Question about a simple combinationnal custom instruction

The combinational block is now fully functional. Now I'm trying to build a multicycle custom instruction which performs filtering in 4 clock cycles. The waveform of my component is available in the attached files.

I have also attached the configuration of the interface in component editor; I specified the number of clock cycles required.

Here is my problem : when I go to the software part and I try my filter with that code :

# include <system.h>

# include <unistd.h>

int main()

{

int i;

int res=20;

//int j;

int tab[25]={256,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0};

for(i=0;i<25;i++)

{

res=ALT_CI_MY_FIR_CI_INST(tab[i]);

printf("Iteration %d ; mem = %d \n",i,res);

}

return 0;

}

I get the following results :

Iteration 0 ; mem = 0

Iteration 1 ; mem = -14

Iteration 2 ; mem = 10

Iteration 3 ; mem = 12

Iteration 4 ; mem = 16

Iteration 5 ; mem = 21

Iteration 6 ; mem = 25

Iteration 7 ; mem = 29

Iteration 8 ; mem = 32

Iteration 9 ; mem = 32

Iteration 10 ; mem = 32

Iteration 11 ; mem = 29

Iteration 12 ; mem = 25

Iteration 13 ; mem = 21

Iteration 14 ; mem = 16

Iteration 15 ; mem = 12

Iteration 16 ; mem = 10

Iteration 17 ; mem = -14

Iteration 18 ; mem = 0

Iteration 19 ; mem = 0

Iteration 20 ; mem = 0

Iteration 21 ; mem = 0

Iteration 22 ; mem = 0

Iteration 23 ; mem = 0

Iteration 24 ; mem = 0

These results are correct but there is a lag of one sample. Where could be the problem ?