help and advice for verilog code

[Adnan86]

Hi
I write this code for adding 4 number but with prefix sum. It means, we need all sum : sum0 , sum1 . ....

module prefix_sum(
input [3:0] data_in,
input clk, rst, ld,
output reg [3:0] sum [0:3]
    );
reg [3:0] temp = 4'b0000;
reg [1:0] num  = 2'b0;

always @ (posedge clk) begin
	if (rst) begin
		sum[0] <= 4'b0000;
		sum[1] <= 4'b0000;
		sum[2] <= 4'b0000;
		sum[3] <= 4'b0000;
	end else if (!ld) begin
		num <= 2'b0;
	end else begin
		temp <= temp + data_in;
		sum[num] <= temp;
		num <= num + 1;
	end
end

endmodule

but the code dont work.
I will appreciate for any advice.
thanks a lot

 

[vGoodtimes]

what is the issue? My guess will be temp never being reset, but that might be intentional.

 

[slutarius]

My guess would be:
You can not get the sum at the same time with the data_in changing and ld==1 ?
If you want the 3 assignment work in serial, you can make the temp as blocking assignment.

But, please post your problem on.

 

[Adnan86]

when i checked syntax, i just have warning that : Port sum must not be declared to be an array
but whenever want write testbench and check behavioral syntax the code dont work at all. and give me this warning :
Port sum must not be declared to be an array
and
PAD symbol "clk" is not constrained (LOC) to a specific
location.
and
PAD symbol "clk" has an undefined IOSTANDARD.
I have no idea what it is.

- - - Updated - - -

so i cant see simulation to know how it works

- - - Updated - - -

also for num <= num +1 have this warning : Result of 3-bit expression is truncated to fit in 2-bit target

 

[dpaul]

when i checked syntax, i just have warning that : Port sum must not be declared to be an array
but whenever want write testbench and check behavioral syntax the code dont work at all.

Are you are trying to store incoming data into a 4x4 array in the order they are received as per the state of "ld"?
It would also be nice to indicate your intention of your RTL code.

also for num <= num +1 have this warning : Result of 3-bit expression is truncated to fit in 2-bit target

Yes, the compiler will not allow you to represent dec(4) using a 2 bits!

PAD symbol "clk" is not constrained (LOC) to a specific
location.
and
PAD symbol "clk" has an undefined IOSTANDARD.
I have no idea what it is.

These two you can deal with later. First get your simulation working.

btw - If you have *just warnings* it would still be possible to proceed to simulation (although you are not likely to see the desired results). Is your TB driving all the necessary i/p signals?

 

[Adnan86]

Are you are trying to store incoming data into a 4x4 array in the order they are received as per the state of "ld"?
It would also be nice to indicate your intention of your RTL code.


Yes, the compiler will not allow you to represent dec(4) using a 2 bits!


These two you can deal with later. First get your simulation working.

btw - If you have *just warnings* it would still be possible to proceed to simulation (although you are not likely to see the desired results). Is your TB driving all the necessary i/p signals?

Yes, I want to store results in 4 number with 4 bits in Sum[num].
Actually i receive 4 data_in and in every clk sum them together and store all results (prefix sum) .

for num <= num + 1; that you said :Yes, the compiler will not allow you to represent dec(4) using a 2 bits!
do you have any advice for this.

and the problem is that i dont know why with these warning i cant get simulation results.

 

[dpaul]

You need to increase the reg size of your sum in that case. As I understand from #1, your num size should be equal to your array depth which has 4 bits, i.e depth=16.
So in your RTL if have to add to your code before incr num, something like,
if (num< 4'b1111)
increment num;
else
reset num;

When you are writing RTL you should think in parallel what h/w shall your code generate, rather than just obtaining the target functionality. Then you'll minimize such compiler warning messages.

 

[Adnan86]

You need to increase the reg size of your sum in that case. As I understand from #1, your num size should be equal to your array depth which has 4 bits, i.e depth=16.
So in your RTL if have to add to your code before incr num, something like,
if (num< 4'b1111)
increment num;
else
reset num;

When you are writing RTL you should think in parallel what h/w shall your code generate, rather than just obtaining the target functionality. Then you'll minimize such compiler warning messages.

no i have just 4 state. so 2 bits for num is good.
i want to get 4 number in data_in
and sum them in every clock and store result after any clock.

 

[ThisIsNotSam]

check the logic for num, it is not being reset. don't initialise regs like that, it's a bad practice.

 

[ads-ee]

check the logic for num, it is not being reset. don't initialise regs like that, it's a bad practice.

This is in the FPGA section, both Altera and Xilinx guarantee the power up state of a FF initialized on the declaration line. the issue is without a reset you can not reliably guarantee the state of the FF when you reset the other FFs. My take on this is: datapath - no resets, control logic - resets. num is used as an index into the array so is part of the control logic.

when i checked syntax, i just have warning that : Port sum must not be declared to be an array
but whenever want write testbench and check behavioral syntax the code dont work at all. and give me this warning :
Port sum must not be declared to be an array

Standard Verilog doesn't support passing arrays through module ports. You will have to set your simulator and synthesis tools to use System Verilog. As of 2015.2 Xilinx simulator still does not accept arrays in module ports regardless of the SV switch (don't know if it is supported in 2016.2).

PAD symbol "clk" is not constrained (LOC) to a specific
location.
and
PAD symbol "clk" has an undefined IOSTANDARD.
I have no idea what it is so i cant see simulation to know how it works.

These are calling for constraints on the ports in your xdc file. If you don't have constraints on all your pins you will see these types of warnings. But why are you running anything else besides simulation, you should not be trying to implement the design at this time.

also for num <= num +1 have this warning : Result of 3-bit expression is truncated to fit in 2-bit target

maximum value of reg [1:0] num; is 2'b11, if you add 1 to this you end up with 3'b100, which according to Verilog assignment rules gets truncated to 2'b00.

- - - Updated - - -

The worst problem is the lack of any kind of initialization on temp. Once temp is used the first time after power-up the state of temp can never be returned to 0 therefore the design can never recompute a new sum after it's done it once (well technically you could add 4 values that zero out temp).

I get the impression that the design isn't done using a hardware paradigm as it seems to lack any approach to how it works on repeated sets of new inputs. It's more like a console level software program that you launch and it runs once and returns control to the console window.

 

[ThisIsNotSam]

Still, it is bad practice, as evidenced by OP's way of 'describing hardware'. Just add a reset.

 

[ads-ee]

I'm also having a difficult time determining what you are trying to accomplish, your first post and subsequent explanations don't clear this up.

Are you trying to set the sum[0], sum[1], sum[2], & sum[3] values with intermediate summations of data_in.

e.g.

data_in => 1, 3, 2, and 5

sum[0] = 1
sum[1] = 4 (1+3)
sum[2] = 6 (1+3+2)
sum[3] = 11 (1+3+2+6)

or are you trying to do something entirely different than that?

- - - Updated - - -

Still, it is bad practice, as evidenced by OP's way of 'describing hardware'. Just add a reset.

Well in the case of num

	end else if (!ld) begin
		num <= 2'b0;

it can be set to 0 when ld == 0, though this seems kind of backwards as I would assume ld stands for LOAD. Which seems to mean that when you are loading you do the storing...hmm, maybe that is the intention to initialize the array with intermediate summations? If this is the case the temp should have been added to this same section of code so that it is initialized to 0 before adding four data_in values together.

 

[dpaul]

For the above obvious reasons I had stopped commenting after #8.
I think the OP needs time to soak-in all these & hopefully, he'll come back tomorrow with a better understanding and better explanation of what he wants to do.

 

[Adnan86]

I'm also having a difficult time determining what you are trying to accomplish, your first post and subsequent explanations don't clear this up.

Are you trying to set the sum[0], sum[1], sum[2], & sum[3] values with intermediate summations of data_in.

e.g.

data_in => 1, 3, 2, and 5

sum[0] = 1
sum[1] = 4 (1+3)
sum[2] = 6 (1+3+2)
sum[3] = 11 (1+3+2+6)

or are you trying to do something entirely different than that?

yes , as you said my point is :

data_in => 1, 3, 2, and 5

sum[0] = 1
sum[1] = 4 (1+3)
sum[2] = 6 (1+3+2)
sum[3] = 11 (1+3+2+6)

- - - Updated - - -



Well in the case of num

	end else if (!ld) begin
		num <= 2'b0;

it can be set to 0 when ld == 0, though this seems kind of backwards as I would assume ld stands for LOAD. Which seems to mean that when you are loading you do the storing...hmm, maybe that is the intention to initialize the array with intermediate summations? If this is the case the temp should have been added to this same section of code so that it is initialized to 0 before adding four data_in values together.

yes my point is :

data_in => 1, 3, 2, and 5

sum[0] = 1
sum[1] = 4 (1+3)
sum[2] = 6 (1+3+2)
sum[3] = 11 (1+3+2+6)

about num, i reset it with ld.

 

[ads-ee]

yes my point is :

data_in => 1, 3, 2, and 5

sum[0] = 1
sum[1] = 4 (1+3)
sum[2] = 6 (1+3+2)
sum[3] = 11 (1+3+2+6)

I just guessed that this was what you were trying to do.

Let this be a lesson to you and others, give an example of what you expect your code to do, instead of a rambling explanation that doesn't translate well and a functionality that can't be determined from "broken" code.

I still don't quite know what it is that you think is wrong with the simulation or if it is even a simulation problem you are having...no testbench and implementation warnings...probably means you are not running simulation on the design?

 

[Adnan86]

I just guessed that this was what you were trying to do.

Let this be a lesson to you and others, give an example of what you expect your code to do, instead of a rambling explanation that doesn't translate well and a functionality that can't be determined from "broken" code.

I still don't quite know what it is that you think is wrong with the simulation or if it is even a simulation problem you are having...no testbench and implementation warnings...probably means you are not running simulation on the design?

yes you right, i thought its enough to say just add 4 number with prefix sum. its my mistake, sorry about that.
BTW
This is main code with little change :

module prefix_sum(
input [3:0] data_in,
input clk, rst, ld,
output reg [3:0] sum [0:3]
    );
reg [3:0] temp ;
reg [1:0] num ;

always @ (posedge clk) begin
	if (rst) begin
		sum[0] <= 4'b0000;
		sum[1] <= 4'b0000;
		sum[2] <= 4'b0000;
		sum[3] <= 4'b0000;
		temp <= 4'b0000;
		num  <= 2'b00;
	end else if (!ld) begin
		num <= 2'b00;
	end else if (num <= 2'b11) begin
		temp <= temp + data_in;
		sum[num] <= temp;
		num <= num + 1;
		
	end
end


endmodule

and this is test :

module test_sum;

	// Inputs
	reg [3:0] data_in;
	reg clk;
	reg rst;
	reg ld;

	// Instantiate the Unit Under Test (UUT)
	prefix_sum uut (
		.data_in(data_in), 
		.clk(clk), 
		.rst(rst), 
		.ld(ld)
	);
	initial repeat (10) clk = ~clk;
	initial begin
		// Initialize Inputs
		data_in = 0;
		clk = 0;
		rst = 1;
		ld = 0;
		#15 ld = 1; rst = 0; data_in = 4'b0010;
		#10  data_in = 4'b0011;
		#10  data_in = 4'b0001;
		#10  data_in = 4'b0100;
		//#20 $stop;
		

		// Wait 100 ns for global reset to finish
		#100;
        
		// Add stimulus here

	end
      
endmodule

And I use ISE 14.7 version.
I have any error in main code just have warning that i said it before, but whenever i run test i have these errors :
Module <prefix_sum> ignored due to previous errors.
and
Port sum must not be declared to be an arrayVerilog
and I cant simulate it.
just this, I hope you get my points

 

[ads-ee]

You can't use arrays in ports at all in ISE 14.7 it has virtually zero support for any System Verilog syntax.

If you want to pass an array to a module you will have to flatten it.

---

Code Verilog - [expand]
1
output reg [4*4-1:0]  sum,

---

and access it like

---

Code Verilog - [expand]
1
sum[4*num +:4]  // the 4*num groups by 4 bits and +:4 means that the LSB is 4*num and the number of bits is 4 so 3:0

---

You could also reassign the flattened port to an array with a for loop too.