59, it's kind of expensive when I can buy an Arduino Nano for about $5.

A fair assessment (not only the part I quoted).

You can by the way buy bare FE310 chips for $25 for 5, and there is also the "LoFive" design done by a 3rd party. The board design is open source, you can make it yourself, and there have been a couple of GroupGets campaigns for assembled LoFives for $25.

https://groupgets.com/manufacturers/qwerty-embedded-design/products/lofive-risc-v

I just looked briefly but I did want to be fair because I just might be interested.  Sure, the board is pricey but so what, education has never been free.  I have a couple of dozen development boards going back 15 years or more.  I don't regret buying any of them.  OTOH, I'm not sure what to do with some of them...

There's quite a bit of documentation and it seems quite thorough.  I don't recall seeing such detail of the internal logic with any other device.

I mainly playing with FPGAs and the PSOC 6 at the moment but that can change in an instant.  Sometimes I think I bounce around like a BB in a bowling alley.

Bare chips for $5 seems very interesting. I prefer to use pre-manufactured boards but I have designed some from time to time.  Mostly I design daughter cards to break out various IO from the factory headers.
 

Offline brucehoult

  • Frequent Contributor
  • **
  • Posts: 759
  • Country: nz
  • Currently at SiFive, previously Samsung R&D
Re: Next step from AVR
« Reply #57 on: October 19, 2018, 02:31:50 pm »
Bare chips for $5 seems very interesting. I prefer to use pre-manufactured boards but I have designed some from time to time.  Mostly I design daughter cards to break out various IO from the factory headers.

Bare chips are here: https://www.crowdsupply.com/sifive/hifive1
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf