usparc.cc 1.1 KB

12345678910111213141516171819202122232425262728293031
  1. solaris 2.5.1 usparc 167mhz?? - SC4.0 cc -fast -Xa -xO5
  2. For the ultra sparc, SunC 4.0 cc -fast -Xa -xO5, running 'des_opts'
  3. gives a speed of 475,000 des/s while 'speed' gives 417,000 des/s.
  4. I believe the difference is tied up in optimisation that the compiler
  5. is able to perform when the code is 'inlined'. For 'speed', the DES
  6. routines are being linked from a library. I'll record the higher
  7. speed since if performance is everything, you can always inline
  8. 'des_enc.c'.
  9. [ 16-Jan-06 - I've been playing with the
  10. '-xtarget=ultra -xarch=v8plus -Xa -xO5 -Xa'
  11. and while it makes the des_opts numbers much slower, it makes the
  12. actual 'speed' numbers look better which is a realistic version of
  13. using the libraries. ]
  14. options des ecb/s
  15. 16 r1 p 475516.90 100.0%
  16. 16 r2 p 439388.10 92.4%
  17. 16 c i 427001.40 89.8%
  18. 16 c p 419516.50 88.2%
  19. 4 r2 p 409491.70 86.1%
  20. 4 r1 p 404266.90 85.0%
  21. 4 c p 398121.00 83.7%
  22. 4 c i 370588.40 77.9%
  23. 4 r1 i 362742.20 76.3%
  24. 16 r2 i 331275.50 69.7%
  25. 16 r1 i 324730.60 68.3%
  26. 4 r2 i 63535.10 13.4% <-- very very weird, must be cache problems.
  27. -DDES_UNROLL -DDES_RISC1 -DDES_PTR