The result are different between the NXP simulator and the embedded system
Here are the two results for the first FFT16 with windowing:
From simulator:
0xa000; 0xfff31c +i 0x000dd6
0xa001; 0x0000af +i 0x0025f7
0xa002; 0x002316 +i 0x0020b1
0xa003; 0x0027a0 +i 0x0000a7
0xa004; 0x000ff9 +i 0xfff8c1
0xa005; 0x0008dc +i 0x000afa
0xa006; 0x0019a6 +i 0x0014a6
0xa007; 0x002b2e +i 0x000751
0xa008; 0x002688 +i 0xffeed4
0xa009; 0x000de9 +i 0xffeabf
0xa00a; 0x000880 +i 0xffff30
0xa00b; 0x001c5d +i 0xfffed0
0xa00c; 0x0017b7 +i 0xffe55d
0xa00d; 0xfffa10 +i 0xffe634
0xa00e; 0xfff364 +i 0xfffd30
0xa00f; 0xfff948 +i 0x0003d1
From embedded system:
0xa000; 0xfff31c +i 0x000dd6
0xa001; 0xfff948 +i 0x0003d1
0xa002; 0xfff364 +i 0xfffd30
0xa003; 0xfffa10 +i 0xffe634
0xa004; 0x0017b7 +i 0xffe55d
0xa005; 0x001c5d +i 0xfffed0
0xa006; 0x000880 +i 0xffff30
0xa007; 0x000de9 +i 0xffeabf
0xa008; 0x002688 +i 0xffeed4
0xa009; 0x002b2e +i 0x000751
0xa00a; 0x0019a6 +i 0x0014a6
0xa00b; 0x0008dc +i 0x000afa
0xa00c; 0x000ff9 +i 0xfff8c1
0xa00d; 0x0027a0 +i 0x0000a7
0xa00e; 0x002316 +i 0x0020b1
0xa00f; 0x0000af +i 0x0025f7
As you can see, the results are globally the same one but the order of the data is not the same in the memory.
For example, the data at address 0xa001 in the simulator is at the address 0xa00f in the embedded system. And so on…
This example computes a FFT16 of complexes in rounded mode(256 * FFT16).
The input data is at addresses 0xA000-0xAFFF in OPRAM, and the result is also at the addresses 0xA000 – 0xAFFF
With the SPT tools box in Matlab, we have a similar result as the NXP simulator.
Do you aware of this behavior?
Answered by private communication. The fix is available in 1.3.0 release