In article <2025Jul5.104922@mips.complang.tuwien.ac.at>,   
   Anton Ertl  wrote:   
      
   >One interesting aspect is that VFX 5.x finally includes an FP package   
   >by default, and it started by including an SSE2-based FP package which   
   >supports a deep FP stack.  However, MPE received customer complaints   
   >about the lower number of significant digits in SSE2 (binary64)   
   >vs. 387 (80-bit FP values), so they switched the default to the   
   >387-based FP package that only has 8 FP stack items.  Apparently no   
   >MPE customer complains about that limitation.   
      
   Interesting indeed! I would rather expect customers to complain   
   about not IEEE compliance.   
      
   >   
   >OTOH, iForth-5.1-mini uses the 387 instructions, but stores FP stack   
   >items in memory at least on call boundaries.  Maybe Marcel Hendrix can   
   >give some insight into what made him take this additional   
   >implementation effort.   
      
   Once an assembler is in place, using only the internal stack   
   fp merely costs 23 screens in ciforth. This includes transcendental   
   functions, that are mostly a wrapper around an assembler instruction:   
       CODE FCOS FCOS, NEXT, END-CODE   
   The most involved are hyperbolic sine etc, that must be constructed   
   by combining exponentials and demands ranges.   
      
   I investigated the instruction set, and I found no way to detect   
   if the 8 registers stack is full.   
   This would offer the possibility to spill registers to memory only   
   if it is needed.   
      
   >- anton   
   --   
   The Chinese government is satisfied with its military superiority over USA.   
   The next 5 year plan has as primary goal to advance life expectancy   
   over 80 years, like Western Europe.   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)