-Stripped out the frame buffer and dp_offload_tx_len_calc;

-Added optional post_split FIFO.

libraries/base/dp/src/vhdl/dp_offload_tx.vhd

@@ -29,7 +29,9 @@
 --   interface is not used, the default settings passed via g_hdr_field_arr are used.
 -- . NOTE: dp_frame_rd DOES NOT SUPPORT 1 WORD FRAMES due to its separate s_sof and
 --   s_eof states.
-
+-- Remarks:
+-- . Note that an output (fill) FIFO may be needed if the packets have gaps in them!
+--   The Ethernet cores require this so an EOP-triggered FIFO may added to them.
 
 LIBRARY IEEE, common_lib, work, mm_lib;
 USE IEEE.STD_LOGIC_1164.ALL;
@@ -50,7 +52,8 @@ ENTITY dp_offload_tx IS
     g_def_nof_blocks_per_packet : NATURAL;
     g_output_fifo_depth         : NATURAL;
     g_hdr_field_arr             : t_common_field_arr;
-    g_hdr_field_ovr_init        : STD_LOGIC_VECTOR
+    g_hdr_field_ovr_init        : STD_LOGIC_VECTOR;
+    g_use_post_split_fifo       : BOOLEAN := FALSE
   ); 
   PORT (
     mm_rst               : IN  STD_LOGIC;
@@ -84,6 +87,8 @@ ARCHITECTURE str OF dp_offload_tx IS
   CONSTANT c_dp_split_val_latency        : NATURAL := 1;
   CONSTANT c_dp_packet_merge_val_latency : NATURAL := 2;
   CONSTANT c_nof_header_words            : NATURAL := field_slv_len(g_hdr_field_arr) / g_data_w;
+  CONSTANT c_dp_field_blk_snk_data_w : NATURAL := field_slv_out_len(field_arr_set_mode(g_hdr_field_arr , "RW"));
+  CONSTANT c_dp_field_blk_src_data_w : NATURAL := g_data_w;
 
   SIGNAL dp_split_snk_in_arr          : t_dp_sosi_arr(g_nof_streams-1 DOWNTO 0);
   SIGNAL dp_split_snk_out_arr         : t_dp_siso_arr(g_nof_streams-1 DOWNTO 0);
@@ -91,8 +96,11 @@ ARCHITECTURE str OF dp_offload_tx IS
   SIGNAL dp_split_src_out_2arr        : t_dp_sosi_2arr_2(g_nof_streams-1 DOWNTO 0);
   SIGNAL dp_split_src_in_2arr         : t_dp_siso_2arr_2(g_nof_streams-1 DOWNTO 0);
 
+  SIGNAL dp_packet_merge_snk_in_arr   : t_dp_sosi_arr(g_nof_streams-1 DOWNTO 0);
+  SIGNAL dp_packet_merge_snk_out_arr  : t_dp_siso_arr(g_nof_streams-1 DOWNTO 0);
+
   SIGNAL dp_packet_merge_src_out_arr  : t_dp_sosi_arr(g_nof_streams-1 DOWNTO 0);
-  SIGNAL dp_packet_merge_src_in_arr   : t_dp_siso_arr(g_nof_streams-1 DOWNTO 0) := (OTHERS=>c_dp_siso_rdy);
+  SIGNAL dp_packet_merge_src_in_arr   : t_dp_siso_arr(g_nof_streams-1 DOWNTO 0);
 
   SIGNAL dp_field_blk_snk_in_arr      : t_dp_sosi_arr(g_nof_streams-1 DOWNTO 0);
   SIGNAL dp_field_blk_snk_out_arr     : t_dp_siso_arr(g_nof_streams-1 DOWNTO 0);
@@ -103,15 +111,15 @@ ARCHITECTURE str OF dp_offload_tx IS
   SIGNAL hdr_fields_in_piped_arr      : t_slv_1024_arr(g_nof_streams-1 DOWNTO 0);
   SIGNAL hdr_fields_to_field_blk      : t_slv_1024_arr(g_nof_streams-1 DOWNTO 0);
 
-  TYPE t_dp_fifo_sc_usedw_arr IS ARRAY(g_nof_streams-1 DOWNTO 0) OF STD_LOGIC_VECTOR(ceil_log2(g_output_fifo_depth)-1 DOWNTO 0);
-  SIGNAL dp_fifo_sc_usedw_arr         : t_dp_fifo_sc_usedw_arr;
+--  TYPE t_dp_fifo_sc_usedw_arr IS ARRAY(g_nof_streams-1 DOWNTO 0) OF STD_LOGIC_VECTOR(ceil_log2(g_output_fifo_depth)-1 DOWNTO 0);
+--  SIGNAL dp_fifo_sc_usedw_arr         : t_dp_fifo_sc_usedw_arr;
 
-  SIGNAL dp_frame_rd_frm_req_arr      : STD_LOGIC_VECTOR(g_nof_streams-1 DOWNTO 0);
+--  SIGNAL dp_frame_rd_frm_req_arr      : STD_LOGIC_VECTOR(g_nof_streams-1 DOWNTO 0);
 
-  SIGNAL dp_frame_rd_snk_in_arr       : t_dp_sosi_arr(g_nof_streams-1 DOWNTO 0);
-  SIGNAL dp_frame_rd_snk_out_arr      : t_dp_siso_arr(g_nof_streams-1 DOWNTO 0);
+--  SIGNAL dp_frame_rd_snk_in_arr       : t_dp_sosi_arr(g_nof_streams-1 DOWNTO 0);
+--  SIGNAL dp_frame_rd_snk_out_arr      : t_dp_siso_arr(g_nof_streams-1 DOWNTO 0);
 
-  SIGNAL dp_frame_rd_src_out_arr      : t_dp_sosi_arr(g_nof_streams-1 DOWNTO 0);
+--  SIGNAL dp_frame_rd_src_out_arr      : t_dp_sosi_arr(g_nof_streams-1 DOWNTO 0);
 
   SIGNAL dp_concat_snk_in_2arr        : t_dp_sosi_2arr_2(g_nof_streams-1 DOWNTO 0);
   SIGNAL dp_concat_snk_out_2arr       : t_dp_siso_2arr_2(g_nof_streams-1 DOWNTO 0);
@@ -130,13 +138,13 @@ ARCHITECTURE str OF dp_offload_tx IS
   TYPE t_nof_words_per_block_arr IS ARRAY(g_nof_streams-1 DOWNTO 0) OF STD_LOGIC_VECTOR(ceil_log2(g_max_nof_words_per_block+1)-1 DOWNTO 0);
   SIGNAL nof_words_per_block_arr : t_nof_words_per_block_arr; 
 
-  SIGNAL nof_words_per_block_stat_arr : t_nof_words_per_block_arr; 
-  SIGNAL nof_words_per_block_stat_piped_arr : t_nof_words_per_block_arr; 
+--  SIGNAL nof_words_per_block_stat_arr : t_nof_words_per_block_arr; 
+--  SIGNAL nof_words_per_block_stat_piped_arr : t_nof_words_per_block_arr; 
 
   TYPE t_nof_blocks_per_packet_arr IS ARRAY(g_nof_streams-1 DOWNTO 0) OF STD_LOGIC_VECTOR(ceil_log2(g_max_nof_blocks_per_packet+1)-1 DOWNTO 0);
   SIGNAL nof_blocks_per_packet_arr : t_nof_blocks_per_packet_arr; 
 
-  SIGNAL nof_blocks_per_packet_stat_arr : t_nof_blocks_per_packet_arr; 
+--  SIGNAL nof_blocks_per_packet_stat_arr : t_nof_blocks_per_packet_arr; 
 
   SIGNAL reg_mosi_arr : t_mem_mosi_arr(g_nof_streams-1 DOWNTO 0);
   SIGNAL reg_miso_arr : t_mem_miso_arr(g_nof_streams-1 DOWNTO 0);
@@ -147,11 +155,11 @@ ARCHITECTURE str OF dp_offload_tx IS
   SIGNAL reg_hdr_ovr_mosi_arr : t_mem_mosi_arr(g_nof_streams-1 DOWNTO 0);
   SIGNAL reg_hdr_ovr_miso_arr : t_mem_miso_arr(g_nof_streams-1 DOWNTO 0); 
 
+  SIGNAL dp_pipeline_hdr_fields_snk_in_arr  : t_dp_sosi_arr(g_nof_streams-1 DOWNTO 0);
+  SIGNAL dp_pipeline_hdr_fields_snk_out_arr : t_dp_siso_arr(g_nof_streams-1 DOWNTO 0);
 
-
-  CONSTANT c_dp_field_blk_snk_data_w : NATURAL := field_slv_out_len(field_arr_set_mode(g_hdr_field_arr , "RW"));
-  CONSTANT c_dp_field_blk_src_data_w : NATURAL := g_data_w;
-
+  SIGNAL dp_pipeline_hdr_fields_src_out_arr : t_dp_sosi_arr(g_nof_streams-1 DOWNTO 0);
+  SIGNAL dp_pipeline_hdr_fields_src_in_arr  : t_dp_siso_arr(g_nof_streams-1 DOWNTO 0);
      
 BEGIN
 
@@ -189,7 +197,7 @@ BEGIN
       clk             => dp_clk,
 
       nof_symbols     => nof_words_per_block_arr(i),
-      nof_symbols_out => nof_words_per_block_stat_arr(i),
+--      nof_symbols_out => nof_words_per_block_stat_arr(i),
 
       snk_out         => dp_split_snk_out_arr(i),
       snk_in          => dp_split_snk_in_arr(i),
@@ -202,6 +210,33 @@ BEGIN
 
   END GENERATE;
 
+  ---------------------------------------------------------------------------------------
+  -- Use a post-split FIFO when the source has no flow control and its inter-block gap
+  -- is shorter than required for the header to be inserted.
+  ---------------------------------------------------------------------------------------
+  gen_dp_fifo_sc : FOR i IN 0 TO sel_a_b(g_use_post_split_fifo, g_nof_streams-1, -1) GENERATE
+    u_dp_fifo_sc : ENTITY work.dp_fifo_sc
+    GENERIC MAP (
+      g_data_w    => g_data_w,
+      g_fifo_size => g_max_nof_words_per_block 
+     )
+    PORT MAP (
+      rst        => dp_rst,
+      clk        => dp_clk,
+  
+      snk_in     => dp_split_src_out_2arr(i)(1),
+      snk_out    => dp_split_src_in_2arr(i)(1),
+  
+      src_out    => dp_packet_merge_snk_in_arr(i),
+      src_in     => dp_packet_merge_snk_out_arr(i)
+    );
+  END GENERATE;
+
+  no_dp_fifo_sc : FOR i IN 0 TO sel_a_b(g_use_post_split_fifo, -1, g_nof_streams-1) GENERATE
+    dp_packet_merge_snk_in_arr(i) <= dp_split_src_out_2arr(i)(1);
+    dp_split_src_in_2arr(i)(1)    <= dp_packet_merge_snk_out_arr(i);
+  END GENERATE;
+
   ---------------------------------------------------------------------------------------
   -- Merge nof_blocks_per_packet
   ---------------------------------------------------------------------------------------
@@ -215,14 +250,18 @@ BEGIN
       clk         => dp_clk,
   
       nof_pkt     => nof_blocks_per_packet_arr(i),
-      nof_pkt_out => nof_blocks_per_packet_stat_arr(i),
+--      nof_pkt_out => nof_blocks_per_packet_stat_arr(i),
  
-      snk_out     => dp_split_src_in_2arr(i)(1),
-      snk_in      => dp_split_src_out_2arr(i)(1),
+      snk_out     => dp_packet_merge_snk_out_arr(i),
+      snk_in      => dp_packet_merge_snk_in_arr(i),
   
       src_in      => dp_packet_merge_src_in_arr(i),
       src_out     => dp_packet_merge_src_out_arr(i)
     );
+
+    dp_concat_snk_in_2arr(i)(0) <= dp_packet_merge_src_out_arr(i);
+    dp_packet_merge_src_in_arr(i) <= dp_concat_snk_out_2arr(i)(0);
+
   END GENERATE;
 
   ---------------------------------------------------------------------------------------
@@ -242,168 +281,31 @@ BEGIN
   -- hdr_fields_in_piped_arr(i).............________X_________X_________X____
   --
   ---------------------------------------------------------------------------------------
-  gen_common_pipeline_hdr_fields_in_arr : FOR i IN 0 TO g_nof_streams-1 GENERATE
-    u_common_pipeline : ENTITY common_lib.common_pipeline
-    GENERIC MAP (
-      g_pipeline  => c_dp_split_val_latency + c_dp_packet_merge_val_latency,
-      g_in_dat_w  => field_slv_len(g_hdr_field_arr),
-      g_out_dat_w => field_slv_len(g_hdr_field_arr)
-    )
-    PORT MAP (
-      rst     => dp_rst,
-      clk     => dp_clk,
-      in_dat  => hdr_fields_in_arr(i)(field_slv_len(g_hdr_field_arr)-1 DOWNTO 0),
-      out_dat => hdr_fields_in_piped_arr(i)(field_slv_len(g_hdr_field_arr)-1 DOWNTO 0)
-    );
-  END GENERATE;
 
-  ---------------------------------------------------------------------------------------
-  -- If the user defined internally used fields in g_hdr_field_arr, wire the corresponding
-  -- internal status signals to these header fields with the correct delay, so they are
-  -- valid at dp_packet_merge_src_out_arr(i).sop like the other header fields.
-  ---------------------------------------------------------------------------------------
-  gen_common_pipeline_stream : FOR i IN 0 TO g_nof_streams-1 GENERATE
-    u_common_pipeline : ENTITY common_lib.common_pipeline
-    GENERIC MAP (
-      g_pipeline  => c_dp_packet_merge_val_latency,
-      g_in_dat_w  => ceil_log2(g_max_nof_words_per_block+1),
-      g_out_dat_w => ceil_log2(g_max_nof_words_per_block+1)
-    )
-    PORT MAP (
-      rst     => dp_rst,
-      clk     => dp_clk,
-      in_dat  => nof_words_per_block_stat_arr(i),
-      out_dat => nof_words_per_block_stat_piped_arr(i)
-    );
-  END GENERATE;
-    
-  gen_dp_offload_tx_len_calc_stream : FOR i IN 0 TO g_nof_streams-1 GENERATE
-    u_dp_offload_tx_len_calc : ENTITY work.dp_offload_tx_len_calc
-    GENERIC MAP (
-      g_hdr_field_arr => g_hdr_field_arr,
-      g_data_w        => g_data_w
-    )
-    PORT MAP (
-      rst  => dp_rst,
-      clk  => dp_clk,
-  
-      nof_words_per_block   => nof_words_per_block_stat_piped_arr(i),
-      nof_blocks_per_packet => nof_blocks_per_packet_stat_arr(i),
-  
-      udp_total_length => udp_total_length_arr(i),
-      ip_total_length  => ip_total_length_arr(i)
-    );
-  END GENERATE;
-
-  p_overwrite_fields: PROCESS(nof_words_per_block_stat_piped_arr, hdr_fields_in_piped_arr, nof_blocks_per_packet_stat_arr, udp_total_length_arr, ip_total_length_arr)
-  BEGIN
-    hdr_fields_to_field_blk <= hdr_fields_in_piped_arr;
-    FOR i IN 0 TO g_nof_streams-1 LOOP
-      IF field_exists(g_hdr_field_arr, "nof_words_per_block") THEN
-        hdr_fields_to_field_blk(i)(field_hi(g_hdr_field_arr, "nof_words_per_block") DOWNTO field_lo(g_hdr_field_arr, "nof_words_per_block")) <= RESIZE_UVEC(nof_words_per_block_stat_piped_arr(i), field_size(g_hdr_field_arr, "nof_words_per_block"));
-      END IF;
-      IF field_exists(g_hdr_field_arr, "nof_blocks_per_packet") THEN
-        hdr_fields_to_field_blk(i)(field_hi(g_hdr_field_arr, "nof_blocks_per_packet") DOWNTO field_lo(g_hdr_field_arr, "nof_blocks_per_packet")) <= RESIZE_UVEC(nof_blocks_per_packet_stat_arr(i), field_size(g_hdr_field_arr, "nof_blocks_per_packet"));
-      END IF;
-      IF field_exists(g_hdr_field_arr, "udp_total_length") THEN
-        hdr_fields_to_field_blk(i)(field_hi(g_hdr_field_arr, "udp_total_length") DOWNTO field_lo(g_hdr_field_arr, "udp_total_length")) <= RESIZE_UVEC(udp_total_length_arr(i), field_size(g_hdr_field_arr, "udp_total_length"));
-      END IF;
-      IF field_exists(g_hdr_field_arr, "ip_total_length") THEN
-        hdr_fields_to_field_blk(i)(field_hi(g_hdr_field_arr, "ip_total_length") DOWNTO field_lo(g_hdr_field_arr, "ip_total_length")) <= RESIZE_UVEC(ip_total_length_arr(i), field_size(g_hdr_field_arr, "ip_total_length"));
-      END IF;
-    END LOOP;
-  END PROCESS;
-
-  ---------------------------------------------------------------------------------------
-  -- Frame buffer 
-  ---------------------------------------------------------------------------------------
-  gen_dp_fifo_sc : FOR i IN 0 TO g_nof_streams-1 GENERATE
-    u_dp_fifo_sc_frame_buffer : ENTITY work.dp_fifo_sc
-    GENERIC MAP (
-      g_data_w      => g_data_w,
-      g_use_ctrl    => TRUE,
-      g_fifo_size   => g_output_fifo_depth + 10
-    )
-    PORT MAP (
-      rst         => dp_rst,
-      clk         => dp_clk,
-
-      usedw       => dp_fifo_sc_usedw_arr(i),
+  gen_dp_pipeline_hdr_fields_in_arr : FOR i IN 0 TO g_nof_streams-1 GENERATE
 
-      snk_out     => dp_packet_merge_src_in_arr(i),   
-      snk_in      => dp_packet_merge_src_out_arr(i),
+    -- FIXME - Put this DP bus on the entity instead of hdr_fields_in_arr.
+    dp_pipeline_hdr_fields_snk_in_arr(i).data(field_slv_len(g_hdr_field_arr)-1 DOWNTO 0) <= hdr_fields_in_arr(i)(field_slv_len(g_hdr_field_arr)-1 DOWNTO 0);
+    dp_pipeline_hdr_fields_snk_in_arr(i).valid <= snk_in_arr(i).sop;
 
-      src_in      => dp_frame_rd_snk_out_arr(i),
-      src_out     => dp_frame_rd_snk_in_arr(i)
-    );
-
-    -- Release frame at EOP. Also release the frame when FIFO gets too full (if user 
-    -- created packets that are too large) to prevent lockups. 
-    dp_frame_rd_frm_req_arr(i) <= '1' WHEN TO_UINT(dp_fifo_sc_usedw_arr(i)) > g_output_fifo_depth ELSE dp_packet_merge_src_out_arr(i).eop;      
-    --dp_frame_rd_frm_req_arr(i) <= dp_packet_merge_src_out_arr(i).eop;
-
-    -- Xon bypasses dp_frame_rd here
-    dp_frame_rd_snk_out_arr(i).xon <= dp_concat_snk_out_2arr(i)(0).xon;
-
-    u_dp_frame_rd : ENTITY work.dp_frame_rd
+    u_dp_pipeline : ENTITY work.dp_pipeline
     GENERIC MAP (
-      g_dat_w          => g_data_w,
-      g_empty_w        => 0,
-      g_channel_w      => 0,
-      g_error_w        => 0,
-      g_frm_cnt_max    => g_max_nof_blocks_per_packet  -- sufficient to ensure no frm_req are missed that would cause the FIFO to fill with frames and eventually overflow
+      g_pipeline  => c_dp_split_val_latency + c_dp_packet_merge_val_latency
     )
     PORT MAP (
       rst     => dp_rst,
       clk     => dp_clk,
 
-      frm_req          => dp_frame_rd_frm_req_arr(i),
-      frm_flush        => '0',
-      frm_ack          => OPEN,
-      frm_busy         => OPEN,
-      frm_err          => OPEN,
-      frm_done         => OPEN,
-      
-      rd_req           => dp_frame_rd_snk_out_arr(i).ready,
-      
-      rd_dat           => dp_frame_rd_snk_in_arr(i).data(g_data_w-1 DOWNTO 0),
-      rd_sync          => dp_frame_rd_snk_in_arr(i).sync,
-      rd_val           => dp_frame_rd_snk_in_arr(i).valid,
-      rd_sof           => dp_frame_rd_snk_in_arr(i).sop,
-      rd_eof           => dp_frame_rd_snk_in_arr(i).eop,
-      
-      out_dat          => dp_frame_rd_src_out_arr(i).data(g_data_w-1 DOWNTO 0),
-      out_sync         => dp_frame_rd_src_out_arr(i).sync,
-      out_val          => dp_frame_rd_src_out_arr(i).valid,
-      out_sof          => dp_frame_rd_src_out_arr(i).sop,
-      out_eof          => dp_frame_rd_src_out_arr(i).eop
-    );
+      snk_in  => dp_pipeline_hdr_fields_snk_in_arr(i),
+      snk_out => OPEN, --dp_pipeline_hdr_fields_snk_out_arr(i), -- snk_in is only valid at the SOP so we won't need this
 
-    -- dp_frame_rd does not have output flow control
-    -- . Depth: 
-    --   . c_nof_header_words for dp_concat
-    --   . 12 to buffer during non-ready cycles of eth_hdr.vhd in eth.vhd
-    --   . g_output_fifo_depth to buffer when ethernet MAC itself is not ready and its input FIFO is almost full
-    --     . This is the same depth as the FIFO in front of dp_frame_rd...
-    --   . 10 as margin
-    u_dp_fifo_sc_flow_control : ENTITY work.dp_fifo_sc
-    GENERIC MAP (
-      g_data_w      => g_data_w,
-      g_use_ctrl    => TRUE,
-      g_fifo_size   => c_nof_header_words + 12 + g_output_fifo_depth + 10 
-    )
-    PORT MAP (
-      rst         => dp_rst,
-      clk         => dp_clk,
-  
-      snk_out     => OPEN,   
-      snk_in      => dp_frame_rd_src_out_arr(i),
-  
-      src_in      => dp_concat_snk_out_2arr(i)(0),
-      src_out     => dp_concat_snk_in_2arr(i)(0)
+      src_out => dp_pipeline_hdr_fields_src_out_arr(i),
+      src_in  => dp_pipeline_hdr_fields_src_in_arr(i)
     );
   END GENERATE;
 
+--  hdr_fields_to_field_blk(i)(field_slv_len(g_hdr_field_arr)-1 DOWNTO 0) <= dp_pipeline_hdr_fields_src_out_arr(i).data(field_slv_len(g_hdr_field_arr)-1 DOWNTO 0);
+    
   ---------------------------------------------------------------------------------------
   -- Create header block & concatenate header to offload stream.
   -- . hdr_fields_in_arr(i) is clocked in at snk_in_arr(i).sop after which it can be
@@ -411,8 +313,12 @@ BEGIN
   ---------------------------------------------------------------------------------------
   gen_dp_field_blk : FOR i IN 0 TO g_nof_streams-1 GENERATE
 
-  dp_field_blk_snk_in_arr(i).data(c_dp_field_blk_snk_data_w-1 DOWNTO 0) <= hdr_fields_to_field_blk(i)(field_slv_len(g_hdr_field_arr)-1 DOWNTO 0);
-  dp_field_blk_snk_in_arr(i).valid <= dp_packet_merge_src_out_arr(i).sop;
+    -- Flow control for the header source. Should not be needed.
+    dp_pipeline_hdr_fields_src_in_arr(i) <= dp_field_blk_snk_out_arr(i);
+
+--    dp_field_blk_snk_in_arr(i).data(c_dp_field_blk_snk_data_w-1 DOWNTO 0) <= hdr_fields_to_field_blk(i)(field_slv_len(g_hdr_field_arr)-1 DOWNTO 0);
+--    dp_field_blk_snk_in_arr(i).valid <= dp_packet_merge_src_out_arr(i).sop;
+    dp_field_blk_snk_in_arr(i) <= dp_pipeline_hdr_fields_src_out_arr(i);
 
     u_dp_field_blk : ENTITY work.dp_field_blk
     GENERIC MAP (
@@ -428,7 +334,8 @@ BEGIN
       mm_rst       => mm_rst,
       mm_clk       => mm_clk,
   
-      snk_in       => dp_field_blk_snk_in_arr(i),
+      snk_in       => dp_field_blk_snk_in_arr(i), -- Valid for only one cycle.
+      snk_out      => dp_field_blk_snk_out_arr(i),
 
 --      slv_in       => hdr_fields_to_field_blk(i)(field_slv_len(g_hdr_field_arr)-1 DOWNTO 0),
 --      slv_in_val   => dp_packet_merge_src_out_arr(i).sop,