GigaDevice RISC-V microcontroller

[technix]

One can always claim interoperability which is protected under anti-trust law (which we also have and we just struck Alibaba for ~US$3B anti-trust fine, and that is exactly the law that permitted Compaq and other IBM-compatible PC makers to exist alongside IBM.)

The first wave of IBM-PC clones were illegal. Here in New Zealand a PC-clone company called Exzel (?) were wiped out once IBM came after them and I'm sure there were some cases in the USA also.

It is legal to copy the hardware, but the BIOS code is copyrighted.

It wasn't until Phoenix clean-room reimplemented the BIOS -- one set of engineers documented what the IBM BIOS actually did, including bugs, and another set wrote bug-compatible code -- that IBM couldn't stop them. (COMPAQ earlier did the same thing, but only for their own use)


This is the main reason there were never any legal Mac clones. The IBM PC BIOS was 2k in size and rather badly written. The Mac ROM was 64k and written and rewritten to get it that small by very clever people such as Bill Atkinson and Andy Hertzfeld. QuickDraw was a work of genius, and at least one essential part, the Region Manager (responsible for representing arbitrary 2D shapes and clipping drawing to them and manipulating them with efficient union, difference etc operations), was patented.

That is why Chinese companies can make STM32 compatible products with its entire pinout, register layout and ID codes identical, and ST can do nothing with it unless they have still valid patents on some of the peripherals. Sadly for STM32F103 since all IP on that chip is either bought from a third party or lifted from the now patent-free STM8, there is no valid patent at all for them to base a suit on. The only thing ST can claim is the copyright on the built-in bootloader ROM, which is just way too easy to circumnavigate especially since those Chinese companies have zero willingness to keep the bootloader compatible and just rewrote it. This is likely also why you don't see code-compatible STM32F4 clones, as that chip has at least a redesigned GPIO controller that may be under patents, and no Chinese company will run the risk of being caught in a patent suit on that.

Just think ST being IBM, GD and CKS being Compaq, and Artery being that company that delivers PC compatibles with Phoenix BIOS and an NEC V20 in it. They run the exactly same code and have exactly the same external peripheral support. They bought the same components from the same external company (Intel for CPU and motherboard chips in IBM compatibles, versus ARM/Synopsys for CPU/USB IP in STM32 compatibles,) the compatible system makers reimplemented everything else (those custom IBM chips in PC versus that Faraday chip for IBM clones, versus reimplemented STM8 IP found in Chinese chips,) and there are also the brave ones that try to make improvements without breaking compatibility (the NEC V20 in a PC comaptible, versus the 240MHz Cortex-M4F in Artery chips)

As of the GD RISC-V chips, basically someone fitted an 65C816 onto an PC motherboard with an adapter.

[technix]

Should GD provided better software support, I would actually prefer GD32F103 over STM32F103, at least simply for the fact that it is using the Cortex-M3 r2p1 core instead of r1p1, which means there are some bugfixes and Cortex-M4 backported features.

[olkipukki]

There are now two ESP32s announced that are purely RISC-V, along with the earlier one with the main core still xtensa plus a low power RISC-V core.
...
I think the first ESP32s have already come through.

That's right, ESP32-C3 ES already available to a generic public.

On another level, the Allwinner D1 is already in mass-production.

Really? Did I miss something?

I cannot find D1 availability yet in the product portfolio, only announced in News dated 15th of April without mentioned a chip release date...

https://www.allwinnertech.com/index.php?c=news&a=index&id=289


re: CPU
I'm not sure why most offers  pre-build (BeagleV, HiFive etc) options only compare to PolarFire where you can buy either their dev board or off-shelve chips.  The latest allows DIY in house...
Hopefully D1 will change a game.

[brucehoult]

On another level, the Allwinner D1 is already in mass-production.

Really? Did I miss something?

I cannot find D1 availability yet in the product portfolio, only announced in News dated 15th of April without mentioned a chip release date...

https://www.allwinnertech.com/index.php?c=news&a=index&id=289

I mean it's going through TSMC right now, not that it's already in Mouser.

At the moment they have some initial chips and are distributing EVBs. I gave URLs where RVboards and Sipeed offered ssh access to one (which I've taken up, from both). I'm on a short list of a dozen people Sipeed will send EVBs to once they have some more. Allwinner has respun the board because only up to 512 MB RAM was working and they want to send them out to developers with more, so it's going to be a couple of weeks more.

I think Sipeed and Pine64 and embedded users will suck up the entire chip supply this year, and they'll be available at retail only in 2022.

re: CPU
I'm not sure why most offers  pre-build (BeagleV, HiFive etc) options only compare to PolarFire where you can buy either their dev board or off-shelve chips.  The latest allows DIY in house...
Hopefully D1 will change a game.

Not quite sure what you're getting at there. PolarFire SoC is 6 or 9 months ahead in ramping up manufacturing so you can get both chips and boards. But the last time I checked you could still only get the 500k LE version.

StarFive 7110 (4x SiFive U74) should be interesting too. AntMicro announced a module using it today, as well the the already known BeagleV (which I have a beta version of).

https://antmicro.com/blog/2021/04/arv-som-announcement/

The 7110 is not taped out yet, as they're waiting for feedback from the 300 beta BeagleVs they've distributed with the dual core no GPU 7100. [no GPU, but it does have HDMI output using the DSP (?) to drive it. I haven't tried that, I'm just using SSH to it via ethernet]

[brucehoult]

I also saw a weird claim today that the D1 and T1033 (dual-core Cortex-A7) are actually the same chip and either the RISC-V part of the ARM part is hidden or disabled depending.

I have *no* idea if there's anything to that.

[EEVblog]

Mods, please just lock this thread, it is completely unacceptable the way our collogues are talking again. And again. And again. There is no way to do intelligent discussion about this.

This until recently has been a valuable thread for those interested in GD chips.
If the tread gets locked then these same pro- and anti-China people will just go and pollute another thread and get it spoiled and/or locked too. We've already lost one to their bickering.
Better to ban some of them I think.

Blueskull will not bother us again.

[technix]

I also saw a weird claim today that the D1 and T1033 (dual-core Cortex-A7) are actually the same chip and either the RISC-V part of the ARM part is hidden or disabled depending.

I have *no* idea if there's anything to that.

Well it makes sense if they are hedging bets - should one of the two CPU options fail the silicon can still be sold somehow.

[technix]

I don't want to offend anyone but this probably is because most chinese people don't know how to write software in general Chinese people are not known for their software skills.

Chinese progammers do have some of the best coding skills in the world, but you won't see them in the embedded world because...

They know how to build cheap hardware which works (but that too is based on what software such as Synopsys spit out), but they seem to know next to nothing very little about software. For example, it seems like every week or so another chinese SBC is released without good software / linux support.

... software engineering and electronics engineering are considered two strictly separate areas of study, so those excellent coders have zero clue what a microcontroller is and usually works for Tencent or Alibaba which have barely any real interest in IoT other than ways to further spy on the general populace.

Also here in the Chinese education system computer science and programming are lumped together with math and theoretical physics - in fact most of the older professors in computer science have a background in pure mathematics; meanwhile electronics engineering is lumped with electrical engineering and mechanical engineering where programming is simply ignored. The only exception AFAIK is Lanzhou University which lacked a separate computer science program and placed that under electronics engineering department.

[coppice]

I also saw a weird claim today that the D1 and T1033 (dual-core Cortex-A7) are actually the same chip and either the RISC-V part of the ARM part is hidden or disabled depending.

I have *no* idea if there's anything to that.

There seem to be several MCUs with both an ARM and a RISC/V core. The core is a small percentage of an MCU die, so that makes a lot of sense. For dies where the cores are a bigger percentage of the total area it seems less likely. How much effort was put into making the RISC/V core designs bus compatible with ARM, so they could be swapped out easily?

[coppice]

For example, it seems like every week or so another chinese SBC is released without good software / linux support.

Chinese companies are no different to western companies in this regard. Look at the mess most MCU vendors make of their supporting software. Try talking to some people who develop things like drivers to support new ICs. The IC can only succeed as part of a complete package, but few vendors want to assign priority developing a really polished driver that lets the silicon shine. The only time the driver developer is held in high regard is when the silicon turns out to have a nasty bug, and everyone looks to the driver developer for a quick workaround.

[technix]

There seem to be several MCUs with both an ARM and a RISC/V core. The core is a small percentage of an MCU die, so that makes a lot of sense. For dies where the cores are a bigger percentage of the total area it seems less likely. How much effort was put into making the RISC/V core designs bus compatible with ARM, so they could be swapped out easily?

Well you can just convert whatever bus interface RISC-V uses into AXI4 and it will work in an existing AXI bus matrix. While ARM owns AXI4, AFAIK that license is super cheap.

[brucehoult]

There seem to be several MCUs with both an ARM and a RISC/V core. The core is a small percentage of an MCU die, so that makes a lot of sense. For dies where the cores are a bigger percentage of the total area it seems less likely. How much effort was put into making the RISC/V core designs bus compatible with ARM, so they could be swapped out easily?

Well you can just convert whatever bus interface RISC-V uses into AXI4 and it will work in an existing AXI bus matrix. While ARM owns AXI4, AFAIK that license is super cheap.

I think probably every RISC-V core that is actually used anywhere supports AXI for talking to peripherals.

AMBA (which includes AXI-3 and AXI-4) requires a license but it's zero cost. The only important restriction is that you're not allowed to include a CPU core that is code-compatible with an ARM instruction set unless you have separately licensed that.

[technix]

There seem to be several MCUs with both an ARM and a RISC/V core. The core is a small percentage of an MCU die, so that makes a lot of sense. For dies where the cores are a bigger percentage of the total area it seems less likely. How much effort was put into making the RISC/V core designs bus compatible with ARM, so they could be swapped out easily?

Well you can just convert whatever bus interface RISC-V uses into AXI4 and it will work in an existing AXI bus matrix. While ARM owns AXI4, AFAIK that license is super cheap.

I think probably every RISC-V core that is actually used anywhere supports AXI for talking to peripherals.

AMBA (which includes AXI-3 and AXI-4) requires a license but it's zero cost. The only important restriction is that you're not allowed to include a CPU core that is code-compatible with an ARM instruction set unless you have separately licensed that.

If that is the case, then I would assume there would be zero incentive not to create RISC-V based cores that uses AXI and AHB as the bus interface, since that would mean that existing AXI and AHB IP would work immediately.

Just wondering whether one can create a chip using multiple different architectures of CPU cores simultaneously... RISC-V have AXI version, obviously Cortex-A speaks AXI. And AFAIK Loongson has a MIPS-compatible core that is licensable and speaks AXI. What if we have a quad-core RV64G, a quad-core Cortex-A53 and a quad-core Loongson GS464 MIPS64 core complexes all on the same bus on the same chip, and somehow all of them can be powered on at the same time...?

[brucehoult]

Just wondering whether one can create a chip using multiple different architectures of CPU cores simultaneously... RISC-V have AXI version, obviously Cortex-A speaks AXI. And AFAIK Loongson has a MIPS-compatible core that is licensable and speaks AXI. What if we have a quad-core RV64G, a quad-core Cortex-A53 and a quad-core Loongson GS464 MIPS64 core complexes all on the same bus on the same chip, and somehow all of them can be powered on at the same time...?

Sure, why not? It's no different to having 12 cores of the same type.

Espressif already have an ESP chip with one xtensa core and one RISC-V core on the same bus. NXP have a chip with two small cores (one each ARM M0+ and RISC-V) and two bigger cores (one each M4F and RISC-V).

[technix]

Just wondering whether one can create a chip using multiple different architectures of CPU cores simultaneously... RISC-V have AXI version, obviously Cortex-A speaks AXI. And AFAIK Loongson has a MIPS-compatible core that is licensable and speaks AXI. What if we have a quad-core RV64G, a quad-core Cortex-A53 and a quad-core Loongson GS464 MIPS64 core complexes all on the same bus on the same chip, and somehow all of them can be powered on at the same time...?

Sure, why not? It's no different to having 12 cores of the same type.

Espressif already have an ESP chip with one xtensa core and one RISC-V core on the same bus. NXP have a chip with two small cores (one each ARM M0+ and RISC-V) and two bigger cores (one each M4F and RISC-V).

The problem is, how do the software juggle this mess of having 4 cores each using a different instruction set?

[brucehoult]

Hopefully they have different RESET vectors :-)

[technix]

Hopefully they have different RESET vectors :-)

At least ARM and RISC-V supports relocatable RESET vectors but all of them would power up with RESET vectors at address zero.

[tautech]

Mods, please just lock this thread, it is completely unacceptable the way our collogues are talking again. And again. And again. There is no way to do intelligent discussion about this.

This until recently has been a valuable thread for those interested in GD chips.
If the tread gets locked then these same pro- and anti-China people will just go and pollute another thread and get it spoiled and/or locked too. We've already lost one to their bickering.
Better to ban some of them I think.

Blueskull will not bother us again.

That is truly sad when one of the most qualified EE's here gets treated like this by other members.
I can only hope you will reconsider any permanent ban.

[technix]

That is truly sad when one of the most qualified EE's here gets treated like this by other members.

I can only hope that such treatment don't fundamentally came from the flag he is flying, especially in this time when international geopolitical tensions are at its maximum, and our field of profession is directly impacted and straight up named in a few official documents.

I can only hope you will reconsider any permanent ban.

It appeared to me that some other members here also acted less than rational here, yet they are allowed to walk away with just a slap on the wrist...

And on the topic of PC, someone's PC is someone else's illegal speech punishable up to death. So just be aware of that.

[tunk]

It appeared to me that some other members here also acted less than rational here, yet they are allowed to walk away with just a slap on the wrist...

I cannot help think the same - or were the others banned as well?

[technix]

It appeared to me that some other members here also acted less than rational here, yet they are allowed to walk away with just a slap on the wrist...

I cannot help think the same - or were the others banned as well?

Apparently not.

[EEVblog]

It appeared to me that some other members here also acted less than rational here, yet they are allowed to walk away with just a slap on the wrist...

I cannot help think the same - or were the others banned as well?

He had been baned before and has a history of years of provacations and reports from users, I was not going to allow this to continue.
Not to mention that one of his last posts was an admission that he directly targetted myself and the forum and reported me to forum advertisers in order to get their ads pulled.
Now please get this thread back on topic or it gets locked.

[coppice]

It appeared to me that some other members here also acted less than rational here, yet they are allowed to walk away with just a slap on the wrist...

I cannot help think the same - or were the others banned as well?

He had been baned before and has a history of years of provacations and reports from users, I was not going to allow this to continue.
Not to mention that one of his last posts was an admission that he directly targetted myself and the forum and reported me to forum advertisers in order to get their ads pulled.
Now please get this thread back on topic or it gets locked.

He also had years of being provoked by same rather unsavoury characters on this forum.

[tautech]

It appeared to me that some other members here also acted less than rational here, yet they are allowed to walk away with just a slap on the wrist...

I cannot help think the same - or were the others banned as well?

He had been baned before and has a history of years of provacations and reports from users, I was not going to allow this to continue.
Not to mention that one of his last posts was an admission that he directly targetted myself and the forum and reported me to forum advertisers in order to get their ads pulled.
Now please get this thread back on topic or it gets locked.

He also had years of being provoked by same rather unsavoury characters on this forum.

And despite politics not expressly allowed on the forum it flies under the radar when attacks on any country should be stomped on, factual or not.
If Dave was a generation older he would have seen the same trash talk of Japan in the 60's and 70's yet he jumps on members from the fastest growing nation we are ever likely to see in our lifetimes.

[EEVblog]

And despite politics not expressly allowed on the forum it flies under the radar when attacks on any country should be stomped on, factual or not.
If Dave was a generation older he would have seen the same trash talk of Japan in the 60's and 70's yet he jumps on members from the fastest growing nation we are ever likely to see in our lifetimes.

FFS Tautech, just stop it.
I asked nicely to get this thread back on topic.