Instructions that I missed

[Alienthe]

Has any of the FPGA wizards here perused the instruction set?

[Alienthe]

Arlet wrote:

Alienthe wrote:

How about licensing costs? ARM makes a pretty penny and I would still expect ARM cores to be more expensive than 6502.

Licensing costs could be a problem for ASIC development, which is where 6502 cores (or similar) would have an advantage. If you buy off-the-shelf ARM devices, like I do, the licensing cost would be rolled in the price of the device. In practice, these prices are quite competitive.

Seems Andes Technology has found the market ripe and is possibly snapping after ARM's lunch. They have been working for a few years and surfaced recently with a few licensable cores and systems.

http://www.andestech.com/en/index/index.htm

I haven't found detailed descriptions of the ISA but what I find indicated 16, 32 and 64 bit features. It seems reasonable to assume these cores are bigger and more costly than 6502, yet the Andes devices are in part promoted as 8 bit alternatives. So there should be still a lot of life left in the 6502 a a commercially used core.

[BigEd]

Hmm, they look very ARMish. I don't see any 64bit features. And I suspect the "16-bit instruction set" just means a short-length instruction stream, like Thumb, which still operates on 32bit registers and memory.

See
http://git.denx.de/?p=u-boot.git;a=blob ... ADME.NDS32
http://git.denx.de/?p=u-boot.git;a=blob ... ADME.N1213
http://ieeexplore.ieee.org/xpl/login.js ... er=6212638

oops: I notice

Quote:

- Multiply-add and multiple-sub with 64-bit accumulator.

[Alienthe]

BigEd wrote:

Hmm, they look very ARMish.

That was my thought too.

As for the 64 bit accumulator that could in theory be like in 6809 a concatenation of two 32-bit registers.

Unfortunately I don't have IEEE access, could you give a summary?

[BigEd]

(I only had the IEEE abstract, sorry)

[Alienthe]

Seems the 8-bit market remains of interest. I had never heard of Cortus before but now it seems they too have a 32 bit CPU with 16 registers for replacing 8-bit microcontrollers. At first glance it too looks similar to the Andes Technology cores. They boast of having a compact core but I would believe the genuine article from WDC is still more compact.

The sales volume stated by WDC probably makes others eager to get a share of this market.

[BigDumbDinosaur]

Alienthe wrote:

The sales volume stated by WDC probably makes others eager to get a share of this market.

WDC does have a competitive advantage at present—several, actually. One is that their development costs were amortized long ago, which does a lot for profitability. The result is that cost per part (don't forget that they are a fabless company) or per core license can be very low and still produce good revenue. This advantage can make it tough for others to move in. If anyone puts cost pressure on WDC they are in a excellent position to counter.

Another advantage that WDC has is that a huge body of 65xx code exists and is in daily use in all sorts of embedded applications (my microwave oven being one of them). Any company producing 65xx-based embedded devices would probably be hesitant about switching to different technology, since they might be faced with rewriting tested code or worse yet, starting from scratch on new code.

Finally, the 65C02 core has been certified for critical medical applications. If you know anyone who has an implanted defibrillator (I know two such individuals) it is highly likely that their defib unit depends on a 'C02 in ASIC form. No company that produces such a device is going to lightly entertain switching to a different MPU architecture—too much has already been invested in a functional design.

[Arlet]

I imagine it would be very hard to steal any of WDC's market share for people still using existing 6502 code. And if you need more power, and are willing to pick a new CPU, it's hard to avoid ARM. ARM has a wide range of different core options, all very well tested and proven in billions of devices, with excellent tool support.

[GARTHWILSON]

Quote:

Seems the 8-bit market remains of interest. I had never heard of Cortus before but now it seems they too have a 32 bit CPU with 16 registers for replacing 8-bit microcontrollers.

The microcontroller stuff I've done for commercial products spent an extremely small percentage of the computing time handling values of more than 8 bits, meaning that performancewise, the number of bits (ie, 8) has not presented any significant performance bottleneck for these.  The other reason to go to wider processors with lots of registers seems to just be the matter of abandoning assembly language in favor of higher-level languages.  With structure macros however, much of the benefit of higher-level languages comes down to be enjoyed in assembly.  If you're doing a lot of heavy math, then the higher-level languages are easier to work with; but when it's mostly I/O-intensive bit-twiddling and simple register comparisons and no math more complex than addition and subtraction, assembly does well.  I've done the bit-twiddling-intensive stuff in the best BASICs too, and it's kind of a pain compared to assembly.  I have completed one assembly-language project with the structure macros so far, and wow—what a difference.