linear assembly optimization problem on ccs v5

when i  reuse c64x platform linear assembly (.sa files) on c66x platform,

the .sa function consume much more cycles on c66x platform(ccs v5) than on c64x (ccs v3.3). (18000cycles on ccs v5 , 2000 cycles on ccs v3)

Then i compare the .asm files generated from the same .sa file , and find out the optimization pipeline of ccs v3.3 is much better than the ccs v5.

compiler version:  7.3.0(ccs v5),  6.1.11(ccs v3). 

ps: on my ccs v5 project optmilization level o3 is already applied

i'm ancious about this problem, can anyone give me advice abt this issue?

Thanks!