It is now becoming more and more obvious that the SuperH architecture has no future.

1) A 64-bit SuperH instruction set has been announced fifteen years ago but no actual SH-5 hardware was ever launched.
2) Development of new SuperH dual-core processors has been ceased about five years ago.
3) Most new SuperH designs have been based on the SH-2A core used in the automotive industry. (Casio uses SH-3/SH-4A)
4) The SuperH series is no longer actively promoted on Renesas' website.
5) There are quite a few documents describing how to migrate SuperH software to newer Renesas products.
6) The RX series uses a proprietary RXv1/RXv2 CISC instruction set and could be a replacement for SH-2 cores.
7) The RZ series combines an ARM core with some proprietary modules and could be a replacement for SH-3/SH-4 cores.

So what does this mean?
If Casio calculators are not going to change much in the next decade, then nothing will happen, but if they do and if a more powerful processor is needed, then Casio will most likely switch to Renesas' RZ series and as such, the ARM instruction set will be used.

An interesting article on the new RZ series and the collaboration with ARM Limited:
https://www.renesas.com/en-eu/about/web-magazine/edge/solution/15-rz.html

"Migrating from SH-4A to Cortex-A" document which might become important in the future:
http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.dai0314a/index.html

"The discussion in this document uses the SH7724 series of devices as a reference point for devices supporting the SH-4A architecture. This is a single-core device aimed at mobile applications." (Casio's latest SH7305 processor is a variant of the SH7724)
Fascinating, thank you for sharing these details. I'm interested to know about some of the other applications in which Renesas SuperH processors are used, because even as someone in Systems I hadn't heard of the architecture before working with the Prizm. It seems to me that Casio would probably hurt themselves to continue to use such a niche architecture in their calculators, and conversely, if they switch to ARM processors, then I see no reason why they would continue purchasing Renesas's ARM processors in particular. Do we know if Renesas and Casio have some particular business relationship?
That's pretty interesting. I think it would make sense for them to continue to use the processor in their ASICs (right?) if they maintain the rights to the architecture (I'm pretty sure this is sorta how this works), so they don't have to restructure their whole product.

I could see many of the same arguments being used to argue that it was a silly thing for TI to use the ez80 in the CE. The ez80 is an uncommon enough architecture, at least in consumer-level devices, such that the only real reason for TI to use it was source compatibility with the z80 and the slight speed boost, which is mainly from the LCD interface anyway.
I guess the Casio<->Renesas relationship is pretty strong.

They are using custom Renesas processors, custom Renesas display drivers/controllers, and even the SRAM is rumored to come from Renesas. Also, Renesas is located in Japan, so Casio may get those components a lot cheaper. And I don't think that Casio holds any rights on the SuperH architecture.
I began suspecting SuperH was coming to a dead end when two or three years ago I looked into the architecture details and realized that at a time when ARM was clearly leading the market for embedded processors and 64-bit ARM was coming to market, SuperH had nothing equivalent materialized, as you point out.

I don't know of any consumer devices using it besides Casio calculators and electronic dictionaries (I'm not saying they don't exist, just that I don't know about them, which speaks to the low popularity of the architecture). Everything that is not ARM tends to be MIPS or some sort of closed platform like those used in the RX series.

I really wonder what makes Casio stay using SuperH (presumably, the fact that their engineers already know the architecture all too well? and making software easier to port, I guess). And they seem to try to use Japanese stuff whenever possible, even though nowadays (the Prizm at least) their stuff is made in China. For the filesystem on the Prizm, they went with a Japanese solution. Another thing is, they always go for custom chips / customization of existing Renesas cores even though I don't see why standard Renesas stuff wouldn't work too. The SH7305 appears to still have some Casio-specific CPU modules, which do stuff like BCD math, but I'm not even sure if that's still used or if it's all done in software.

The good thing about SuperH is that as patents for its older versions, people are free to do open hardware implementations of them, with the advantage of having vast toolchain support for the architecture. But I don't see what that would bring over an arch like RISC-V that's royalty-free since its inception.

RE the last TeamFX post: yes, that appears to be the case. I don't see how custom cores would be economical otherwise, especially when apparently part of the features go unused (and yes, I know the same SH7305 is also used in dictionaries...). It's almost as if Renesas was begging them to order custom cores! Perhaps Casio products account for a major part of Renesas CPU sales?
Quote:
I really wonder what makes Casio stay using SuperH.

Well, everything works and porting stuff to a new architecture costs a lot of time and money.

Originally, they moved to SuperH because the HP Xpander was utilizing an SH-3 CPU and they hired Saltire Software which were responsible for some of the software designs when the product was canceled in 2001. Casio already had some experience with SH-3 used in their Cassiopeia A-10/11/20 PDAs in 1997/1998. The SH-3-based PV-S1600 and ClassPad 300 were launched in late 2002 and early 2003, respectively.
I just realized the patents for SH4 expire this year. With no SuperH successor on the horizon, this may be an additional reason for Renesas to want to migrate costumers to other architectures they have more control over. Continuing on my previous point about SuperH patents, this is great news for open hardware (SH4 is a well-supported architecture with a MMU, no less!) and I expect that, soon, royalty-free SH4 designs will begin to appear.

Should Casio's relationship with Renesas go the wrong way, with some creativity they could even use this to their advantage: they could find another partner to produce SH4 chips based on those royalty-free designs and that way stick with SH4 for a few more years. But this is an unrealistic outcome, because it is so unlikely - it implies the end of the Casio-Renesas relationship, when it is so strong (as everything TeamFX mentioned would indicate).
IMO nearly all other alternatives would be preferable: stay with Renesas and SH4, stay with Renesas but with another (Renesas-specific?) architecture, or move to ARM and forget Renesas.
RISC-V might be a good replacement for SuperH. It is becoming more and more popular with every year.
Unlike ARM or MIPS, RISC-V is completely open source (similar to a BSD license) and requires no fees or NDAs.

https://en.wikipedia.org/wiki/RISC-V
The SH2 and SH4 architectures may become more relevant. See: http://j-core.org/ and https://lwn.net/Articles/647636/. It is similar in spirit to RISC-V.
Sure, but it might be too late as RISC-V is taking off fast!
I'm also not sure if SuperH is a good choice for all that Internet of things stuff that is coming.
Let's face it, it's a design from the '90s... Razz
Switching to RISC-V or a J-core-based solution would require switching to a supplier other than Renesas, something they might not be in a position to do. From what I've seen, certain Japanese companies much prefer domestic solutions even when there are companies in other countries doing better (and sometimes cheaper) stuff. I think it has to do with their culture.

J-core appears to be basically dead in the water, judging by the level of activity on their mailing list and stuff. It's not like there's a lot of things left to do, anyway: there's a VHDL implementation, the Linux kernel apparently has mainline support for it, and it seems to excel at what it is: basically a rename of the SH2 ISA with a couple of backported instructions. From here I guess they can implement the "new" stuff in SH4, but after that, there's not much to do... SuperH will always be a decades-old design, and if they want to "innovate"... for innovation there's RISC-V and its future successors.

If Casio moved to J-core that would mean their relationship with the SuperH ISA is actually stronger than the relationship with Renesas... I think it's the other way around: Renesas first, SuperH second. They would also need to find someone to build the chips for them, because as far as I know there are no J-core ASICs in the wild.

RISC-V is taking off fast, but I don't see Casio jumping in so soon. In five years, maybe, but right now I think it may be perceived as too risky of a move. ("risky", mainly because it's new and might be perceived as lacking real-world testing/usage) Plus, the whole "open-source" thing may scare corporate types - even though it's just a innocuous BSD license...

My guesses for the next year or two are that any new graphic calc they release will either iterate over previous SH7305 designs (eventually, at higher and higher clock speeds - and bigger batteries - until the practical limit for the core design is reached), or they may move to a ARM offering by Renesas.
In fact, they may have already moved, and are now busy porting their completely bug-free code to the glorious ARM...
My understanding, from someone I know who builds embedded stuff around Renesas chips for a living, is that Renesas is pitching RX as the replacement to SuperH on the lower end, and RZ at the higher end. The SH7305 honestly falls into the RX's bracket more than RZ, but no telling what Casio might do.

ARM is definitely possible, but I wouldn't rule out RX as a possibility. It's a pretty good chip, too; I've been hacking away at the RX63N in particular lately. I wouldn't mind that kind of chip in a calculator...
The RX series appears to only support on-chip RAM with sizes that are fractions of a MB. Powerful, yes, but not memory plentiful - it could be enough if Casio's software was designed for little memory, like TI's software for old Z80 calcs, but that is not the case.
The SH7305 on the Prizm works with 2 MB of RAM, and on the fx-CP400 Classpad, 16 MB. The fx-CG50 certainly won't have less than 2 MB of RAM. It's 2017, not 1997: I don't think Casio developers would enjoy working with 256 KB of memory or less.
Apparently only the RZ series can deal with more memory - those can have big internal SRAM and work with external memory with address spaces over 64 MB, and if Casio had to choose between RX and RZ I think they'd definitely go with RZ.
This is a bigger hardware and software transition for Casio. Similar to their transition from NEC V30MX and V30MZ to SH-3 in 2002. So they could be even considering moving away from Renesas.

Btw, Renesas SP has been acquired by Synaptics. Casio's Prizm and ClassPad II display controllers came from Renesas SP.
  
Register to Join the Conversation
Have your own thoughts to add to this or any other topic? Want to ask a question, offer a suggestion, share your own programs and projects, upload a file to the file archives, get help with calculator and computer programming, or simply chat with like-minded coders and tech and calculator enthusiasts via the site-wide AJAX SAX widget? Registration for a free Cemetech account only takes a minute.

» Go to Registration page
Page 1 of 1
» All times are UTC - 5 Hours
 
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum

 

Advertisement