fpga: lib: rfnoc: Make RFNoC packet gates removable

This adds a parameter to chdr_xb_ingress_buff and to chdr_stream_output
to make their packet gates optional. They are included by default.

Having the packet gate in chdr_xb_ingress_buff can reduce contention in
the crossbar. Having the packet gate in chdr_stream_output can reduce
the latency for non-data packets (e.g., stream status updates).


Original-commit: 67e0a5a340c14cb0c6ee170fb17be2c09737d723

lib/rfnoc/core/chdr_stream_output.v

@@ -17,6 +17,7 @@
 // Parameters:
 //   - CHDR_W      : Width of the CHDR bus in bits
 //   - MTU         : Log2 of the maximum number of lines in a packet
+//   - EN_PKT_GATE : Enable packet gate to make each packet contiguous
 //
 // Signals:
 //   - m_axis_chdr_* : Output CHDR stream (AXI-Stream)
@@ -25,7 +26,8 @@
 
 module chdr_stream_output #(
   parameter CHDR_W      = 256,
-  parameter MTU    = 10
+  parameter MTU         = 10,
+  parameter EN_PKT_GATE = 1
 )(
   // Clock, reset and settings
   input  wire              clk,
@@ -84,15 +86,31 @@ module chdr_stream_output #(
   reg               chdr_out_tlast, chdr_out_tvalid;
   wire              chdr_out_tready;
 
+  if (EN_PKT_GATE) begin : gen_pkt_gate
     axi_packet_gate #(
-    .WIDTH(CHDR_W), .SIZE(MTU), .USE_AS_BUFF(0)
+      .WIDTH      (CHDR_W),
+      .SIZE       (MTU   ),
+      .USE_AS_BUFF(0     )
     ) chdr_pkt_gate_i (
-    .clk(clk), .reset(rst), .clear(1'b0),
+      .clk     (clk               ),
-    .i_tdata(chdr_out_tdata), .i_tlast(chdr_out_tlast), .i_terror(1'b0),
+      .reset   (rst               ),
-    .i_tvalid(chdr_out_tvalid), .i_tready(chdr_out_tready),
+      .clear   (1'b0              ),
-    .o_tdata(m_axis_chdr_tdata), .o_tlast(m_axis_chdr_tlast),
+      .i_tdata (chdr_out_tdata    ),
-    .o_tvalid(m_axis_chdr_tvalid), .o_tready(m_axis_chdr_tready)
+      .i_tlast (chdr_out_tlast    ),
+      .i_terror(1'b0              ),
+      .i_tvalid(chdr_out_tvalid   ),
+      .i_tready(chdr_out_tready   ),
+      .o_tdata (m_axis_chdr_tdata ),
+      .o_tlast (m_axis_chdr_tlast ),
+      .o_tvalid(m_axis_chdr_tvalid),
+      .o_tready(m_axis_chdr_tready)
     );
+  end else begin : gen_no_pkt_gate
+    assign m_axis_chdr_tdata  = chdr_out_tdata;
+    assign m_axis_chdr_tlast  = chdr_out_tlast;
+    assign m_axis_chdr_tvalid = chdr_out_tvalid;
+    assign chdr_out_tready    = m_axis_chdr_tready;
+  end
 
   // ---------------------------------------------------
   //  Flow Control State

lib/rfnoc/crossbar/chdr_xb_ingress_buff.v

@@ -18,13 +18,16 @@
 //   MTU         : Maximum transmission unit, in WIDTH-sized words, is 2**MTU
 //   DEST_W      : Width of the destination routing information (TDEST)
 //   NODE_ID     : Numeric identifier for this port
+//   EN_PKT_GATE : Enable packet gate to make each packet contiguous. This
+//                 reduces congestion in the crossbar.
 //
 
 module chdr_xb_ingress_buff #(
   parameter       WIDTH       = 64,
   parameter       MTU         = 10,
   parameter       DEST_W      = 4,
-  parameter [9:0] NODE_ID = 0
+  parameter [9:0] NODE_ID     = 0,
+  parameter       EN_PKT_GATE = 1
 ) (
   input  wire               clk,
   input  wire               reset,
@@ -69,6 +72,7 @@ module chdr_xb_ingress_buff #(
 
   // The axi_packet_gate queues up an entire packet before letting it go out.
   // This reduces congestion in the crossbar for slowly-built packets.
+  if (EN_PKT_GATE) begin : gen_pkt_gate
     axi_packet_gate #(
       .WIDTH (WIDTH),
       .SIZE  (MTU)
@@ -86,6 +90,23 @@ module chdr_xb_ingress_buff #(
       .o_tvalid (gate_o_tvalid),
       .o_tready (gate_o_tready)
     );
+  end else begin : gen_no_pkt_gate
+    axi_fifo_flop2 #(
+      .WIDTH(WIDTH+1)
+    ) axi_fifo_flop2_i (
+      .clk     (clk                         ),
+      .reset   (reset                       ),
+      .clear   (1'b0                        ),
+      .i_tdata ({gate_i_tlast, gate_i_tdata}),
+      .i_tvalid(gate_i_tvalid               ),
+      .i_tready(gate_i_tready               ),
+      .o_tdata ({gate_o_tlast, gate_o_tdata}),
+      .o_tvalid(gate_o_tvalid               ),
+      .o_tready(gate_o_tready               ),
+      .space   (                            ),
+      .occupied(                            )
+    );
+  end
 
 
   //---------------------------------------------------------------------------