Hi,
In Cliff Cummings's paper "Simulation and Synthesis Techniques for Asynchronous FIFO Design" (at the web page: www.sunburst-design.com/papers), there is clear description on FULL/EMPTY design. As to almost_full/almost_empty, the author indicates to use
the second adder after Gray-to-binary logic and use this value to substract rptr to get an almost_full signal. A less-than operation insures the almost_full is set for the full range.

I would like to generate almost_full/almost_empty with programmable settings, so I updated the design as these:

// -----------------------------------------------
// GRAYSTYLE2 pointer
always @(posedge wclk or negedge wrst_n) begin
if (!wrst_n)
{wbin, wptr} <= 0;
else
{wbin, wptr} <= {wbinnext, wgraynext};
end

always @(posedge wclk or negedge wrst_n) begin // this is added
if (!wrst_n)
wbin_p1 <= 0;
else
wbin_p1 <= wgraynextp1;
end

// Memory write-address pointer (okay to use binary to address memory)
assign waddr = wbin[ADDRSIZE-1:0];
assign wbinnext = wbin + (winc & ~wfull);
assign wgraynext = (wbinnext >> 1) ^ wbinnext;
// this is added to generate the second adder
assign wgraynextp1 = ((wbinnext + prog_diff) >> 1) ^ (wbinnext + prog_diff);

assign wfull_val = (wgraynext == {~wq2_rptr[ADDRSIZE:ADDRSIZE-1],wq2_rptr[ADDRSIZE-2:0]});
// this is added to generate almost_full
assign awfull_val = (wbin_p1 - wq2_rptr) <= prog_diff;

In my simulation I set prog_diff to 4. The fifo depth is set to 32. The simulation doesn't generate correct almost_full signal. Even in the start of the simulation, when fifo is start to write, almost_full is asserted.

It seems I can't generate the signal simply as what author indicated. Can anybody give me some suggestion? Thanks very much.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On Sunday, June 21, 2020 at 6:47:33 AM UTC-6, chenyo...@gmail.com wrote:

Hi,
In Cliff Cummings's paper "Simulation and Synthesis Techniques for Asynchronous FIFO Design" (at the web page: www.sunburst-design.com/papers), there is clear description on FULL/EMPTY design. As to almost_full/almost_empty, the author indicates to use

the second adder after Gray-to-binary logic and use this value to substract rptr to get an almost_full signal. A less-than operation insures the almost_full is set for the full range.

I would like to generate almost_full/almost_empty with programmable settings, so I updated the design as these:

// -----------------------------------------------
// GRAYSTYLE2 pointer
always @(posedge wclk or negedge wrst_n) begin
if (!wrst_n)
{wbin, wptr} <= 0;
else
{wbin, wptr} <= {wbinnext, wgraynext};
end

always @(posedge wclk or negedge wrst_n) begin // this is added
if (!wrst_n)
wbin_p1 <= 0;
else
wbin_p1 <= wgraynextp1;
end

// Memory write-address pointer (okay to use binary to address memory)
assign waddr = wbin[ADDRSIZE-1:0];
assign wbinnext = wbin + (winc & ~wfull);
assign wgraynext = (wbinnext >> 1) ^ wbinnext;
// this is added to generate the second adder
assign wgraynextp1 = ((wbinnext + prog_diff) >> 1) ^ (wbinnext + prog_diff);

assign wfull_val = (wgraynext == {~wq2_rptr[ADDRSIZE:ADDRSIZE-1],wq2_rptr[ADDRSIZE-2:0]});
// this is added to generate almost_full
assign awfull_val = (wbin_p1 - wq2_rptr) <= prog_diff;

In my simulation I set prog_diff to 4. The fifo depth is set to 32. The simulation doesn't generate correct almost_full signal. Even in the start of the simulation, when fifo is start to write, almost_full is asserted.

It seems I can't generate the signal simply as what author indicated. Can anybody give me some suggestion? Thanks very much.

Some code seems to be missing: what is wq2_rptr? Making an asynchronous FIFO is a pain; I'd try to get some already-verified code. For many cases you might also be able to instantiate a FIFO primitive and use the model for simulation. Xilinx has
these built in.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)