From: bc@freeuk.com   
      
   On 30/11/2025 09:51, Philipp Klaus Krause wrote:   
   > Am 30.11.25 um 10:05 schrieb Michael S:   
      
   >> * Zilog Z80 based MCUs   
   >   
   > This one gets complicated. The original Z80 had a 4-bit ALU, but is   
   > widely considered 8-bit, and I'd agree.   
      
   That's news to me. Are you thinking of the 4040 as the original? Z80 was   
   a souped-up version of 8080: a superset with better technical specs.   
      
      
   > It has an 8-bit data bus, a 16-   
   > bit address bus. Most instructions operate on 8-bit data, but there are   
   > some that operate on 16 bits.   
      
   My first compiler for Z80 (not for C) supported u8, i16 and f24 types   
   (in modern parlance). That f24 just happened to be simpler for emulating   
   floating point.   
      
   Later this became u8, i16, u32 (I think) and f32. I never had a need for   
   odd sizes.   
      
   (Although some of the memory used was only 4-bit or 6-bits wide, this   
   was read or written via 8-bit data. For example 8KB of memory occupied a   
   16KB address range, because it was convenient for video. So there were   
   savings in memory, but via different means!)   
      
   I am looking at retargeting Z80 now (it'll be emulated), but I don't   
   have plans for non-power-of-two types.   
      
   If I wanted to emulate eZ80 with its 24-bit registers, then I'd need to   
   consider how best to do that. It's not as simple as just bolting on an   
   i24 or u24 type in the HLL, or (what's being discussed) employing an   
   ambitious type that can have an arbitrary N bit size.   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)