From: user5857@newsgrouper.org.invalid   
      
   John Levine  posted:   
      
   > It appears that Terje Mathisen   said:   
   > >My "Algorithms and Data Structures" professor told us that IBM had   
   > >patented a zero-time sort chip, which they then promptly buried since   
   > >their internal stats said that over all IBM mainframes, 60% of the CPU   
   > >hours were spent in various forms of sorting.   
   > >   
   > >Terje   
   > >   
   > >PS. The chip was a ladder of comparators, so that you could use a   
   > >channel to stream data into it, then when you were done (or the chip was   
   > >full) you would run the channel transfer in the opposite direction to   
   > >retrieve the sorted data.   
   >   
   > Sounds like an urban legend to me.  For one thing, on machines of that   
   > era no useful sort fit in core so it spent most of its time doing disk   
   > and tape I/O anyway.   
      
   I remember hearing about the chip (circa 1986-8).   
   I agree that the number of entries would not have been enough to justify.   
   I also agree that Interrupt response time after the I/O would have   
   prevented any useful performance advantage.   
      
   By the mid-1990s the entry count could have been large enough to justify.   
      
   > For another, if it really did make a difference, they would have   
   > called it the Improved Sort Performance Facility and sold it as an   
   > extra cost option.   
   >   
   > These days z/Series has the optional Enhanced-Sort Facility added in   
   > September 2020 which adds a complex SORT LISTS instruction that does   
   > in-memory sorts or merges presuably in microcode so it can run at full   
   > memory speed.   
   >   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)