From d95d0252e0b5e71e9ed8710e04dd5d56ee405bdc Mon Sep 17 00:00:00 2001
From: Wade Fife <wade.fife@ettus.com>
Date: Fri, 24 May 2024 14:17:13 -0500
Subject: [PATCH] fpga: lib: Add axis_combine

Original-commit: ab19d99cb0b4b74327bc31ecb5bc27ffcf94dec0
---
 lib/axi/Makefile.srcs  |   1 +
 lib/axi/axis_combine.v | 116 +++++++++++++++++++++++++++++++++++++++++
 2 files changed, 117 insertions(+)
 create mode 100644 lib/axi/axis_combine.v

diff --git a/lib/axi/Makefile.srcs b/lib/axi/Makefile.srcs
index 97eb6e4..6ab092c 100644
--- a/lib/axi/Makefile.srcs
+++ b/lib/axi/Makefile.srcs
@@ -33,6 +33,7 @@ axis_upsizer.v \
 axis_downsizer.v \
 axis_width_conv.v \
 axis_split.v \
+axis_combine.v \
 axis_packetize.v \
 axis_pkt_throttle.sv \
 axis_data_if_packetize.v \
diff --git a/lib/axi/axis_combine.v b/lib/axi/axis_combine.v
new file mode 100644
index 0000000..89c5fcd
--- /dev/null
+++ b/lib/axi/axis_combine.v
@@ -0,0 +1,116 @@
+//
+// Copyright 2023 Ettus Research, a National Instruments Brand
+//
+// SPDX-License-Identifier: LGPL-3.0-or-later
+//
+// Module: axis_combine
+//
+// Description:
+//
+//   Combines AXI stream buses into one output bus with concatenated data signals.
+//
+// Parameters:
+//
+//   SIZE           : The number of inputs streams to combine.
+//   WIDTH          : The width of TDATA on the input streams, if they are all the
+//                    same width. If they are different widths, then use WIDTH_VEC
+//                    instead.
+//   WIDTH_VEC      : A vector of widths corresponding to each stream's TDATA width.
+//                    Each number in this vector must be 32 bits wide. This defaults
+//                    to WIDTH bits for all inputs.
+//   USER_WIDTH     : The width of TUSER on the input streams, if they are all the
+//                    same width. If they are different widths, then use USER_WIDTH_VEC
+//                    instead.
+//   USER_WIDTH_VEC : A vector of widths corresponding to each stream's TUSER width.
+//                    Each number in this vector must be 32 bits wide. This defaults
+//                    to USER_WIDTH bits for all inputs.
+//   FIFO_SIZE_LOG2 : Log2 the size of the FIFO to use internally for each stream.
+//
+
+`default_nettype none
+
+module axis_combine #(
+  parameter               SIZE           = 2,
+  parameter               WIDTH          = 32,
+  parameter               USER_WIDTH     = 1,
+  parameter [32*SIZE-1:0] WIDTH_VEC      = {SIZE{WIDTH[31:0]}},
+  parameter [32*SIZE-1:0] USER_WIDTH_VEC = {SIZE{USER_WIDTH[31:0]}},
+  parameter               FIFO_SIZE_LOG2 = 0
+) (
+  input  wire clk,
+  input  wire reset,
+
+  // Input streams
+  input  wire [len(SIZE-1,     WIDTH_VEC)-1:0] s_axis_tdata,
+  input  wire [len(SIZE-1,USER_WIDTH_VEC)-1:0] s_axis_tuser,
+  input  wire [                      SIZE-1:0] s_axis_tlast,
+  input  wire [                      SIZE-1:0] s_axis_tvalid,
+  output wire [                      SIZE-1:0] s_axis_tready,
+
+  // Output streams
+  output wire [len(SIZE-1,     WIDTH_VEC)-1:0] m_axis_tdata,
+  output wire [len(SIZE-1,USER_WIDTH_VEC)-1:0] m_axis_tuser,
+  output wire                                  m_axis_tlast,
+  output wire                                  m_axis_tvalid,
+  input  wire                                  m_axis_tready
+);
+
+  // Helper function to calculate the combined length of the lower 'n' ports
+  // based on the concatenated widths (each 32-bits wide) stored in width_vector.
+  // Note: If n is negative, returns 0.
+  function automatic integer len(input integer n, input [32*SIZE-1:0] width_vector);
+    integer i, total;
+  begin
+    total = 0;
+    if (n >= 0) begin
+      for (i = 0; i <= n; i = i + 1) begin
+        total = total + width_vector[32*i +: 32];
+      end
+    end
+    len = total;
+  end
+  endfunction
+
+  wire [len(SIZE-1,     WIDTH_VEC)-1:0] int_tdata;
+  wire [len(SIZE-1,USER_WIDTH_VEC)-1:0] int_tuser;
+  wire [                      SIZE-1:0] int_tlast;
+  wire [                      SIZE-1:0] int_tvalid;
+  wire [                      SIZE-1:0] int_tready;
+
+  // Generate a FIFO for each stream
+  genvar i;
+  generate
+    for (i = 0; i < SIZE; i = i + 1) begin
+      axi_fifo #(
+        .WIDTH (WIDTH_VEC[32*i +: 32] + USER_WIDTH_VEC[32*i +: 32] + 1),
+        .SIZE  (FIFO_SIZE_LOG2)
+      ) axi_fifo (
+        .clk      (clk),
+        .reset    (reset),
+        .clear    (),
+        .i_tdata  ({ s_axis_tlast[i],
+                     s_axis_tuser[len(i,USER_WIDTH_VEC)-1 : len(i-1,USER_WIDTH_VEC)],
+                     s_axis_tdata[len(i,     WIDTH_VEC)-1 : len(i-1,     WIDTH_VEC)] }),
+        .i_tvalid (s_axis_tvalid[i]),
+        .i_tready (s_axis_tready[i]),
+        .o_tdata  ({ int_tlast[i],
+                     int_tuser[len(i,USER_WIDTH_VEC)-1 : len(i-1,USER_WIDTH_VEC)],
+                     int_tdata[len(i,     WIDTH_VEC)-1 : len(i-1,     WIDTH_VEC)] }),
+        .o_tvalid (int_tvalid[i]),
+        .o_tready (int_tready[i]),
+        .space    (),
+        .occupied ()
+      );
+    end
+  endgenerate
+
+  assign int_tready      = { SIZE {m_axis_tvalid & m_axis_tready} };
+
+  assign m_axis_tvalid   = &int_tvalid;
+  assign m_axis_tdata    = int_tdata;
+  assign m_axis_tuser    = int_tuser;
+  assign m_axis_tlast    = int_tlast;
+
+endmodule
+
+`default_nettype wire