C6678软件仿真与程序在开发板跑，指令运行周期数相差较大，其原因是什么？是工程配置的问题还是其他原因？需要哪些优化？

 
软件仿真与程序在开发板跑，指令运行周期相差较大，其原因是什么？是工程配置的问题还是其他原因？需要哪些优化？
测试平台C6678LE开发板，程序单核运行
具体测试结果如下：
软件仿真结果 ，simulation      
FFT点数                         运行时间                运行指令周期数数 
N = 8   radix = 2  TIME: 0.171000 us  CYCLES:171 
N = 16   radix = 4  TIME: 0.199000 us  CYCLES:199 
N = 32   radix = 2  TIME: 0.334000 us  CYCLES:334 
N = 64   radix = 4  TIME: 0.468000 us  CYCLES:468 
N = 128  radix = 2  TIME: 0.937000 us  CYCLES:937 
N = 256  radix = 4  TIME: 1.607000 us  CYCLES:1607 
N = 512  radix = 2  TIME: 3.676000 us  CYCLES:3676 
N = 1024  radix = 4  TIME: 7.018000 us  CYCLES:7018 
N = 2048  radix = 2  TIME: 16.543000 us  CYCLES:16543 
N = 4096  radix = 4  TIME: 32.685000 us  CYCLES:32685 
N = 8192  radix = 2  TIME: 76.258000 us  CYCLES:76258 
N = 16384  radix = 4  TIME: 152.048000 us  CYCLES:152048 
N = 32768  radix = 2  TIME: 348.709000 us  CYCLES:348709 
N = 65536  radix = 4  TIME: 696.883000 us  CYCLES:696883
开发板运行结果：
数据存放在DDR3中 定义DDR3段的时候采用.DDR3:load>>DDR3 采用load指令
FFT点数                         运行时间                运行指令周期数数 
N = 8   radix = 2  TIME: 0.215000 us  CYCLES:215 
N = 16   radix = 4  TIME: 0.243000 us  CYCLES:243 
N = 32   radix = 2  TIME: 0.401000 us  CYCLES:401 
N = 64   radix = 4  TIME: 0.617000 us  CYCLES:617 
N = 128  radix = 2  TIME: 1.207000 us  CYCLES:1207 
N = 256  radix = 4  TIME: 2.168000 us  CYCLES:2168 
N = 512  radix = 2  TIME: 4.793000 us  CYCLES:4793 
N = 1024  radix = 4  TIME: 9.256000 us  CYCLES:9256 
N = 2048  radix = 2  TIME: 21.034000 us  CYCLES:21034 
N = 4096  radix = 4  TIME: 51.208000 us  CYCLES:51208 
N = 8192  radix = 2  TIME: 266.581000 us  CYCLES:266581 
N = 16384  radix = 4  TIME: 539.959000 us  CYCLES:539959 
N = 32768  radix = 2  TIME: 1512.223000 us  CYCLES:1512223 
N = 65536  radix = 4  TIME: 3096.024000 us  CYCLES:3096024