So I have an array of 4 RAM modules that I want to be able to read/write to based on two different selector signals. Right now I'm instantiating the RAM using intermediary signals:
genvar i;
generate
for (i = 0; i < regnum; i=i+1) begin: regs
rfram_generic rf (clk,rst,ce_a_int[i],addr_a_int[i],do_a_int[i],
ce_b_int[i],addr_b_int[i],do_b_int[i],
ce_w_int[i],we_w_int[i],addr_w_int[i],
di_w_int[i]);
end
endgenerate
And I want to select the RAM to use using either head or tail signals (2-bit vectors). Any ideas how to do this?
I'm new here and can't comment on questions yet, but in response to Marty: most FPGA synthesis tools will translate internal multisource signals with tristate values into MUX-like logic, where possible. See, for instance: a description of old tristate-to-logic behaviour that sounds accurate to me.
As a recommendation to Adam, you're probably better off making this explicit in your code by performing the masking yourself rather than using the tristates. This will improve portability, give you more predictable results, and serve as self-documentation if anybody ever has to revisit your code.
However, making some guesses based on your solution, it would probably make sense to simply mask off the clock enable on the write port and mux the output of the read port. For instance:
reg [WIDTH-1:0] do_a,do_b;
always @(*) do_a = do_a_int[head];
always @(*) do_b = do_b_int[tail];
generate
genvar i;
for (i = 0; i < regnum; i=i+1) begin: regs
rfram_generic rf (clk,rst,
ce_a,addr_a,do_a_int[i],
ce_b,addr_b,do_b_int[i],
ce_w,head==i?we_w:1'b0,addr_w,di_w);
end
endgenerate
This will probably result in less complex logic (i.e., better area and delay) than your solution.