Why I'm keep getting error in the output signal? - verilog

Question

I write a simple Verilog code for calculating inner product between 2 vectors. I don't understand why, but I get an error in the module's output - red signal. There is probably some problem with the result variable.
Does anyone know what I'm missing?
The code -
`resetall
`timescale 1ns/10ps
`include "../hdl/params.v"
module ZCalculationParllel(features_vec, weights_vec, bias, predict);
// Feature vector in size of |NUMBER_OF_PIXELS| * |PIXEL_PRECISION|
input [(`NUMBER_OF_PIXELS * `PIXEL_PRECISION) - 1 : 0] features_vec;
// Weights vector in size of |NUMBER_OF_PIXELS| * |WEIGHT_BIAS_PRECISION|
input [(`NUMBER_OF_PIXELS * `WEIGHT_BIAS_PRECISION) - 1 : 0] weights_vec;
// Bias vector in size |WEIGHT_BIAS_PRECISION|
input [`WEIGHT_BIAS_PRECISION - 1 : 0] bias;
// The output value for prediction
//output predict; in case we want return one bit represent prediction
output [(2 * `PIXEL_PRECISION + $bits(`NUMBER_OF_PIXELS)) - 1 : 0] predict;
// Accomulator array for saving the multiplication result before the adders
wire [`PIXEL_PRECISION_PLUS_WEIGHT_BIAS_PRECISION - 1 : 0] multiplications [0 : `NUMBER_OF_PIXELS - 1];
wire [(2 * `PIXEL_PRECISION + $bits(`NUMBER_OF_PIXELS)) - 1 : 0] result;
genvar k;
generate
for (k = 0; k < `NUMBER_OF_PIXELS; k = k + 1)
begin: elementMul
assign multiplications[k] = features_vec[(k + 1) * `PIXEL_PRECISION - 1 -: `PIXEL_PRECISION] * weights_vec[(k + 1) * `WEIGHT_BIAS_PRECISION - 1 -: `WEIGHT_BIAS_PRECISION];
end
endgenerate
assign result = 2;
genvar i;
generate
for (i = 0; i < `NUMBER_OF_PIXELS; i = i + 1)
begin: elementAdd
assign result = result + multiplications[i];
end
endgenerate
assign predict = result;
endmodule
The test -
`resetall
`timescale 1ns/10ps
`include "../hdl/ZCalculationParllel.v"
`include "../hdl/params.v"
module ZCalculationParllel_tb ;
reg [(`NUMBER_OF_PIXELS * `PIXEL_PRECISION) - 1 : 0] features;
reg [(`NUMBER_OF_PIXELS * `WEIGHT_BIAS_PRECISION) - 1 : 0] weights;
reg [`WEIGHT_BIAS_PRECISION - 1 : 0] b;
wire [(2 * `PIXEL_PRECISION + $bits(`NUMBER_OF_PIXELS)) - 1 : 0] ans;
ZCalculationParllel z (
.features_vec(features),
.weights_vec(weights),
.bias(b),
.predict(ans)
);
initial
begin
b = 8'b00000010;
features[7:0] = 8'b00000010;
features[15:8] = 8'b00000001;
features[24:16] = 8'b00000010;
weights[7:0] = 8'b00000010;
weights[15:8] = 8'b00000001;
weights[24:16] = 8'b00000010;
end
endmodule

Answer 1

assign are applied concurrency, not sequentially. Xs will result if any of the assignments are conflicting.
Replace your result related code with something like the following:
reg [(2 * `PIXEL_PRECISION + $bits(`NUMBER_OF_PIXELS)) - 1 : 0] result;
integer i;
always #* begin
result = 2;
for (i = 0; i < `NUMBER_OF_PIXELS; i = i + 1) begin
result = result + multiplications[i];
end
end