cmprs_afi_mux_ptr.v 11.3 KB
Newer Older
1 2 3 4 5 6 7
/*!
 * <b>Module:</b>cmprs_afi_mux_ptr
 * @file cmprs_afi_mux_ptr.v
 * @date 2015-06-28  
 * @author Andrey Filippov     
 *
 * @brief Maintain 4-channel chunk pointers (before AXI)
8 9
 * Advance 32-byte chunk pointers for each AXI burst and each frame (4*2=8 pointers)
 *
10 11 12 13
 * @copyright Copyright (c) 2015 Elphel, Inc.
 *
 * <b>License:</b>
 *
14 15 16 17 18 19 20 21 22 23 24 25
 * cmprs_afi_mux_ptr.v is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation, either version 3 of the License, or
 * (at your option) any later version.
 *
 *  cmprs_afi_mux_ptr.v is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program.  If not, see <http://www.gnu.org/licenses/> .
26 27 28 29 30 31
 *
 * Additional permission under GNU GPL version 3 section 7:
 * If you modify this Program, or any covered work, by linking or combining it
 * with independent modules provided by the FPGA vendor only (this permission
 * does not extend to any 3-rd party modules, "soft cores" or macros) under
 * different license terms solely for the purpose of generating binary "bitstream"
32
 * files and/or simulating the code, the copyright holders of this Program give
33 34
 * you the right to distribute the covered work without those independent modules
 * as long as the source code for them is available from the FPGA vendor free of
Andrey Filippov's avatar
Andrey Filippov committed
35
 * charge, and there is no dependence on any encrypted modules for simulating of
36 37 38
 * the combined code. This permission applies to you if the distributed code
 * contains all the components and scripts required to completely simulate it
 * with at least one of the Free Software programs.
39
 */
40 41 42 43 44 45 46 47 48
`timescale 1ns/1ps

module  cmprs_afi_mux_ptr(
    input                         hclk,               // global clock to run axi_hp @ 150MHz, shared by all compressor channels
    // Write dual port 8x27 channel start address/length RAM
    input                  [26:0] sa_len_di,          // data to write per-channle buffer sa/length in chunks
    input                  [ 2:0] sa_len_wa,          // channel address to write sa/lengths
    input                         sa_len_we,          // write enable sa/length data
    
49
    input                         en,                 // 0 - resets, 0->1 resets all pointers. While reset allows write response
50 51
    input                  [ 3:0] reset_pointers,     // per-channel - reset pointers
    input                         pre_busy_w,         // combinatorial signal - one before busy[0] (depends on ptr_resetting)
52 53
    input                  [ 1:0] pre_winner_channel, // channel that won arbitration for AXI access, valid 1 cycle ahead of @ pre_busy_w
//    input                  [ 1:0] winner_channel,     // channel that won arbitration for AXI access, valid @ pre_busy_w
54
    input                         need_to_bother,     // wants to start access if address and data FIFO permit
55
    input                   [1:0] chunk_inc_want_m1,  // how much to increment chunk pointer (0..3) +1 - valid with busy[0] (w/o rollover)
56 57
    
    input                         last_burst_in_frame, // valid with busy[0] (last_burst_in_frame<=last_chunk_w[winner2])
58
                                                       // last_burst_in_frame is invalid when rollover splits burst in 2
59 60 61 62
    input                  [ 3:0] busy,               // one cycle less than sending 1-4 bursts, [1] - delayed by 1, [2] - by 2 
    output                        ptr_resetting,      // pointers will be reset next cycle (2-cycle-long pulse)
    output reg             [26:0] chunk_addr,         // chunk absolute address, valid with busy[1]
    input                  [ 2:0] chunk_ptr_ra,       // chunk pointer read address {eof, chn[1:0]}
63 64 65 66
    output                 [26:0] chunk_ptr_rd,       // chunk pointer read data (non-registered
//    output                 [ 2:0] max_inc             // maximal increment to rollover (limited by 4)        
    output                 [ 2:0] max_wlen            // maximal wlen[3:2], MSB - limited by rollover        

67 68 69 70 71 72 73 74 75 76
);
    reg   [3:0] reset_rq;               // request to reset pointers when ready
    reg   [3:0] reset_rq_pri;           // one-hot reset rq 
    wire  [1:0] reset_rq_enc;           // encoded reset_rq_pri
    wire        start_resetting_w;
    reg   [1:0] resetting;              // resetting chunk_pointer and eof_pointer
    reg   [2:0] ptr_wa;                 // pointer memory write port address, msb - eof/current, 2 LSB - channel
    reg         ptr_we;                 // pointer memory write enable
    reg  [26:0] ptr_ram[0:7];           // pointer (current and eof) memory (in 32-byte chunks
    wire [26:0] ptr_ram_di;             // data to be written to ptr_ram
77
    reg  [26:0] sa_len_ram[0:7];        // start chunk/num cunks in a buffer (write port @mclk)
78
    reg  [26:0] chunk_ptr_inc;          // incremented by 1..4 chunk pointer
79
//    reg  [27:0] rollover_r;         // incremented chunk pointer, decremented by length (MSB - sign)
80
    reg         en_d;                   //enable delayed by 1 cycle
81
    wire [ 2:0] sa_len_ra;              // start/len read address (0..3 - start addresses, 4..7 - lengths)
82

83 84 85 86 87 88
    reg  [ 2:0] max_inc_ram[0:3];       // maximal increment to rollover (limited by 4)     
    wire [ 1:0] pre_chunk_inc_m1;
    reg  [ 2:0] chunk_inc;
    wire [26:0] chunks_to_rollover;
    
    reg   [3:0] chunks_to_rollover_r;   // [3] >=8 
89
// SuppressWarnings VEditor unused 
90
    wire  [3:0] chunks_to_rollover_m1;
91
    wire  [3:0] pre_chunks_to_rollover_m1;
92 93 94 95
    reg         max_inc_ram_we;
    reg   [1:0] max_inc_ram_wa;
    wire        rollover_w; // this cycle causes rollover - valid at pre_busy_w
    reg         rollover_r; // this cycle causes rollover - valid at busy[0] and late
96 97
    reg  [ 1:0] winner_channel;     // channel that won arbitration for AXI access, valid @ pre_busy_w
    
98 99
    
    wire        ptr_ram_wa = ptr_ram[ptr_wa]; // SuppressThisWarning VEditor debug - just to view 
100 101 102 103 104
//    wire  [2:0] max_wlen_di;   // data to write to max_inc_ram and bypass register
    reg   [2:0] max_wlen_r;    // memory registered output
    reg   [2:0] max_wlen_same; // used to bypass max_inc_ram for the same channel
//    reg         use_same_max_wlen; // valid @ pre_busy_w
    reg   [1:0] last_max_written; // channel for which max_wlen was written to RAM
105 106 107 108 109 110
    
    assign ptr_resetting = resetting[0];
    assign sa_len_ra= {busy[1],ptr_wa[1:0]};
    
    assign reset_rq_enc = {reset_rq_pri[3] | reset_rq_pri[2],
                           reset_rq_pri[3] | reset_rq_pri[1]};
111
    assign ptr_ram_di= (resetting[1] ||rollover_r)  ? 27'b0 : chunk_ptr_inc ;
112 113 114

    assign chunk_ptr_rd = ptr_ram[chunk_ptr_ra];
    assign start_resetting_w = en && !busy[0] && !resetting[0] && (|reset_rq) && !need_to_bother;
115
    
116 117 118
///    assign max_wlen = max_inc_ram[winner_channel]; // valid @pre_busy_w
//    assign max_wlen = (last_max_written == winner_channel) ? (max_inc_ram_we? max_wlen_di: max_wlen_same) :max_wlen_r ; // valid @pre_busy_w
    assign max_wlen = (last_max_written == winner_channel) ? max_wlen_same :max_wlen_r ; // valid @pre_busy_w
119 120 121 122 123 124 125
    
//chunk_inc_want_m1
    assign pre_chunk_inc_m1 = (max_wlen[1:0] >= chunk_inc_want_m1)? chunk_inc_want_m1 : max_wlen[1:0];
    assign rollover_w =       !max_wlen[2] && (max_wlen[1:0] <= chunk_inc_want_m1);
    
    assign chunks_to_rollover = sa_len_ram[sa_len_ra] -  ptr_ram_di;
    assign chunks_to_rollover_m1 = chunks_to_rollover_r -1;
126 127 128
//    assign max_wlen_di = (|chunks_to_rollover_m1[3:2])?3'h7:{1'b0,chunks_to_rollover_m1[1:0]};    
    // 1 cycle ahead of chunks_to_rollover_m1
    assign pre_chunks_to_rollover_m1 =  {|chunks_to_rollover[26:3],chunks_to_rollover[2:0]} - 1;  
129
    always @ (posedge hclk) begin
130 131 132 133 134
        winner_channel <= pre_winner_channel;
        max_wlen_r <= max_inc_ram[pre_winner_channel]; // valid @pre_busy_w
        if       (!en)                 last_max_written <= 0;
        else if  (ptr_we & ~ptr_wa[2]) last_max_written <= ptr_wa[1:0];
        
135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164
        en_d <= en;
        // ===== calculate and rollover channel addresses ====
        // clear (during "resetting" or update 8x27 RAM that holds chunk pointers for the current burst and currenty frame
        // uses (reads) additional memory with per-channel start addresses and lengths (both measured in 32-byte chunks
        // read port of the ptr_ram will be used to monitor pointers through the status network
        // TODO: Make sure correct RAM modules are inferred
        // write length RAM
        if (sa_len_we) sa_len_ram[sa_len_wa] <= sa_len_di;
        if (start_resetting_w)  reset_rq_pri <= {reset_rq[3] & ~(|reset_rq[2:0]),
                                                 reset_rq[2] & ~(|reset_rq[1:0]),
                                                 reset_rq[1] &     ~reset_rq[0],
                                                 reset_rq[0]};
                                                 
        if (en && !en_d) reset_rq <= 4'hf;
        else             reset_rq <= reset_pointers | (reset_rq  & ~({4{resetting[0] &~ resetting[1]}} & reset_rq_pri));
        
        if (!en) resetting <= 0;
        else     resetting <= {resetting[0], start_resetting_w | (resetting[0] & ~resetting[1])};
        
        if (resetting == 2'b1)               ptr_wa[1:0] <= reset_rq_enc;
        else if (pre_busy_w)                 ptr_wa[1:0] <= winner_channel;
        
        if (start_resetting_w || pre_busy_w) ptr_wa[2]  <= 0;
        else if (ptr_we)                     ptr_wa[2]  <= 1;
        
        ptr_we <= resetting[0] ||   //  a pair of cycles to reset chunk pointer and frame chunk pointer
                  (busy[1] && !busy[2]) || // always update chunk pointer
                  (busy[2] && !busy[3] && last_burst_in_frame); // optionally update frame chunk pointer (same value)

        if (busy[0] && !busy[1]) begin // first clock of busy
165
            // calculate full address for current AXI burst (valid 1 clk after busy)
166 167 168
            chunk_addr <= ptr_ram[ptr_wa] + sa_len_ram[sa_len_ra];
            chunk_ptr_inc <= ptr_ram[ptr_wa] + chunk_inc;
        end                  
169 170 171
//        if (busy[1] && !busy[2]) begin // first clock of busy
//            chunk_ptr_rovr <={1'b0,chunk_ptr_inc} - {1'b0,sa_len_ram[sa_len_ra]}; // sa_len_ra now points at length
//        end
172 173
        // write to ptr_ram (1 or 2 locations - if eof)
        if (ptr_we) ptr_ram[ptr_wa] <= ptr_ram_di;
174 175 176 177 178 179 180 181 182 183 184 185
        
        if (pre_busy_w) chunk_inc <= {1'b0, pre_chunk_inc_m1} + 1;
        
        if (pre_busy_w) rollover_r <= rollover_w;
        
//    wire [26:0] chunks_to_rollover;
    
//    reg   [3:0] chunks_to_rollover_r;   // [3] >=8 
        if (ptr_we) chunks_to_rollover_r <= {|chunks_to_rollover[26:3],chunks_to_rollover[2:0]};
        
        max_inc_ram_we <= ptr_we & ~ptr_wa[2];
        max_inc_ram_wa <= ptr_wa[1:0];
186 187 188 189 190 191 192 193 194 195
        
        // set 1 cycle earlier 
        if (ptr_we & ~ptr_wa[2]) max_wlen_same <= (|pre_chunks_to_rollover_m1[3:2])?3'h7:{1'b0,pre_chunks_to_rollover_m1[1:0]};
        
        if (max_inc_ram_we) begin
            max_inc_ram[max_inc_ram_wa] <= max_wlen_same; // max_wlen_di; //  (|chunks_to_rollover_m1[3:2])?3'h7:{1'b0,chunks_to_rollover_m1[1:0]};
//            max_wlen_same <=               max_wlen_di;
        end
        
        
196 197 198 199
    end
endmodule