From: redelm@ev1.net.invalid   
      
   R.Wieser  wrote in part:   
   > Robert   
   >>>> Well, please make sure the pointer is correct   
   >>> :-) And how do you propose that should be done ?   
   >>> It sounds like a great idea, but ...   
   >> use MOV.   
   >   
   > How would that change anything ?  If the target for   
   > an FXSAVE is wrong enough that it causes an exception,   
   > how /wouldn't/ that be in the same way wrong for a MOV ?   
   > (lets forget about alignment for a moment)   
      
   It is a purer memory test.  I thought there was question   
   of whether FXSAVE was available or supported on your CPU.   
   This checks opcode encoding too.   
      
   > It would even be making the problem larger, as you would   
   > than need to pick a REG value too - and wonder if it perhaps   
   > is having a negative influence on the result.   
      
   All GP registers should be available at all times.   
      
   > FWI, I tried several R/M values, none of which wanted   
   > to work.  Bad luck I guess.   
      
   Encoding should not be a guessing game.   
   The odds are bad, <1% .   
      
   > In retrospect I should perhaps have tried loading all the   
   > common registers with the same value and tried all R/M   
   > values until something worked.  On success it would be   
   > a case of determining which register is the source, and   
   > than look back at the instruction set to find a match -   
   > and from it figure out what the/my mistake was.   
      
   x86 has quirky indirect addressing modes that   
   are unlikely to yield to trial-and-error.   
      
   -- Robert   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)