XPost: alt.folklore.computers   
   From: terje.mathisen@tmsw.no   
      
   Peter Flass wrote:   
   > On 8/6/25 10:25, John Levine wrote:   
   >> According to Anton Ertl :   
   >>> Lawrence D'Oliveiro  writes:   
   >>>> Not aware of any platforms that do/did ILP64.   
   >>>   
   >>> AFAIK the Cray-1 (1976) was the first 64-bit machine, ...   
   >>   
   >> The IBM 7030 STRETCH was the first 64 bit machine, shipped in 1961,   
   >> but I would be surprised if anyone had written a C compiler for it.   
   >>   
   >> It was bit addressable but memories in those days were so small that a    
   >> full bit   
   >> address was only 24 bits.  So if I were writing a C compiler, pointers    
   >> and ints   
   >> would be 32 bits, char 8 bits, long 64 bits.   
   >>   
   >> (There is a thing called STRETCH C Compiler but it's completely    
   >> unrelated.)   
   >    
   > I don't get why bit-addressability was a thing? Intel iAPX 432 had it,   
   > too, and it seems like all it does is drastically shrink your address    
   > space and complexify instruction and operand fetch to (maybe) save a few    
   > bytes.   
      
   Bit addressing, presumably combined with an easy way to mask the    
   results/pick an arbitrary number of bits less or equal to register    
   width, makes it easier to impement compression/decompression/codecs.   
      
   However, since the only thing needed to do the same on current CPUs is a   
   single shift after an aligned load, this feature costs far too much in    
   reduced address space compared to what you gain.   
      
   In the real world, all important codecs (like mp4 or aes crypto) end up    
   as dedicated hardware, either AES opcodes or a standalone VLSI slice    
   capable of CABAC decoding. The main reason is energy: A cell phone or    
   laptop cannot stream video all day without having hardware support for    
   the decoding task.   
      
   One possibly relevant anecdote: Back in the later 1990'ies, when Intel    
   was producing the first quad core Pentium Pro-style cpus, I showed them    
   that it was in fact possible for one of those CPUs to decode a maximum    
   h264 bitstream, with 40 Mbit/s of CABAC coded data, in pure software.   
   (Their own sw engineers had claimed that every other frame of a 60 Hz HD   
   video would have to be skipped.)   
      
   What Intel did was to license h264 decoding IP since that would use far    
   less power and leave 3 of the 4 cores totally idle.   
      
   Terje   
      
   --    
   -    
   "almost all programming can be viewed as an exercise in caching"   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)