From: andrews@sdf.org   
      
   On 2024-02-11, The Natural Philosopher  wrote:   
   >   
   > An assemblers is - or ought to be - a 1:1 translator from human readable   
   > to machine readable commands.   
      
   There is _plenty_ of scope for an assembler to choose whatever   
   opcodes it wants.  Consider the simplest case of a NOP - some   
   architectures have a specific NOP opcode, others it is simply a   
   shorthand for an operation that does nothing, e.g. add 0 to a   
   register and so on.  Regardless there are plenty of alternatives   
   that can be chosen.  Things get more opaque once addressing modes,   
   the size of jumps, immediate operands and so on are considered.   
   I'm reminded of the warning in the A86 manual "this assembler   
   generates a unique fingerprint in these cases, which I can detect   
   in the binary whether you are registered or not".   
      
   > If the compiler you use to compile   
   >> your clean room compiler is poisoned then so will be the compiled compiler   
   >> despite your clean room code. That's the Thompson trap.   
      
   There's a lot of mysticism that has been attached to that over the   
   years, mostly by people who have never read the original report.   
   It wasn't some magical AI code fairy that could identify that you   
   were compiling any abstract compiler or login program and automatically   
   conjure up appropriate code for the circumstances, it used   
   fingerprinting IIRC at the token level (i.e. after the code is   
   broken in to "words", but before parsing to figure out how those   
   "words" are associated with each other).  An independent implementation   
   of functionally equivalent code, or even the same code after heavy   
   edits over time, would not be affected.   
      
   > The ARM is a special CPU that was designed initially to beat the 6502   
   > and walk all over z80s and 8080s.   
      
   No, it was designed for the Archimedes, pure and simple. The 8086   
   was already one the market thus the rest of the industry essentially   
   leapfrogged 16 bit and jumped straight to 32 bit.  To this day 16   
   bitters are few and far between.  There's 8086-80286, MSP430, and...   
   err... Well there's the original 68000 but that was 32 bit from a   
   software viewpoint.   
      
   > Because they couldn't afford massive wafers, it was strictly limited in   
   > hardware. All they could do was a very basic instruction set and a lot   
   > of on-chip registers. And a three instruction set pipeline and clock it   
   > as fast as it would go.  And a 32 bit address bus. To take advantage of   
   > a lot more RAM that was getting cheaper by the day. The low power was   
   > simply a cost saving measure - a plastic cased low dissipation chip was   
   > *cheaper*.   
   >   
   > And a few - maybe only one - very bright boys (sophie wilson) , looked   
   > at the absolute minimum of what those instructions had to do.   
      
   That's a very romanticised view, it often happens in science and   
   engineering when one of the characters has an interesting personal   
   story, Alan Turing and Stephen Hawking would be others that come   
   to mind.  The feature set was a committee effort, the high level   
   design was Roger/Sophie Wilson and the low level Steve Furber.   
   But as above, it was designed for the Archimedes, no more and no   
   less.   
      
   The primary design objectives were a low per-unit cost (not design   
   cost as sometimes stated) and a minimum of glue logic between major   
   subsystems.  I recall seeing a "triangle" diagram with the corners   
   cut off, the centre of the triangle was the CPU, the corners were   
   memory controller, graphics, and peripheral bus.   
      
   You're correct to identify a plastic package as a design criteria,   
   from memory the target was £2/chip which implied that over a ceramic   
   one.  None of the group had any chip design experience, they knew   
   a plastic package meant no more than a 1-2W power dissipation, but   
   had no idea what that meant in terms of design.  Thus they optimised   
   for power at every opportunity and undercut the target by orders   
   of magnitude.   
      
   The other dimension to lowering the cost of the package was reducing   
   pin out to the bare minimum, hence the 24 bit (not 32 bit) address   
   bus.  Size of the wafer was an irrelevance since they never baked   
   their own chips, die size yes they wanted to keep small to lower   
   cost but not an over-riding consideration - it wasn't that much   
   smaller than many other designs of the period.   
      
   This is from my lecture notes and also a couple of pints while at   
   Uni 25 years ago.  The lecturer for hardware design was none other   
   than Steve Furber who co-designed and literally wrote the book on   
   the thing.   
      
   --   
   Andrew Smallshaw   
   andrews@sdf.org   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)