LI77657: SUBOPTIMAL CODE FOR VECTOR LONG LONG ADD

XL C/C++ for Linux Fix Pack 8 (October 2015 Update) for 12.1
XL C/C++ for Linux Fix Pack 4 (October 2013 Update) for 12.1
XL C/C++ for Linux Fix Pack 5 (December 2013 Update) for 12.1
XL C/C++ for Linux Fix Pack 6 (February 2014 Update) for 12.1
XL C/C++ for Linux Fix Pack 7 (May 2014 Update) for 12.1

APAR status

• Closed as program error.

Error description

• For the simple test case below,
  extern "C" vector unsigned long long add64(vector unsigned long
  long a,
  vector unsigned long long b)
  {
    return a + b;
  }
  
  Output:
        | 000000           add64
      2|                   a,b,vs34,vs35
      4| 000000 vadduwm    vs32=vs34,vs35
      4| 000004 vaddcuw    vs33=vs34,vs35
      4| 000008 ld         gr3=.+CONSTANT_AREA(gr2,0)
      4| 00000C xxsldwi    vs0=vs33,vs33,1
      4| 000010 lxvdsx     vs1=+CONSTANT_AREA(gr3,0)
      4| 000014 xxland     vs33=vs0,vs1
      4| 000018 vadduwm    vs34=vs32,vs33
      5| 00001C bclr       lr
  
  Compiler could generate a code sequence that replaces xxsldwi +
  xxland with 1 vperm call.
  

Local fix

• Use the vec_addc and vec_perm builtins.
  

Problem summary

• USERS AFFECTED:
  Users using -qarch=pwr7 -qaltivec with vector long long types
  may be affected by this issue.
  
  PROBLEM DESCRIPTION:
  The compiler produces inefficient code for vector long long add
  routines.
  

Problem conclusion

• Fixed vector long long add to generate fewer instructions for
  -qarch=pwr7.
  

Temporary fix

Comments

APAR Information

• APAR is sysrouted FROM one or more of the following:

  IV36964

• APAR is sysrouted TO one or more of the following:

Fix information

• Fixed component name

  XL C/C++ FOR LI

• Fixed component ID

  5725C7300

Applicable component levels

• RC10 PSN IV36964

     UP06/09/13