While from a theoretical perspective, I tend to favor the adoption of
32-bit Architectures for their greater flexibility, I also dislike the wasting of resources that universal adoption would imply.
Hoping for some comment from informed people (myself not having any direct involvement in embedded systems), I present the following bullet points. (Note that I did not include the include the code size advantage that is sometimes claimed for 32-bit Architectures [Apparently, the older smaller Architectures do have a larger average code size for many uses, but I am not comfortable including such since it seems to be an artifact of such Architectures. On the other hand, I _did_ include Architectural legacy factors in favor of 8-bitters {legacy software, ISA familiarity, vendor relationships, and even lack of patent barriers}, so this might be a bit unfair.]. )
32-bit advantages:
*Greater flexibility of use
**Potential for future addition of functionality
**Reduced pressure to optimize for resource utilization (lower development cost and potential for simpler [more reliable and more provable] design)
**Potential for using the same part for multiple products (volume discounts, inventory control benefits [demand variability averaged over more products, parts usable for new products already in stock, less complex inventory--fewer disastrous errors])
**Potential for broader familiarity with ISA and systems within the company and throughout the industry
**Potential for higher volume/lower cost from larger market
*Larger processor
**Encourages use of newer process and greater integration to reduce the fractional cost of the processor
***Potentially lower power consumption
***Smaller end-product form factor and potentially less expensive packaging/shielding
***Potentially greater reliability
**Easier programming
***ISA support of more hardware-expensive operations
***Larger, orthogonal register set
***Greater performance (and possibly larger memory) can reduce required optimization effort
*Simpler programming
**Large flat address space
**Reduced need for software extended precision arithmetic (possibly also reduced need for overflow guards with intermediate results)
8-bit advantages:
*History
**Support of legacy device software
**Existing familiarity with ISA and systems
**Established vendor relationships
**Better vendor understanding of market requirements (ISA and systems tweaked for target markets)
**Expiration of patents (as well as simpler processor) reduces barriers to entry, increasing potential for competition
*Processor size
**Smaller processor encourages less extensive integeration
***Use of older process technology reduces device cost
***Easier to justify cost of custom additions
***Potentially less expensive field replaceable units
***Potentially greater isolation of failures
**Potential for lower power consumption by work minimization and lower transistor count (leakage)
**Potential for greater reliability (less to go wrong)
*Lesser flexibility of use
**Greater restraint against feature-creap
**Greater incentive to optimize resource utilization (mainly beneficial for higher volume products)
**Potentially greater reliability by less emphasis on microarchitectural and device techniques to increase performance, reduce area, and/or reduce power consumption
Since I am not directly involved with embedded systems, the above is doubtless missing some significant points. Let the sharing of information begin!
Paul A. Clayton just a technophile