home bbs files messages ]

Forums before death by AOL, social media and spammers... "We can't have nice things"

   comp.lang.asm.x86      Ahh, the lost art of x86 assembly      4,675 messages   

[   << oldest   |   < older   |   list   |   newer >   |   newest >>   ]

   Message 3,964 of 4,675   
   Anton Ertl to Melzzzzz   
   Re: FMA4 on Matisse?   
   21 Oct 19 08:36:49   
   
   From: anton@nospicedham.mips.complang.tuwien.ac.at   
      
   Melzzzzz  writes:   
   >FMA4 works on my 2700X. Question is: is that supported   
   >still on Zen2?   
   >Instructions are not reported at all by cpuid but works.   
   >I had some nice speedup with FMA4 on Zen.   
   >I don't know why FMA3 won when FMA4 is clearly superior?   
      
   Intel supports FMA3, but not FMA4.   
      
   If your question is why they do that, I can only speculate:   
      
   1) Well-placed register-register moves can have latency 0 and cost no   
   execution unit resources, only some front-end resources, because the   
   register renamer eliminates them.  So the advantage of FMA4 on Intel   
   may be miniscule (not sure if the Zen register renamer works the same   
   way; if so, maybe you can get similar performance with FMA3 if you   
   arrange the move appropriately).   
      
   2) FMA4 may require complications in the instruction decoder and   
   register renamer that the Intel engineers were not prepared to pay   
   for, given the small benefit.   
      
   - anton   
   --   
   M. Anton Ertl                    Some things have to be seen to be believed   
   anton@mips.complang.tuwien.ac.at Most things have to be believed to be seen   
   http://www.complang.tuwien.ac.at/anton/home.html   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)   

[   << oldest   |   < older   |   list   |   newer >   |   newest >>   ]


(c) 1994,  bbs@darkrealms.ca