AVR Freaks

Or 18-bit using two 9-bit bytes!   See PDP-10

Yes LE, no, all instruction fetches are 1 word at a time (16 bit) All op codes are 16 bit (1 word) or 32 bit (2 word), yet it is considered to be an 8 bit micro, confused?  

Welcome to AVR!

Once you figure out the opcodes are all 1 or 2 words in size, it begins to make sense, there is no need for byte addressing within the program memory space, and as it's Harvard Arch, there

are separate memory spaces for program and data, memory fetch size can be different between the various memory spaces....

Fly over Jim

See AVR Instruction Set manual for details

Note generically, locations (addresses) and amount stored there can be decoupled.

For example you could make some digital system having an address of only 1 byte width (total 256 different possible addresses) and each address contains 7 bytes of information to be retrieved.

This was my point a few comments back...an address is just saying where you will get the information (partitioning of storage)...how much is there could be any amount, depending upon the arrangement of memory resources.

You could have 2k different addresses of 64 bit words each.   Those 64 bit words could be opcodes (kinda wide!!), and maybe reserve 11 or 22 of those bits to reference the 2k address space (for opcodes needing to use addressing).   

My CDC6400 COMPASS is way rusty, but conceptually not much different from AVR in terms of approach.  Word size wider, so there were 1/2/4 instructions per word versus the AVR's 1 or 2 words per instruction.  From Wikipedia:

Instructions were either 15 or 30 bits long, so there could be up to four instructions per 60-bit word. A 60-bit word could contain any combination of 15-bit and 30-bit instructions that fit within the word, but a 30-bit instruction could not wrap to the next word. The op codes were six bits long. The remainder of the instruction was either three three-bit register fields (two operands and one result), or two registers with an 18-bit immediate constant. All instructions were 'register to register'. For example, the following COMPASS (assembly language) code loads two values from memory, performs a 60-bit integer add, then stores the result: ...

Who exactly is "they", and what do you want "them" to change?  Give an example.  I'm sure that "they" would walk-back everything to the 1990's and change all of the documents relating to the invention of the AVR and change all the datasheets and go back even further and change the instruction set to give each instruction half the range (think about branch targets) so that you do not need to do any arithmetic.  While you are at it, I find IEEE 754 awfully confusing, as with the AVR an extra bit is forced/assumed for both exponent and mantissa.  Please have it changed.  From Wikipedia:

Further, the exponent is not represented directly, but a bias is added so that the smallest representable exponent is represented as 1, with 0 used for subnormal numbers. For numbers with an exponent in the normal range (the exponent field being neither all ones nor all zeros), the leading bit of the significand will always be 1. Consequently, a leading 1 can be implied rather than explicitly present in the memory encoding, and under the standard the explicitly represented part of the significand will lie between 0 and 1. This rule is called leading bit convention, implicit bit convention, or hidden bit convention. This rule allows the binary format to have an extra bit of precision. The leading bit convention cannot be used for the subnormal numbers as they have an exponent outside the normal exponent range and scale by the smallest represented exponent as used for the smallest normal numbers.