I need to infer an 8 bit accumulator (acc8) using Verilog on the Xilinx Webpack. The Library guide seems to contain syntax errors. I could not get the tool to infer a loadable accumulator, no matter how I play around with the implementation. I get an adder using 10 slices, instead of 5 slices I should get when an accumulator is inferred. Does anybody know the solution?
You have to be a little careful because the carry logic in the slice follows the LUT, so in order to get in one level of logic you need to visualize the load preceeding the add. To do that, one input of the adder has an and gate so that it is forced to zero when load is active, the other input is a mux to select your load value or the addend (could be the same, depending on your design). Note in that case that there is logic in front of both add inputs. That same logic needs to preceed the carry mux DI input, which does have an AND gate available (the mult and). In order to use that, your load signal has to be active low. Some synthesis tools will infer the right structure as long as it is realizable in the hardware, while others need you to be more explicit.
Y K wrote:
--
--Ray Andraka, P.E. President, the Andraka Consulting Group, Inc.
401/884-7930 Fax 401/884-7950 email snipped-for-privacy@andraka.com"They that give up essential liberty to obtain a little temporary safety deserve neither liberty nor safety." -Benjamin Franklin, 1759
This pointed me at a workaround: The problem is indeed around the load input. The following code shows three versions: The Xilinx "Verilog" version (self explanatory), the same written in Verilog, and one where reset and load are or'ed together. Only the third one works. I prefer a pure Verilog solution for future portability reasons. The code is going to a product line that may live for 20 years, and I don't want to rewrite it everytime I have to design a board using a new FPGA family. The workaround is good enough for me, but perhaps Xilinx can fix the documentation and XST code too?
Thank you Yishai Kagan yk_four_zeroes snipped-for-privacy@hotmail.com (convert to digits to get the correct e-mail).
Here it is: module accumulator8(input C,R,L,D,CE,ADD, input [7:0] B, output reg [7:0] Q); /* The following does not work for obvious reasons. // Verilog Inference Code copied from the Libraries Guide Page 163 always @ (posedge C) begin if (R) Q realizable in the hardware, while others need you to be more explicit.
This pointed me at a workaround: The problem is indeed around the load input. The following code shows three versions: The Xilinx "Verilog" version (self explanatory), the same written in Verilog, and one where reset and load are or'ed together. Only the third one works. I prefer a pure Verilog solution for future portability reasons. The code is going to a product line that may live for 20 years, and I don't want to rewrite it everytime I have to design a board using a new FPGA family. The workaround is good enough for me, but perhaps Xilinx can fix the documentation and XST code too?
Thank you Yishai Kagan yk_four_zeroes snipped-for-privacy@hotmail.com (convert to digits to get the correct e-mail).
Here it is: module accumulator8(input C,R,L,D,CE,ADD, input [7:0] B, output reg [7:0] Q); /* The following does not work for obvious reasons. // Verilog Inference Code copied from the Libraries Guide Page 163 always @ (posedge C) begin if (R) Q realizable in the hardware, while others need you to be more explicit.
This pointed me at a workaround: The problem is indeed around the load input. The following code shows three versions: The Xilinx "Verilog" version (self explanatory), the same written in Verilog, and one where reset and load are or'ed together. Only the third one works. I prefer a pure Verilog solution for future portability reasons. The code is going to a product line that may live for 20 years, and I don't want to rewrite it everytime I have to design a board using a new FPGA family. The workaround is good enough for me, but perhaps Xilinx can fix the documentation and XST code too?
Thank you Yishai Kagan yk_four_zeroes snipped-for-privacy@hotmail.com (convert to digits to get the correct e-mail).
Here it is: module accumulator8(input C,R,L,D,CE,ADD, input [7:0] B, output reg [7:0] Q); /* The following does not work for obvious reasons. // Verilog Inference Code copied from the Libraries Guide Page 163 always @ (posedge C) begin if (R) Q realizable in the hardware, while others need you to be more explicit.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.