[WinAVR-users] [ winavr-Bugs-1980592 ] in asm: unable to use 64-bit operand

[SourceForge.net <no...@so...>]

Bugs item #1980592, was opened at 2008-05-31 15:09
Message generated for change (Tracker Item Submitted) made by Item Submitter
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=520074&aid=1980592&group_id=68108

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: None
Group: None
Status: Open
Resolution: None
Priority: 5
Private: No
Submitted By: Bernhard Seifert (berni_seifert)
Assigned to: Nobody/Anonymous (nobody)
Summary: in asm: unable to use 64-bit operand

Initial Comment:

using the avr-libc manual I've created a 32-bit byteswap function

uint32_t Math::reverse(uint32_t x)
{
   // Reverse byte order
   asm volatile
   (
      "mov __tmp_reg__, %A0  \n\t"
      "mov %A0, %D0          \n\t"
      "mov %D0, __tmp_reg__  \n\t"
      "mov __tmp_reg__, %B0  \n\t"
      "mov %B0, %C0          \n\t"
      "mov %C0, __tmp_reg__  \n\t"      
   :
      "+r" (x)
   );

   // Return value
   return x;
}

according to the manual %A0 is the lowest byte of parameter 1, %B0 the second-lowest, and so on. %A1 is the lowest byte of the second parameter and so on.
This works fine for up to 32bit values, but the upper 4 bytes of a 64bit value cannot be addressed. %E0 til %H0 does not work, %A1 to %D1 doesn't work neither.

I've found the following workaround to make a 64bit byte swap function:

// Reverse byte order
uint64_t Math::reverse(uint64_t x)
{
   // Reverse byte order
   asm volatile
   (
      "mov __tmp_reg__, r23  \n\t"
      "mov r23, r16          \n\t"
      "mov r16, __tmp_reg__  \n\t"
      "mov __tmp_reg__, r22  \n\t"
      "mov r22, r17          \n\t"
      "mov r17, __tmp_reg__  \n\t"
      "mov __tmp_reg__, r21  \n\t"
      "mov r21, r18          \n\t"
      "mov r18, __tmp_reg__  \n\t"
      "mov __tmp_reg__, r20  \n\t"
      "mov r20, r19          \n\t"
      "mov r19, __tmp_reg__  \n\t"
   :
      "+a" (x)
   );

   // Return value
   return x;   
}


by telling the compiler "+a" (x) the registers r16-r23 are used. But unnecessary moves between the registers make the function slower and larger.
In no manual/forum I've found a solution to this problem!


----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=520074&aid=1980592&group_id=68108