155714Skrissolaris 2.5.1 usparc 167mhz?? - SC4.0 cc -fast -Xa -xO5 255714Skris 355714SkrisFor the ultra sparc, SunC 4.0 cc -fast -Xa -xO5, running 'des_opts' 455714Skrisgives a speed of 475,000 des/s while 'speed' gives 417,000 des/s. 5194206SsimonI believe the difference is tied up in optimisation that the compiler 655714Skrisis able to perform when the code is 'inlined'. For 'speed', the DES 755714Skrisroutines are being linked from a library. I'll record the higher 855714Skrisspeed since if performance is everything, you can always inline 955714Skris'des_enc.c'. 1055714Skris 1155714Skris[ 16-Jan-06 - I've been playing with the 1255714Skris '-xtarget=ultra -xarch=v8plus -Xa -xO5 -Xa' 1355714Skris and while it makes the des_opts numbers much slower, it makes the 1455714Skris actual 'speed' numbers look better which is a realistic version of 1555714Skris using the libraries. ] 1655714Skris 1755714Skrisoptions des ecb/s 1855714Skris16 r1 p 475516.90 100.0% 1955714Skris16 r2 p 439388.10 92.4% 2055714Skris16 c i 427001.40 89.8% 2155714Skris16 c p 419516.50 88.2% 2255714Skris 4 r2 p 409491.70 86.1% 2355714Skris 4 r1 p 404266.90 85.0% 2455714Skris 4 c p 398121.00 83.7% 2555714Skris 4 c i 370588.40 77.9% 2655714Skris 4 r1 i 362742.20 76.3% 2755714Skris16 r2 i 331275.50 69.7% 2855714Skris16 r1 i 324730.60 68.3% 2955714Skris 4 r2 i 63535.10 13.4% <-- very very weird, must be cache problems. 3055714Skris-DDES_UNROLL -DDES_RISC1 -DDES_PTR 3155714Skris 32