Variable Optimized Away

Jun 09, 2011 5 Replies
C

Hello, I am having a problem where Xilinx ISE is trimming away a variable that I need, tempShifts. I can still implement the module but I am not sure what it is doing with my variable. This says to me that tempShifts isn't being trimmed... Explanations?



Also this module uses the shift and add 3 algorithm to convert binary to BCD format. Specifically this converts a 16 bit binary number into



5 BCD digits.

module binaryBCD( input [15:0] binaryIn, //input 16bit number output reg [19:0] bcdOut //5 display BCD output );



reg [35:0] tempShifts; //temporary storage while computing



always @(binaryIn) begin tempShifts = 0; //the method used is shift and add 3 //this requires 16 total shifts, first three are done here tempShifts[18:3] = binaryIn;



repeat(13) begin //for the method if any BCD digit is over 4 add three then shift if (tempShifts[19:16] > 4) tempShifts[19:16] = tempShifts[19:16] + 3; if (tempShifts[23:20] > 4) tempShifts[23:20] = tempShifts[23:20]+ 3; if (tempShifts[27:24] > 4) tempShifts[27:24] = tempShifts[27:24] + 3; if (tempShifts[31:28] > 4) tempShifts[31:28] = tempShifts[31:28] + 3; if (tempShifts[35:32] > 4) tempShifts[35:32] = tempShifts[35:32] + 3;



tempShifts = tempShifts


Vote
Twitter Facebook LinkedIn
G

XST has a brain-dead way of giving warnings about intermediate results. Basically the only bits of tempShifts that are used are effectively renamed to bcdOut. By the way I was under the impression that the repeat operator was not synthesizable. Does this code actually work in the hardware?

-- Gabor

Vote
Twitter Facebook LinkedIn
C

dd 3

first three are done here

y BCD digit is over 4 add three then

9:16] > 4)

tempShifts[19:16] =3D tempShifts[19:16] + 3;

3:20] > 4)

tempShifts[23:20] =3D tempShifts[23:20]+ 3;

7:24] > 4)

tempShifts[27:24] =3D tempShifts[27:24] + 3;

1:28] > 4)

tempShifts[31:28] =3D tempShifts[31:28] + 3;

5:32] > 4)

tempShifts[35:32] =3D tempShifts[35:32] + 3;

empShifts > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0end

Yea I have successfully implemented this code on a Nexys 2 development board, XCS3500E FPGA.

Vote
Twitter Facebook LinkedIn
C

t

add 3

s, first three are done here

any BCD digit is over 4 add three then

[19:16] > 4)

=A0tempShifts[19:16] =3D tempShifts[19:16] + 3;

[23:20] > 4)

=A0tempShifts[23:20] =3D tempShifts[23:20]+ 3;

[27:24] > 4)

=A0tempShifts[27:24] =3D tempShifts[27:24] + 3;

[31:28] > 4)

=A0tempShifts[31:28] =3D tempShifts[31:28] + 3;

[35:32] > 4)

=A0tempShifts[35:32] =3D tempShifts[35:32] + 3;

tempShifts > > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0end

s

To implement this wouldn't it just copy the structure as many times as needed, in this case 13 times?

Vote
Twitter Facebook LinkedIn
A

here

three then

3;
3;
3;
3;
3;

Doesn't this take an awful amount of logic resources ? If this is for a display, or other human interface, there's generally no requirement to update it faster than people can read. In that case, a version with a state machine that requires N clocks would probably work just as well, and be much more compact.

Vote
Twitter Facebook LinkedIn
G

here

three then

3;
3;
3;
3;

I presume it should be the same as: for (i = 0;i < 13; i = i + 1) However in the past I have never tried "repeat" other than in test benches. My "Verilog Golden Reference Guide" has this to say about repeat for synthesis:

"Only synthesizable with some tools, and only then if the loop is 'broken' by a clock event..."

Apparently tools have advanced a bit since this book was written.

-- Gabor

Vote
Twitter Facebook LinkedIn

Join the Discussion

Have something to add? Share your thoughts — no account required.

Didn't find your answer?

Ask the community — no account required