From: nowhere@nevernet.at   
      
   On 06/11/2022 09:33, muta...@gmail.com wrote:   
      
   > I wasn't aware that the 8086 didn't have FS and GS, but I never   
   > attempted to use them anyway.   
   >   
   > The author appears to be mistaken, calling it huge model when   
   > large model meets that description and is more normal.   
   >   
   > However, what about the claim that basically DS and ES are   
   > not enough? DS permanently pointing to the data section   
   > seems reasonable, and ES being loaded when you are   
   > accessing something not in the data section seems   
   > reasonable too.   
      
   there were four segment registers which all could be used to   
   access data: DS,ES,SS and also CS.   
      
   > Is this issue when you are trying to do a memcpy between   
   > two areas that aren't in the data section?   
      
   look at the block instructions:  DS:SI <-> ES:DI   
   and with override: SS:SI and CS:SI (but not for ES:DI)   
      
   > And temporarily reassigning DS is too inefficient? Is the   
   > average application 5% slower because of this? 1%?   
   > 0.1%?   
      
   time penalty depends on many things...   
   main usage for data access is/was:   
   DS current data block   
   ES graphic memory write   
   SS temporary variables   
   CS static values (code inherent)   
      
   > Is it really a mistake to not have FS and GS? I assume   
   > having those would have required more money per   
   > processor?   
      
   FS and GS are second page (prefix 0x0f) instructions.   
   IIRC they weren't available before 80386.   
   __   
   wolfgang   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)