Comparing mclt Bayer with Java results, fixing bugs

dsp/mclt16x16_bayer.v

@@ -41,7 +41,7 @@
 module  mclt16x16_bayer#(
     parameter SHIFT_WIDTH =      7, // bits in shift (7 bits - fractional)
     parameter PIX_ADDR_WIDTH =   9, // number of pixel address width
-//    parameter EXT_PIX_LATENCY =  2, // external pixel buffer a->d latency
+    parameter EXT_PIX_LATENCY =  2, // external pixel buffer a->d latency (may increase to 4 for gamma)
     parameter COORD_WIDTH =     10, // bits in full coordinate 10 for 18K RAM
     parameter PIXEL_WIDTH =     16, // input pixel width (unsigned) 
     parameter WND_WIDTH =       18, // input pixel width (unsigned) 
@@ -104,12 +104,12 @@ module  mclt16x16_bayer#(
     reg                   inv_checker_r3;                
     reg                   inv_checker_r4;                
 
-    wire                         [1:0] signs;        //!< bit 0: sign to add to dtt-cc input, bit 1: sign to add to dtt-cs input
-    wire                        [6:0] phases;        //!< other signals
+    wire                     [1:0] signs;        //!< bit 0: sign to add to dtt-cc input, bit 1: sign to add to dtt-cs input
+    wire                     [6:0] phases;        //!< other signals
     
-    wire signed   [WND_WIDTH-1:0] window_w;
+    wire signed    [WND_WIDTH-1:0] window_w;
  
-    wire  signed [DTT_IN_WIDTH-1:0] data_dtt_in; // multiplexed DTT input data
+    wire signed [DTT_IN_WIDTH-1:0] data_dtt_in; // multiplexed DTT input data
     
        
     reg                      [6:0] dtt_in_cntr; //
@@ -175,7 +175,7 @@ module  mclt16x16_bayer#(
         end
 
 //         if (!phases[14]) dtt_in_cntr <= 0; 
-         if (!dtt_we) dtt_in_cntr <= 0; 
+         if     (!dtt_we) dtt_in_cntr <= 0; 
          else             dtt_in_cntr <= dtt_in_cntr + 1;
    
          start_dtt <= dtt_in_cntr == DTT_IN_DELAY;
@@ -199,23 +199,20 @@ module  mclt16x16_bayer#(
     
     end
     
+`ifdef DSP_ACCUM_FOLD
+        localparam ADDR_DLYL = 4 - EXT_PIX_LATENCY; // 4'h2; // 3 for mpy, 2 - for dsp
+`else        
+        localparam ADDR_DLYL = 5 - EXT_PIX_LATENCY; // 4'h3; // 3 for mpy, 2 - for dsp
+`endif        
+  
+  
       
     mclt_bayer_fold #(
         .SHIFT_WIDTH     (SHIFT_WIDTH),
         .PIX_ADDR_WIDTH  (PIX_ADDR_WIDTH),
-        .ADDR_DLY        (4'h1), // 2 for mpy, 1 - for dsp
+        .ADDR_DLY        (ADDR_DLYL), // 3 for mpy, 2 - for dsp
         .COORD_WIDTH     (COORD_WIDTH),
-//        .PIXEL_WIDTH     (PIXEL_WIDTH),
         .WND_WIDTH       (WND_WIDTH)
-//        .OUT_WIDTH       (OUT_WIDTH),
-//        .DTT_IN_WIDTH    (DTT_IN_WIDTH),
-//        .TRANSPOSE_WIDTH (TRANSPOSE_WIDTH),
-//        .OUT_RSHIFT      (OUT_RSHIFT),
-//        .OUT_RSHIFT2     (OUT_RSHIFT2),
-//        .DSP_B_WIDTH     (DSP_B_WIDTH),
-//        .DSP_A_WIDTH     (DSP_A_WIDTH),
-//        .DSP_P_WIDTH     (DSP_P_WIDTH),
-//        .DEAD_CYCLES     (DEAD_CYCLES)
     ) mclt_bayer_fold_i (
         .clk           (clk),         // input
         .rst           (rst),         // input
@@ -233,7 +230,6 @@ module  mclt16x16_bayer#(
         .signs         (signs),       // output[1:0] 
         .phases        (phases),      // output[7:0]
         .var_pre2_first(var_pre2_first), // output
-//        .var_first     (), // var_first),   // output reg
         .pre_last_in   (pre_last_in_w)// output reg
     );
 
@@ -268,9 +264,11 @@ module  mclt16x16_bayer#(
     wire                        dtt_inc16;
     reg                   [4:0] dtt_out_ram_cntr;
     reg                   [4:0] dtt_out_ram_wah;
-    reg                   [1:0] dtt_out_ram_wpage; // one of 4 pages (128 samples long) being written to
+    reg                   [1:0] dtt_out_ram_wpage;  // one of 4 pages (128 samples long) being written to
+    reg                   [1:0] dtt_out_ram_wpage2; // later by 1 DTT
     wire                        dtt_start_fill; // some data available in DTT output buffer, OK to start consecutive readout
     reg                         dtt_start_first_fill;
+    reg                         dtt_start_second_fill;
     reg                   [1:0] dtt_start_out;  // start read out to sin/cos rotator
     
 
@@ -305,7 +303,11 @@ module  mclt16x16_bayer#(
         
         dtt_start_first_fill <= dtt_start_fill & dtt_first_quad_out;
         
-        if (dtt_start_first_fill) dtt_out_ram_wpage <= dtt_out_ram_wah[4:3]; 
+        dtt_start_second_fill<= dtt_start_fill & ~dtt_first_quad_out;
+        
+        if (dtt_start_first_fill) dtt_out_ram_wpage <= dtt_out_ram_wah[4:3];
+         
+        if (dtt_start_second_fill) dtt_out_ram_wpage2 <= dtt_out_ram_wpage; 
 
         if      (rst)                  dtt_dly_cntr <= 0;
         else if (dtt_start_first_fill) dtt_dly_cntr <= DTT_OUT_DELAY;
@@ -320,21 +322,10 @@ module  mclt16x16_bayer#(
         if      (rst)                   dtt_rd_regen_dv[3:1] <= 0;
         else                            dtt_rd_regen_dv[3:1] <= dtt_rd_regen_dv[2:0];
         
-        if (dtt_start_out[0])           dtt_rd_cntr_pre <= {dtt_out_ram_wpage, 7'b0}; //copy page number
-        
+//        if (dtt_start_out[0])           dtt_rd_cntr_pre <= {dtt_out_ram_wpage, 7'b0}; //copy page number
+        if (dtt_start_out[0])           dtt_rd_cntr_pre <= {dtt_out_ram_wpage2, 7'b0}; //copy page number
         else if (dtt_rd_regen_dv[0]) dtt_rd_cntr_pre <= dtt_rd_cntr_pre + 1;
 
-/*
-        dtt_rd_ra0 <= {dtt_rd_cntr_pre[8:7],
-                       dtt_rd_cntr_pre[6] ^ dtt_rd_cntr_pre[5],
-                       dtt_rd_cntr_pre[5]? (~dtt_rd_cntr_pre[4:0]) : dtt_rd_cntr_pre[4:0],
-                       dtt_rd_cntr_pre[5]};    
-        dtt_rd_ra1 <= {dtt_rd_cntr_pre[8:7],
-                       dtt_rd_cntr_pre[6] ^ dtt_rd_cntr_pre[5],
-                       dtt_rd_cntr_pre[5]? (~dtt_rd_cntr_pre[4:0]) : dtt_rd_cntr_pre[4:0],
-                       ~dtt_rd_cntr_pre[5]};    
-
-*/        
         dtt_rd_ra0 <= {dtt_rd_cntr_pre[8:7],
                        dtt_rd_cntr_pre[0] ^ dtt_rd_cntr_pre[1],
                        dtt_rd_cntr_pre[0]? (~dtt_rd_cntr_pre[6:2]) : dtt_rd_cntr_pre[6:2],
@@ -362,8 +353,8 @@ module  mclt16x16_bayer#(
         .pix_sgn   (signs),        // input[1:0] 
         .window    (window_w),     // input[17:0] signed 
         .var_pre2_first (var_pre2_first),    // input
-        .dtt_in    (data_dtt_in),   // output[24:0] signed
-        .dtt_in_dv (dtt_we) // output reg 
+        .dtt_in    (data_dtt_in),  // output[24:0] signed
+        .dtt_in_dv (dtt_we)        // output reg 
     );
     
    

dsp/mclt_baeyer_fold_accum.v

@@ -37,7 +37,7 @@
  * with at least one of the Free Software programs.
  */
 `timescale 1ns/1ps
-`define DSP_ACCUM_FOLD 1
+//`define DSP_ACCUM_FOLD 1
 module  mclt_baeyer_fold_accum # (
     parameter PIXEL_WIDTH =     16, // input pixel width (unsigned) 
     parameter WND_WIDTH =       18, // input pixel width (unsigned) 
@@ -49,15 +49,13 @@ module  mclt_baeyer_fold_accum # (
 )(
     input                            clk,
     input                            rst,
-//    input                      [5:0] phases,
     input                            pre_phase,
-    input signed   [PIXEL_WIDTH-1:0] pix_d,    //!< pixel data
+    input signed   [PIXEL_WIDTH-1:0] pix_d,    //!< pixel data (should be 1 cycle later for `undef DSP_ACCUM_FOLD
     input                      [1:0] pix_sgn,  //!< bit 0: sign to add to dtt-cc input, bit 1: sign to add to dtt-cs input
     input signed     [WND_WIDTH-1:0] window,
     input                            var_pre2_first,
     output signed [DTT_IN_WIDTH-1:0] dtt_in,
     output                           dtt_in_dv  
-
 );
     reg                           var_pre_first;
     reg                           var_first;
@@ -65,43 +63,6 @@ module  mclt_baeyer_fold_accum # (
     
     reg                      [6:0] phases;
     
-`ifdef DSP_ACCUM_FOLD
-    reg                            dtt_in_dv_dsp_r;
-    reg signed [DTT_IN_WIDTH-1:0]  dtt_in_dsp; 
-`else
-    wire                   [ 1:0] pix_sgn_d;
-    reg            [PIXEL_WIDTH-1:0] pix_dr;         // only for mpy to match dsp
-    reg  signed   [WND_WIDTH-1:0] window_r;
-    reg  signed [PIXEL_WIDTH-1:0] pix_d_r;         // registered pixel data (to be absorbed by MPY)
-    reg                    [ 1:0] pix_sgn_r;
-    
-    reg  signed [PIXEL_WIDTH + WND_WIDTH - 1:0] pix_wnd_r; // MSB not used: positive[PIXEL_WIDTH]*positive[WND_WIDTH]->positive[PIXEL_WIDTH+WND_WIDTH-1]
-    reg  signed              [DTT_IN_WIDTH-1:0] pix_wnd_r2; // pixels (positive) multiplied by window(positive), two MSBs == 2'b0 to prevent overflow
-    // rounding
-//    wire signed              [DTT_IN_WIDTH-3:0] pix_wnd_r2_w = pix_wnd_r[PIXEL_WIDTH + WND_WIDTH - 2 -: DTT_IN_WIDTH - 2]
-    wire signed              [DTT_IN_WIDTH-2:0] pix_wnd_r2_w = pix_wnd_r[PIXEL_WIDTH + WND_WIDTH - 2 -: DTT_IN_WIDTH - 1]
-    `ifdef ROUND
-//                 + pix_wnd_r[PIXEL_WIDTH + WND_WIDTH -DTT_IN_WIDTH]
-                 + pix_wnd_r[PIXEL_WIDTH + WND_WIDTH -DTT_IN_WIDTH -1]
-    `endif
-    ;
-    reg  signed [DTT_IN_WIDTH-1:0] data_cc_r;   
-    reg  signed [DTT_IN_WIDTH-1:0] data_sc_r;
-    reg  signed [DTT_IN_WIDTH-1:0] data_sc_r2; // data_sc_r delayed by 1 cycle 
-    reg                            mode_mux;
-    reg                            dtt_in_dv_r;
-    reg  signed [DTT_IN_WIDTH-1:0] data_dtt_in; // multiplexed DTT input data
-`endif
-    
-    
-    
-    `ifdef DSP_ACCUM_FOLD
-        assign dtt_in =    dtt_in_dsp;
-        assign dtt_in_dv = dtt_in_dv_dsp_r;
-    `else
-        assign dtt_in =    data_dtt_in;
-        assign dtt_in_dv = dtt_in_dv_r;
-    `endif
     always @ (posedge clk) begin
         phases <= {phases[5:0], pre_phase};
 
@@ -119,6 +80,12 @@ module  mclt_baeyer_fold_accum # (
 
 
 `ifdef DSP_ACCUM_FOLD
+    reg                            dtt_in_dv_dsp_r;
+    reg signed [DTT_IN_WIDTH-1:0]  dtt_in_dsp; 
+
+    assign dtt_in =    dtt_in_dsp;
+    assign dtt_in_dv = dtt_in_dv_dsp_r;
+    
     always @ (posedge clk) begin
         if (rst) dtt_in_dv_dsp_r <= 0;
         else     dtt_in_dv_dsp_r <= phases[5];
@@ -223,13 +190,40 @@ module  mclt_baeyer_fold_accum # (
     );
 
 `else
+    wire                   [ 1:0] pix_sgn_d;
+///    reg            [PIXEL_WIDTH-1:0] pix_dr;         // only for mpy to match dsp
+    reg  signed   [WND_WIDTH-1:0] window_r;
+    reg  signed [PIXEL_WIDTH-1:0] pix_d_r;         // registered pixel data (to be absorbed by MPY)
+    reg                    [ 1:0] pix_sgn_r;
+    
+    reg  signed [PIXEL_WIDTH + WND_WIDTH - 1:0] pix_wnd_r; // MSB not used: positive[PIXEL_WIDTH]*positive[WND_WIDTH]->positive[PIXEL_WIDTH+WND_WIDTH-1]
+    reg  signed              [DTT_IN_WIDTH-1:0] pix_wnd_r2; // pixels (positive) multiplied by window(positive), two MSBs == 2'b0 to prevent overflow
+    // rounding
+//    wire signed              [DTT_IN_WIDTH-3:0] pix_wnd_r2_w = pix_wnd_r[PIXEL_WIDTH + WND_WIDTH - 2 -: DTT_IN_WIDTH - 2]
+    wire signed              [DTT_IN_WIDTH-2:0] pix_wnd_r2_w = pix_wnd_r[PIXEL_WIDTH + WND_WIDTH - 2 -: DTT_IN_WIDTH - 1]
+    `ifdef ROUND
+//                 + pix_wnd_r[PIXEL_WIDTH + WND_WIDTH -DTT_IN_WIDTH]
+                 + pix_wnd_r[PIXEL_WIDTH + WND_WIDTH -DTT_IN_WIDTH -1]
+    `endif
+    ;
+    reg  signed [DTT_IN_WIDTH-1:0] data_cc_r;   
+    reg  signed [DTT_IN_WIDTH-1:0] data_sc_r;
+    reg  signed [DTT_IN_WIDTH-1:0] data_sc_r2; // data_sc_r delayed by 1 cycle 
+    reg                            mode_mux;
+    reg                            dtt_in_dv_r;
+    reg  signed [DTT_IN_WIDTH-1:0] data_dtt_in; // multiplexed DTT input data
+
+    assign dtt_in =    data_dtt_in;
+    assign dtt_in_dv = dtt_in_dv_r;
+
     always @ (posedge clk) begin
         if (rst) dtt_in_dv_r <= 0;
         else     dtt_in_dv_r <= phases[6];
 
-        pix_dr <= pix_d;
+///        pix_dr <= pix_d;
         if (phases[1]) begin
-            pix_d_r <=    pix_dr;
+///            pix_d_r <=    pix_dr;
+            pix_d_r <=    pix_d;
             window_r <=   window;
         end
         if (phases[2])  pix_wnd_r <= pix_d_r * window_r; // 1 MSB is extra

dsp/mclt_bayer_fold.v

@@ -41,20 +41,9 @@
 module  mclt_bayer_fold#(
     parameter SHIFT_WIDTH =      7, // bits in shift (7 bits - fractional)
     parameter PIX_ADDR_WIDTH =   9, // number of pixel address width
-//    parameter EXT_PIX_LATENCY =  2, // external pixel buffer a->d latency
-    parameter ADDR_DLY =        4'h2, // extra delay of pixel address to match window delay
+    parameter ADDR_DLY =        4'h3, // extra delay of pixel address to match window delay
     parameter COORD_WIDTH =     10, // bits in full coordinate 10 for 18K RAM
-//    parameter PIXEL_WIDTH =     16, // input pixel width (unsigned) 
     parameter WND_WIDTH =       18 // input pixel width (unsigned) 
-//    parameter OUT_WIDTH =       25, // bits in dtt output
-//    parameter DTT_IN_WIDTH =    25, // bits in DTT input
-//    parameter TRANSPOSE_WIDTH = 25, // width of the transpose memory (intermediate results)    
-//    parameter OUT_RSHIFT =       2, // overall right shift of the result from input, aligned by MSB (>=3 will never cause saturation)
-//    parameter OUT_RSHIFT2 =      0, // overall right shift for the second (vertical) pass
-//    parameter DSP_B_WIDTH =     18, // signed, output from sin/cos ROM
-//    parameter DSP_A_WIDTH =     25,
-//    parameter DSP_P_WIDTH =     48,
-//    parameter DEAD_CYCLES =     14  // start next block immedaitely, or with longer pause
 )(
     input                             clk,          //!< system clock, posedge
     input                             rst,          //!< sync reset
@@ -71,14 +60,12 @@ module  mclt_bayer_fold#(
     output                            pix_page,     //!< copy pixel page (should be externally combined with first color)
     output signed     [WND_WIDTH-1:0] window,       //!< msb==0, always positive
     output                      [1:0] signs,        //!< bit 0: sign to add to dtt-cc input, bit 1: sign to add to dtt-cs input
-//    output                     [14:0] phases,       //!< other signals
     output                      [6:0] phases,       //!< other signals
     output                            var_pre2_first,//!< two ahead of first of 2 fold variants (4 for monochrome, 2 left for checker)
-//    output reg                        var_first,    //!< first of 2 fold variants (4 for monochrome, 2 left for checker)
     output reg                        pre_last_in   //!< pre last data in
 );
+
     reg                         [6:0] in_cntr;      // input phase counter
-//    reg                        [14:0] run_r;        // run phase
     reg                         [6:0] run_r;        // run phase
     
     reg                         [1:0] tile_size_r;  // 0: 16x16, 1 - 18x18, 2 - 20x20, 3 - 22x22 (max for 9-bit addr)
@@ -86,9 +73,11 @@ module  mclt_bayer_fold#(
     reg                         [7:0] top_left_r0;  // index of the 16x16 top left corner
     reg                         [7:0] top_left_r;   // index of the 16x16 top left corner
     reg                         [1:0] valid_rows_r0;// 3 for green, 1 or 2 for R/B - which of the even/odd checker rows contain pixels
-//    reg                         [1:0] valid_rows_r ;// correct latency for window rom
-/// wire                       [ 9:0] fold_addr=    {tile_size_r,inv_checker_r, in_cntr[0],in_cntr[6:1]}; 
-    wire                       [ 9:0] fold_addr=    {tile_size_r,inv_checker_r, (valid_rows_r0==3)?in_cntr[0]:valid_rows_r0[0],in_cntr[6:1]}; 
+    wire                       [ 9:0] fold_addr=    {tile_size_r,inv_checker_r, (valid_rows_r0==3)?
+                                                     in_cntr[0]:
+//                                                     valid_rows_r0[0],
+                                                     ~valid_rows_r0[0],
+                                                     in_cntr[6:1]}; 
     reg             [SHIFT_WIDTH-1:0] x_shft_r0;    // tile pixel X fractional shift (valid @ start) 
     reg             [SHIFT_WIDTH-1:0] y_shft_r0;    // tile pixel Y fractional shift (valid @ start)
     reg             [SHIFT_WIDTH-1:0] x_shft_r;     // matching delay 
@@ -100,28 +89,16 @@ module  mclt_bayer_fold#(
     wire         [PIX_ADDR_WIDTH-1:0] pix_a_w =   {~fold_rom_out[15] & fold_rom_out[7],fold_rom_out[15:8]};
     reg          [PIX_ADDR_WIDTH-1:0] pix_a_r;
     wire                       [ 1:0] sgn_w =     fold_rom_out[16 +: 2];
-//    reg                               blank_r;    // blank window (latency 1 from fold_rom_out)
-//    wire                              blank_d;    //  delayed to matchwindow rom regrst
     wire                              pre_page =  in_cntr == 2; // valid 1 cycle before fold_rom_out
     
     wire                              var_first_d; // adding subtracting first variant of 2 folds    
-//    reg                               var_pre_first;
     assign phases = run_r;
     assign var_pre2_first = var_first_d;
     
-//    wire           [ 3:0] bayer_1hot = { mpix_a_w[4] &  mpix_a_w[0],  
-//                                        mpix_a_w[4] & ~mpix_a_w[0],
-//                                        ~mpix_a_w[4] &  mpix_a_w[0],
-//                                        ~mpix_a_w[4] & ~mpix_a_w[0]};
-                                        
-//    wire                          mpix_use = |(bayer_d & bayer_1hot); //not disabled by bayer, valid with mpix_a_w
-//    wire                          mpix_use_d; // delayed
-//    reg                           mpix_use_r; // delayed
 
    
     always @ (posedge clk) begin
         if (rst) run_r <= 0;
-//        else     run_r <= {run_r[13:0], start | (run_r[0] & ~(&in_cntr[6:0]))};
         else     run_r <= {run_r[5:0], start | (run_r[0] & ~(&in_cntr[6:0]))};
         
         if (!run_r[0]) in_cntr <= 0;
@@ -145,18 +122,6 @@ module  mclt_bayer_fold#(
             
 
         if (run_r[2]) pix_a_r <= pix_a_w + {1'b0, top_left_r};
-        
-
-///        if (in_cntr == 2) valid_rows_r <= valid_rows_r0;
-        
-///        blank_r <= ~(wnd_a_w[0] ? valid_rows_r[1]: valid_rows_r[0]);
-
-///        if (run_r[9]) var_pre_first <= var_first_d;  
-        
-///        if (run_r[10]) begin
-//            var_first <= var_first_d;
-///            var_first <= var_pre_first;
-///        end
          
         pre_last_in <= in_cntr[6:0] == 7'h7d;
          
@@ -192,18 +157,25 @@ module  mclt_bayer_fold#(
     );
 
 // Matching window latency with pixel data latency
-    wire [3:0] addr_dly = ADDR_DLY;
-    dly_var #(
-        .WIDTH(11),
-        .DLY_WIDTH(4)
-    ) dly_pixel_addr_i (
-        .clk  (clk),      // input
-        .rst  (rst),      // input
-//        .dly  (4'h2),     // input[3:0] Delay for external memory latency = 2, reduce for higher 
-        .dly  (addr_dly),     // input[3:0] Delay for external memory latency = 2, reduce for higher 
-        .din  ({pre_page, run_r[3], pix_a_r}), // input[0:0] 
-        .dout ({pix_page,   pix_re, pix_addr}) // output[0:0] 
-    );
+    generate
+        if (ADDR_DLY !=0) begin
+            wire [3:0] addr_dly = ADDR_DLY - 1; // iverilog problem mitigation
+            dly_var #(
+                .WIDTH(11),
+                .DLY_WIDTH(4)
+            ) dly_pixel_addr_i (
+                .clk  (clk),      // input
+                .rst  (rst),      // input
+                .dly  (addr_dly), // input[3:0] Delay for external memory latency = 2, reduce for higher 
+                .din  ({pre_page, run_r[3], pix_a_r}), // input[0:0] 
+                .dout ({pix_page,   pix_re, pix_addr}) // output[0:0] 
+            );
+        end else begin
+            assign pix_page = pre_page;
+            assign pix_re =   run_r[3];
+            assign pix_addr = pix_a_r;
+        end
+    endgenerate
 
 // Latency = 6
     mclt_wnd_mul #(
@@ -238,13 +210,10 @@ module  mclt_bayer_fold#(
     ) dly_var_first_i (
         .clk  (clk),           // input
         .rst  (rst),           // input
-//        .dly  (4'h9),          // input[3:0] 
         .dly  (4'h8),          // input[3:0] 
         .din  (run_r[0] && (in_cntr[0] == 0)),  // input[0:0] 
         .dout (var_first_d)    // output[0:0] 
     );
 
-    
-//
 endmodule
 

dsp/mclt_test_03.tf

@@ -508,7 +508,9 @@ module  mclt_test_03 ();
     );
 
     localparam PIX_ADDR_WIDTH = 9;
-    localparam ADDR_DLY = 2;
+//    localparam ADDR_DLY = 2;
+    localparam EXT_PIX_LATENCY =  2; // external pixel buffer a->d latency (may increase to 4 for gamma)
+    
     reg                   [1:0] TILE_SIZE =   3; // 22;
     reg                         INV_CHECKER = 0;
     reg                   [7:0] TOP_LEFT =    69; // center
@@ -540,6 +542,7 @@ module  mclt_test_03 ();
     mclt16x16_bayer #(
         .SHIFT_WIDTH     (SHIFT_WIDTH),
         .PIX_ADDR_WIDTH  (PIX_ADDR_WIDTH),
+        .EXT_PIX_LATENCY (EXT_PIX_LATENCY), // 2), // external pixel buffer a->d latency (may increase to 4 for gamma)
         .COORD_WIDTH     (COORD_WIDTH),
         .PIXEL_WIDTH     (PIXEL_WIDTH),
         .WND_WIDTH       (WND_WIDTH),
@@ -552,7 +555,7 @@ module  mclt_test_03 ();
         .DSP_A_WIDTH     (DSP_A_WIDTH),
         .DSP_P_WIDTH     (DSP_P_WIDTH),
         .DEAD_CYCLES     (DEAD_CYCLES)
-    ) mclt_bayer_fold_i (
+    ) mclt16x16_bayer_i (
         .clk         (CLK),         // input
         .rst         (RST),         // input
         .start       (start),       // input
@@ -565,7 +568,16 @@ module  mclt_test_03 ();
         .pix_addr    (PIX_ADDR9),   // output[8:0] 
         .pix_re      (PIX_RE),      // output
         .pix_page    (PIX_PAGE),    // output
-        .pix_d       (PIX_D)        // input[15:0] 
+        .pix_d       (PIX_D),        // input[15:0]
+        .pre_busy      (), // output
+        .pre_last_in   (), // output
+        .pre_first_out (), // output
+        .pre_last_out  (), // output
+        .out_addr      (), // output[7:0] 
+        .dv            (), // output
+        .dout0         (), // output[24:0] signed 
+        .dout1         () // output[24:0] signed 
+         
     );
 
 

system_defines.vh

@@ -43,6 +43,9 @@
   `define DEBUG_HISTOGRAMS
   // TODO: Later compare instantiate/infer
   `define INSTANTIATE_DSP48E1 // not yet tesetd/debugged otherwise
+  `define DSP_ACCUM_FOLD 1 // for MCLT  
+  
+  
 // https://forums.xilinx.com/t5/Embedded-Processor-System-Design/AXI4-Bursts-4KB-Address-Boundary-Limitation/td-p/216413
 // Interconnect does not have 4K limit, and compressed data can only go to interconnect (memory), so it is OK to violate AXI specs here
   `define AXI_4K_LIMIT_DISABLE // Current x393 code (only simulation modules) does not have it implemented, defining it causes mismatch synth/sim