From 68922a34d25aa18c732469a5908e82f75fc57db3 Mon Sep 17 00:00:00 2001
From: Konstantin Pavlov <pavlovconst@gmail.com>
Date: Thu, 17 Dec 2020 18:42:52 +0300
Subject: [PATCH] Updated benchmark projects. Added 'benchmark_results.txt'
 that i got on my machine

---
 example_projects/benchmark_results.txt        | 18 +++++
 .../gowin_benchmark/src/dynamic_delay.sv      | 77 +++++++++++++------
 example_projects/gowin_benchmark/src/main.sv  |  1 +
 .../quartus_benchmark/dynamic_delay.sv        | 77 +++++++++++++------
 example_projects/quartus_benchmark/main.sv    |  1 +
 .../sources_1/dynamic_delay.sv                | 77 +++++++++++++------
 .../vivado_benchmark.srcs/sources_1/main.sv   |  1 +
 7 files changed, 177 insertions(+), 75 deletions(-)
 create mode 100644 example_projects/benchmark_results.txt

diff --git a/example_projects/benchmark_results.txt b/example_projects/benchmark_results.txt
new file mode 100644
index 0000000..55fb98a
--- /dev/null
+++ b/example_projects/benchmark_results.txt
@@ -0,0 +1,18 @@
+
+
+Compilation time results for reference benchmark projects
+=========================================================
+
+
+Xeon E5-2630 v4, RAM 32GB, Windows 7
+------------------------------------
+quartus_benchmark - 4m 58s  ( Quartus Lite 17 )
+vivado_benchmark  - 5m 58s  ( Vivado 2019.2 )
+gowin_benchmark   - 4m 15s  ( Gowin_V1.9.6Beta )
+ise_benchmark     - 9m 10s  ( ISE 12.4 )
+
+
+Xeon E5-2630 v4, RAM 32GB, Windows 7, project files on RamDisk
+--------------------------------------------------------------
+quartus_benchmark - 4m 57s
+vivado_benchmark  - 5m 56s
\ No newline at end of file
diff --git a/example_projects/gowin_benchmark/src/dynamic_delay.sv b/example_projects/gowin_benchmark/src/dynamic_delay.sv
index 4c458ef..302fe70 100644
--- a/example_projects/gowin_benchmark/src/dynamic_delay.sv
+++ b/example_projects/gowin_benchmark/src/dynamic_delay.sv
@@ -4,55 +4,82 @@
 //--------------------------------------------------------------------------------
 
 // INFO --------------------------------------------------------------------------------
-//  Dynamic delay for arbitrary signal
+//  Dynamic delay for arbitrary signal.
 //
-//  CAUTION: The module intentionally does NOT implement error handling when
-//           LENGTH is not a multiple of 2. Please handle "out of range"
-//           checks externally.
+//  Incoming data elements have WIDTH bits each. Module does serialization of
+//  input data and outputs flattened bits, based on provided selector value.
+//  You can perform delays bit-wize, not just element-wize.
+//
+//  CAUTION: Be careful selecting last, most-delayed "WIDTH" number of bits.
+//           The module intentionally does NOT implement "out of range"
+//           checks. Please handle them externally.
+
 
 
 /* --- INSTANTIATION TEMPLATE BEGIN ---
 
 dynamic_delay #(
-  .LENGTH( 8 )
-  //.SEL_W( 3 )
-) DD1 (
+  .LENGTH( 3 ),
+  .WIDTH( 4 )
+) M (
   .clk( clk ),
-  .nrst( 1'b1 ),
+  .nrst( nrst ),
   .ena( 1'b1 ),
-  .in(  ),
-  .sel(  ),
-  .out(  )
+  .in( in_data[3:0] ),
+  .sel( sel[3:0] ),
+  .out( out_data[3:0] )
 );
 
 --- INSTANTIATION TEMPLATE END ---*/
 
 
 module dynamic_delay #( parameter
-  LENGTH = 8,               // maximum delay chain width
-  SEL_W = $clog2(LENGTH)    // output selector width
+  LENGTH = 63,                        // maximum delay chain length
+  WIDTH = 4,                          // data width
+
+  SEL_W = $clog2( (LENGTH+1)*WIDTH )  // output selector width
+                                      //  plus one is for zero delay element
 )(
   input clk,
   input nrst,
   input ena,
-  input in,
-  input [SEL_W-1:0] sel,     // output selector
-  output logic out
+  input [WIDTH-1:0] in,               // input data
+                                      // bit in[0] is the "oldest" one
+                                      // bit in[WIDTH] is considered the most recent
+  input [SEL_W-1:0] sel,              // output selector
+  output logic [WIDTH-1:0] out        // output data
 );
 
-logic [(LENGTH-1):0] data = 0;
+
+
+logic [(LENGTH+1)-1:0][WIDTH-1:0] data = '0;
+
+// packed vector includes extra bits
+logic [(LENGTH+1)*WIDTH-1:0] pack_data;
+assign pack_data[(LENGTH+1)*WIDTH-1:0] = data;
 
 integer i;
 always_ff @(posedge clk) begin
-  if (~nrst) begin
-    data[(LENGTH-1):0] <= 0;
-    out <= 0;
-  end else if (ena) begin
-    data[0] <= in;
-    for (i=1; i<LENGTH; i=i+1) begin
-      data[i] <= data[i-1];
+  if( ~nrst ) begin
+    // reset all data except zero element
+    for( i=1; i<(LENGTH+1); i=i+1 ) begin
+      data[i][WIDTH-1:0] <= '0;
     end
-    out <= data[sel[SEL_W-1:0]];
+  end else if (ena) begin
+    for( i=1; i<(LENGTH+1); i=i+1 ) begin
+      data[i][WIDTH-1:0] <= data[i-1][WIDTH-1:0];
+    end
+  end
+end
+
+integer j;
+always_comb begin
+  // zero element assignment
+  data[0][WIDTH-1:0] <= in[WIDTH-1:0];
+
+  // output selector, sel==0 gives non-delayed output
+  for( j=0; j<WIDTH; j=j+1 ) begin
+    out[j] <= pack_data[sel[SEL_W-1:0]+j];
   end
 end
 
diff --git a/example_projects/gowin_benchmark/src/main.sv b/example_projects/gowin_benchmark/src/main.sv
index 9b5430b..cae5bab 100644
--- a/example_projects/gowin_benchmark/src/main.sv
+++ b/example_projects/gowin_benchmark/src/main.sv
@@ -28,6 +28,7 @@ module main(
 
 dynamic_delay #(
   .LENGTH( `LENGTH ),
+  .WIDTH( 1 ),
   .SEL_W( `SEL_W )
 ) dd [`WIDTH-1:0] (
   .clk( {`WIDTH{clk}} ),
diff --git a/example_projects/quartus_benchmark/dynamic_delay.sv b/example_projects/quartus_benchmark/dynamic_delay.sv
index 4c458ef..302fe70 100644
--- a/example_projects/quartus_benchmark/dynamic_delay.sv
+++ b/example_projects/quartus_benchmark/dynamic_delay.sv
@@ -4,55 +4,82 @@
 //--------------------------------------------------------------------------------
 
 // INFO --------------------------------------------------------------------------------
-//  Dynamic delay for arbitrary signal
+//  Dynamic delay for arbitrary signal.
 //
-//  CAUTION: The module intentionally does NOT implement error handling when
-//           LENGTH is not a multiple of 2. Please handle "out of range"
-//           checks externally.
+//  Incoming data elements have WIDTH bits each. Module does serialization of
+//  input data and outputs flattened bits, based on provided selector value.
+//  You can perform delays bit-wize, not just element-wize.
+//
+//  CAUTION: Be careful selecting last, most-delayed "WIDTH" number of bits.
+//           The module intentionally does NOT implement "out of range"
+//           checks. Please handle them externally.
+
 
 
 /* --- INSTANTIATION TEMPLATE BEGIN ---
 
 dynamic_delay #(
-  .LENGTH( 8 )
-  //.SEL_W( 3 )
-) DD1 (
+  .LENGTH( 3 ),
+  .WIDTH( 4 )
+) M (
   .clk( clk ),
-  .nrst( 1'b1 ),
+  .nrst( nrst ),
   .ena( 1'b1 ),
-  .in(  ),
-  .sel(  ),
-  .out(  )
+  .in( in_data[3:0] ),
+  .sel( sel[3:0] ),
+  .out( out_data[3:0] )
 );
 
 --- INSTANTIATION TEMPLATE END ---*/
 
 
 module dynamic_delay #( parameter
-  LENGTH = 8,               // maximum delay chain width
-  SEL_W = $clog2(LENGTH)    // output selector width
+  LENGTH = 63,                        // maximum delay chain length
+  WIDTH = 4,                          // data width
+
+  SEL_W = $clog2( (LENGTH+1)*WIDTH )  // output selector width
+                                      //  plus one is for zero delay element
 )(
   input clk,
   input nrst,
   input ena,
-  input in,
-  input [SEL_W-1:0] sel,     // output selector
-  output logic out
+  input [WIDTH-1:0] in,               // input data
+                                      // bit in[0] is the "oldest" one
+                                      // bit in[WIDTH] is considered the most recent
+  input [SEL_W-1:0] sel,              // output selector
+  output logic [WIDTH-1:0] out        // output data
 );
 
-logic [(LENGTH-1):0] data = 0;
+
+
+logic [(LENGTH+1)-1:0][WIDTH-1:0] data = '0;
+
+// packed vector includes extra bits
+logic [(LENGTH+1)*WIDTH-1:0] pack_data;
+assign pack_data[(LENGTH+1)*WIDTH-1:0] = data;
 
 integer i;
 always_ff @(posedge clk) begin
-  if (~nrst) begin
-    data[(LENGTH-1):0] <= 0;
-    out <= 0;
-  end else if (ena) begin
-    data[0] <= in;
-    for (i=1; i<LENGTH; i=i+1) begin
-      data[i] <= data[i-1];
+  if( ~nrst ) begin
+    // reset all data except zero element
+    for( i=1; i<(LENGTH+1); i=i+1 ) begin
+      data[i][WIDTH-1:0] <= '0;
     end
-    out <= data[sel[SEL_W-1:0]];
+  end else if (ena) begin
+    for( i=1; i<(LENGTH+1); i=i+1 ) begin
+      data[i][WIDTH-1:0] <= data[i-1][WIDTH-1:0];
+    end
+  end
+end
+
+integer j;
+always_comb begin
+  // zero element assignment
+  data[0][WIDTH-1:0] <= in[WIDTH-1:0];
+
+  // output selector, sel==0 gives non-delayed output
+  for( j=0; j<WIDTH; j=j+1 ) begin
+    out[j] <= pack_data[sel[SEL_W-1:0]+j];
   end
 end
 
diff --git a/example_projects/quartus_benchmark/main.sv b/example_projects/quartus_benchmark/main.sv
index ef9fc4c..a162823 100755
--- a/example_projects/quartus_benchmark/main.sv
+++ b/example_projects/quartus_benchmark/main.sv
@@ -28,6 +28,7 @@ module main(
 
 dynamic_delay #(
   .LENGTH( `LENGTH ),
+  .WIDTH( 1 ),
   .SEL_W( `SEL_W )
 ) dd [`WIDTH-1:0] (
   .clk( {`WIDTH{clk}} ),
diff --git a/example_projects/vivado_benchmark/vivado_benchmark.srcs/sources_1/dynamic_delay.sv b/example_projects/vivado_benchmark/vivado_benchmark.srcs/sources_1/dynamic_delay.sv
index 4c458ef..302fe70 100644
--- a/example_projects/vivado_benchmark/vivado_benchmark.srcs/sources_1/dynamic_delay.sv
+++ b/example_projects/vivado_benchmark/vivado_benchmark.srcs/sources_1/dynamic_delay.sv
@@ -4,55 +4,82 @@
 //--------------------------------------------------------------------------------
 
 // INFO --------------------------------------------------------------------------------
-//  Dynamic delay for arbitrary signal
+//  Dynamic delay for arbitrary signal.
 //
-//  CAUTION: The module intentionally does NOT implement error handling when
-//           LENGTH is not a multiple of 2. Please handle "out of range"
-//           checks externally.
+//  Incoming data elements have WIDTH bits each. Module does serialization of
+//  input data and outputs flattened bits, based on provided selector value.
+//  You can perform delays bit-wize, not just element-wize.
+//
+//  CAUTION: Be careful selecting last, most-delayed "WIDTH" number of bits.
+//           The module intentionally does NOT implement "out of range"
+//           checks. Please handle them externally.
+
 
 
 /* --- INSTANTIATION TEMPLATE BEGIN ---
 
 dynamic_delay #(
-  .LENGTH( 8 )
-  //.SEL_W( 3 )
-) DD1 (
+  .LENGTH( 3 ),
+  .WIDTH( 4 )
+) M (
   .clk( clk ),
-  .nrst( 1'b1 ),
+  .nrst( nrst ),
   .ena( 1'b1 ),
-  .in(  ),
-  .sel(  ),
-  .out(  )
+  .in( in_data[3:0] ),
+  .sel( sel[3:0] ),
+  .out( out_data[3:0] )
 );
 
 --- INSTANTIATION TEMPLATE END ---*/
 
 
 module dynamic_delay #( parameter
-  LENGTH = 8,               // maximum delay chain width
-  SEL_W = $clog2(LENGTH)    // output selector width
+  LENGTH = 63,                        // maximum delay chain length
+  WIDTH = 4,                          // data width
+
+  SEL_W = $clog2( (LENGTH+1)*WIDTH )  // output selector width
+                                      //  plus one is for zero delay element
 )(
   input clk,
   input nrst,
   input ena,
-  input in,
-  input [SEL_W-1:0] sel,     // output selector
-  output logic out
+  input [WIDTH-1:0] in,               // input data
+                                      // bit in[0] is the "oldest" one
+                                      // bit in[WIDTH] is considered the most recent
+  input [SEL_W-1:0] sel,              // output selector
+  output logic [WIDTH-1:0] out        // output data
 );
 
-logic [(LENGTH-1):0] data = 0;
+
+
+logic [(LENGTH+1)-1:0][WIDTH-1:0] data = '0;
+
+// packed vector includes extra bits
+logic [(LENGTH+1)*WIDTH-1:0] pack_data;
+assign pack_data[(LENGTH+1)*WIDTH-1:0] = data;
 
 integer i;
 always_ff @(posedge clk) begin
-  if (~nrst) begin
-    data[(LENGTH-1):0] <= 0;
-    out <= 0;
-  end else if (ena) begin
-    data[0] <= in;
-    for (i=1; i<LENGTH; i=i+1) begin
-      data[i] <= data[i-1];
+  if( ~nrst ) begin
+    // reset all data except zero element
+    for( i=1; i<(LENGTH+1); i=i+1 ) begin
+      data[i][WIDTH-1:0] <= '0;
     end
-    out <= data[sel[SEL_W-1:0]];
+  end else if (ena) begin
+    for( i=1; i<(LENGTH+1); i=i+1 ) begin
+      data[i][WIDTH-1:0] <= data[i-1][WIDTH-1:0];
+    end
+  end
+end
+
+integer j;
+always_comb begin
+  // zero element assignment
+  data[0][WIDTH-1:0] <= in[WIDTH-1:0];
+
+  // output selector, sel==0 gives non-delayed output
+  for( j=0; j<WIDTH; j=j+1 ) begin
+    out[j] <= pack_data[sel[SEL_W-1:0]+j];
   end
 end
 
diff --git a/example_projects/vivado_benchmark/vivado_benchmark.srcs/sources_1/main.sv b/example_projects/vivado_benchmark/vivado_benchmark.srcs/sources_1/main.sv
index 9b5430b..cae5bab 100755
--- a/example_projects/vivado_benchmark/vivado_benchmark.srcs/sources_1/main.sv
+++ b/example_projects/vivado_benchmark/vivado_benchmark.srcs/sources_1/main.sv
@@ -28,6 +28,7 @@ module main(
 
 dynamic_delay #(
   .LENGTH( `LENGTH ),
+  .WIDTH( 1 ),
   .SEL_W( `SEL_W )
 ) dd [`WIDTH-1:0] (
   .clk( {`WIDTH{clk}} ),