Subversion Repositories openarty

////////////////////////////////////////////////////////////////////////////////

//

// Filename:    zipcpu.v

//

// Project:     Zip CPU -- a small, lightweight, RISC CPU soft core

//

// Purpose:     This is the top level module holding the core of the Zip CPU

//              together.  The Zip CPU is designed to be as simple as possible.

//      (actual implementation aside ...)  The instruction set is about as

//      RISC as you can get, with only 26 instruction types currently supported.

//      (There are still 8-instruction Op-Codes reserved for floating point,

//      and 5 which can be used for transactions not requiring registers.)

//      Please see the accompanying spec.pdf file for a description of these

//      instructions.

//

//      All instructions are 32-bits wide.  All bus accesses, both address and

//      data, are 32-bits over a wishbone bus.

//

//      The Zip CPU is fully pipelined with the following pipeline stages:

//

//              1. Prefetch, returns the instruction from memory.

//

//              2. Instruction Decode

//

//              3. Read Operands

//

//              4. Apply Instruction

//

//              4. Write-back Results

//

//      Further information about the inner workings of this CPU, such as

//      what causes pipeline stalls, may be found in the spec.pdf file.  (The

//      documentation within this file had become out of date and out of sync

//      with the spec.pdf, so look to the spec.pdf for accurate and up to date

//      information.)

//

//

//      In general, the pipelining is controlled by three pieces of logic

//      per stage: _ce, _stall, and _valid.  _valid means that the stage

//      holds a valid instruction.  _ce means that the instruction from the

//      previous stage is to move into this one, and _stall means that the

//      instruction from the previous stage may not move into this one.

//      The difference between these control signals allows individual stages

//      to propagate instructions independently.  In general, the logic works

//      as:

//

//

//      assign  (n)_ce = (n-1)_valid && (!(n)_stall)

//

//

//      always @(posedge i_clk)

//              if ((i_rst)||(clear_pipeline))

//                      (n)_valid = 0

//              else if (n)_ce

//                      (n)_valid = 1

//              else if (n+1)_ce

//                      (n)_valid = 0

//

//      assign (n)_stall = (  (n-1)_valid && ( pipeline hazard detection )  )

//                      || (  (n)_valid && (n+1)_stall );

//

//      and ...

//

//      always @(posedge i_clk)

//              if (n)_ce

//                      (n)_variable = ... whatever logic for this stage

//

//      Note that a stage can stall even if no instruction is loaded into

//      it.

//

//

// Creator:     Dan Gisselquist, Ph.D.

//              Gisselquist Technology, LLC

//

////////////////////////////////////////////////////////////////////////////////

//

// Copyright (C) 2015-2017, Gisselquist Technology, LLC

//

// This program is free software (firmware): 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.

//

// This program is distributed in the hope that it will be useful, but WITHOUT

// ANY WARRANTY; without even the implied warranty of MERCHANTIBILITY 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.  (It's in the $(ROOT)/doc directory.  Run make with no

// target there if the PDF file isn't present.)  If not, see

// <http://www.gnu.org/licenses/> for a copy.

//

// License:     GPL, v3, as defined and found on www.gnu.org,

//              http://www.gnu.org/licenses/gpl.html

//

//

////////////////////////////////////////////////////////////////////////////////

//

//

//

`define CPU_CC_REG      4'he

`define CPU_PC_REG      4'hf

`define CPU_CLRCACHE_BIT 14     // Set to clear the I-cache, automatically clears

`define CPU_PHASE_BIT   13      // Set if we are executing the latter half of a CIS

`define CPU_FPUERR_BIT  12      // Floating point error flag, set on error

`define CPU_DIVERR_BIT  11      // Divide error flag, set on divide by zero

`define CPU_BUSERR_BIT  10      // Bus error flag, set on error

`define CPU_TRAP_BIT    9       // User TRAP has taken place

`define CPU_ILL_BIT     8       // Illegal instruction

`define CPU_BREAK_BIT   7

`define CPU_STEP_BIT    6       // Will step one (or two CIS) instructions

`define CPU_GIE_BIT     5

`define CPU_SLEEP_BIT   4

// Compile time defines

//

`include "cpudefs.v"

//

//

module  zipcpu(i_clk, i_rst, i_interrupt,

                // Debug interface

                i_halt, i_clear_pf_cache, i_dbg_reg, i_dbg_we, i_dbg_data,

                        o_dbg_stall, o_dbg_reg, o_dbg_cc,

                        o_break,

                // CPU interface to the wishbone bus

                o_wb_gbl_cyc, o_wb_gbl_stb,

                        o_wb_lcl_cyc, o_wb_lcl_stb,

                        o_wb_we, o_wb_addr, o_wb_data, o_wb_sel,

                        i_wb_ack, i_wb_stall, i_wb_data,

                        i_wb_err,

                // Accounting/CPU usage interface

                o_op_stall, o_pf_stall, o_i_count

`ifdef  DEBUG_SCOPE

                , o_debug

`endif

                );

        parameter [31:0] RESET_ADDRESS=32'h0100000;

        parameter       ADDRESS_WIDTH=30,

                        LGICACHE=8;

`ifdef  OPT_MULTIPLY

        parameter       IMPLEMENT_MPY = `OPT_MULTIPLY;

`else

        parameter       IMPLEMENT_MPY = 0;

`endif

`ifdef  OPT_DIVIDE

        parameter       IMPLEMENT_DIVIDE = 1;

`else

        parameter       IMPLEMENT_DIVIDE = 0;

`endif

`ifdef  OPT_IMPLEMENT_FPU

        parameter       IMPLEMENT_FPU = 1,

`else

        parameter       IMPLEMENT_FPU = 0,

`endif

                        IMPLEMENT_LOCK=1;

`ifdef  OPT_EARLY_BRANCHING

        parameter       EARLY_BRANCHING = 1;

`else

        parameter       EARLY_BRANCHING = 0;

`endif

        parameter       WITH_LOCAL_BUS = 1;

        localparam      AW=ADDRESS_WIDTH;

        localparam      [(AW-1):0]       RESET_BUS_ADDRESS = RESET_ADDRESS[(AW+1):2];

        input                   i_clk, i_rst, i_interrupt;

        // Debug interface -- inputs

        input                   i_halt, i_clear_pf_cache;

        input           [4:0]    i_dbg_reg;

        input                   i_dbg_we;

        input           [31:0]   i_dbg_data;

        // Debug interface -- outputs

        output  wire            o_dbg_stall;

        output  reg     [31:0]   o_dbg_reg;

        output  reg     [3:0]    o_dbg_cc;

        output  wire            o_break;

        // Wishbone interface -- outputs

        output  wire            o_wb_gbl_cyc, o_wb_gbl_stb;

        output  wire            o_wb_lcl_cyc, o_wb_lcl_stb, o_wb_we;

        output  wire    [(AW-1):0]       o_wb_addr;

        output  wire    [31:0]   o_wb_data;

        output  wire    [3:0]    o_wb_sel;

        // Wishbone interface -- inputs

        input                   i_wb_ack, i_wb_stall;

        input           [31:0]   i_wb_data;

        input                   i_wb_err;

        // Accounting outputs ... to help us count stalls and usage

        output  wire            o_op_stall;

        output  wire            o_pf_stall;

        output  wire            o_i_count;

        //

`ifdef  DEBUG_SCOPE

        output  reg     [31:0]   o_debug;

`endif

        // Registers

        //

        //      The distributed RAM style comment is necessary on the

        // SPARTAN6 with XST to prevent XST from oversimplifying the register

        // set and in the process ruining everything else.  It basically

        // optimizes logic away, to where it no longer works.  The logic

        // as described herein will work, this just makes sure XST implements

        // that logic.

        //

        (* ram_style = "distributed" *)

`ifdef  OPT_NO_USERMODE

        reg     [31:0]   regset [0:15];

`else

        reg     [31:0]   regset [0:31];

`endif

        // Condition codes

        // (BUS, TRAP,ILL,BREAKEN,STEP,GIE,SLEEP ), V, N, C, Z

        reg     [3:0]    flags, iflags;

        wire    [14:0]   w_uflags, w_iflags;

        reg             break_en, step, sleep, r_halted;

        wire            break_pending, trap, gie, ubreak;

        wire            w_clear_icache, ill_err_u;

        reg             ill_err_i;

        reg             ibus_err_flag;

        wire            ubus_err_flag;

        wire            idiv_err_flag, udiv_err_flag;

        wire            ifpu_err_flag, ufpu_err_flag;

        wire            ihalt_phase, uhalt_phase;

        // The master chip enable

        wire            master_ce;

        //

        //

        //      PIPELINE STAGE #1 :: Prefetch

        //              Variable declarations

        //

        reg     [(AW+1):0]       pf_pc;

        reg     new_pc;

        wire    clear_pipeline;

        assign  clear_pipeline = new_pc;

        wire            dcd_stalled;

        wire            pf_cyc, pf_stb, pf_we, pf_busy, pf_ack, pf_stall, pf_err;

        wire    [(AW-1):0]       pf_addr;

        wire    [31:0]           pf_data;

        wire    [31:0]           pf_instruction;

        wire    [(AW-1):0]       pf_instruction_pc;

        wire    pf_valid, pf_gie, pf_illegal;

        //

        //

        //      PIPELINE STAGE #2 :: Instruction Decode

        //              Variable declarations

        //

        //

        reg             op_valid /* verilator public_flat */,

                        op_valid_mem, op_valid_alu;

        reg             op_valid_div, op_valid_fpu;

        wire            op_stall, dcd_ce, dcd_phase;

        wire    [3:0]    dcd_opn;

        wire    [4:0]    dcd_A, dcd_B, dcd_R;

        wire            dcd_Acc, dcd_Bcc, dcd_Apc, dcd_Bpc, dcd_Rcc, dcd_Rpc;

        wire    [3:0]    dcd_F;

        wire            dcd_wR, dcd_rA, dcd_rB,

                                dcd_ALU, dcd_M, dcd_DIV, dcd_FP,

                                dcd_wF, dcd_gie, dcd_break, dcd_lock,

                                dcd_pipe, dcd_ljmp;

        wire            dcd_valid;

        wire    [AW:0]   dcd_pc /* verilator public_flat */;

        wire    [31:0]   dcd_I;

        wire            dcd_zI; // true if dcd_I == 0

        wire    dcd_A_stall, dcd_B_stall, dcd_F_stall;

        wire    dcd_illegal;

        wire                    dcd_early_branch;

        wire    [(AW-1):0]       dcd_branch_pc;

        wire            dcd_sim;

        wire    [22:0]   dcd_sim_immv;

        //

        //

        //      PIPELINE STAGE #3 :: Read Operands

        //              Variable declarations

        //

        //

        //

        // Now, let's read our operands

        reg     [4:0]    alu_reg;

        wire    [3:0]    op_opn;

        wire    [4:0]    op_R;

        reg     [31:0]   r_op_Av, r_op_Bv;

        reg     [(AW-1):0]       op_pc;

        wire    [31:0]   w_op_Av, w_op_Bv;

        wire    [31:0]   op_A_nowait, op_B_nowait, op_Av, op_Bv;

        reg             op_wR, op_wF;

        wire            op_gie, op_Rcc;

        wire    [14:0]   op_Fl;

        reg     [6:0]    r_op_F;

        wire    [7:0]    op_F;

        wire            op_ce, op_phase, op_pipe, op_change_data_ce;

        // Some pipeline control wires

        reg     op_illegal;

        wire    op_break;

        wire    op_lock;

`ifdef  VERILATOR

        reg             op_sim          /* verilator public_flat */;

        reg     [22:0]   op_sim_immv     /* verilator public_flat */;

`endif

        //

        //

        //      PIPELINE STAGE #4 :: ALU / Memory

        //              Variable declarations

        //

        //

        wire    [(AW-1):0]       alu_pc;

        reg             r_alu_pc_valid, mem_pc_valid;

        wire            alu_pc_valid;

        wire            alu_phase;

        wire            alu_ce /* verilator public_flat */, alu_stall;

        wire    [31:0]   alu_result;

        wire    [3:0]    alu_flags;

        wire            alu_valid, alu_busy;

        wire            set_cond;

        reg             alu_wR, alu_wF;

        wire            alu_gie, alu_illegal;

        wire    mem_ce, mem_stalled;

        wire    mem_pipe_stalled;

        wire    mem_valid, mem_ack, mem_stall, mem_err, bus_err,

                mem_cyc_gbl, mem_cyc_lcl, mem_stb_gbl, mem_stb_lcl, mem_we;

        wire    [4:0]            mem_wreg;

        wire                    mem_busy, mem_rdbusy;

        wire    [(AW-1):0]       mem_addr;

        wire    [31:0]           mem_data, mem_result;

        wire    [3:0]            mem_sel;

        wire    div_ce, div_error, div_busy, div_valid;

        wire    [31:0]   div_result;

        wire    [3:0]    div_flags;

        assign  div_ce = (master_ce)&&(!clear_pipeline)&&(op_valid_div)

                                &&(!mem_rdbusy)&&(!div_busy)&&(!fpu_busy)

                                &&(set_cond);

        wire    fpu_ce, fpu_error, fpu_busy, fpu_valid;

        wire    [31:0]   fpu_result;

        wire    [3:0]    fpu_flags;

        assign  fpu_ce = (master_ce)&&(!clear_pipeline)&&(op_valid_fpu)

                                &&(!mem_rdbusy)&&(!div_busy)&&(!fpu_busy)

                                &&(set_cond);

        wire    adf_ce_unconditional;

        //

        //

        //      PIPELINE STAGE #5 :: Write-back

        //              Variable declarations

        //

        wire            wr_reg_ce, wr_flags_ce, wr_write_pc, wr_write_cc,

                        wr_write_scc, wr_write_ucc;

        wire    [4:0]    wr_reg_id;

        wire    [31:0]   wr_gpreg_vl, wr_spreg_vl;

        wire    w_switch_to_interrupt, w_release_from_interrupt;

        reg     [(AW+1):0]       ipc;

        wire    [(AW+1):0]       upc;

        //

        //      MASTER: clock enable.

        //

        assign  master_ce = ((!i_halt)||(alu_phase))&&(!o_break)&&(!sleep);

        //

        //      PIPELINE STAGE #1 :: Prefetch

        //              Calculate stall conditions

        //

        //      These are calculated externally, within the prefetch module.

        //

        //

        //      PIPELINE STAGE #2 :: Instruction Decode

        //              Calculate stall conditions

`ifdef  OPT_PIPELINED

        assign          dcd_stalled = (dcd_valid)&&(op_stall);

`else // Not pipelined -- either double or single fetch

        assign          dcd_stalled = (dcd_valid)&&(op_stall);

`endif

        //

        //      PIPELINE STAGE #3 :: Read Operands

        //              Calculate stall conditions

        wire    prelock_stall;

`ifdef  OPT_PIPELINED

        reg     cc_invalid_for_dcd;

        always @(posedge i_clk)

                cc_invalid_for_dcd <= (wr_flags_ce)

                        ||(wr_reg_ce)&&(wr_reg_id[3:0] == `CPU_CC_REG)

                        ||(op_valid)&&((op_wF)||((op_wR)&&(op_R[3:0] == `CPU_CC_REG)))

                        ||((alu_wF)||((alu_wR)&&(alu_reg[3:0] == `CPU_CC_REG)))

                        ||(mem_busy)||(div_busy)||(fpu_busy);

        assign  op_stall = (op_valid)&&( // Only stall if we're loaded w/validins

                        // Stall if we're stopped, and not allowed to execute

                        // an instruction

                        // (!master_ce)         // Already captured in alu_stall

                        //

                        // Stall if going into the ALU and the ALU is stalled

                        //      i.e. if the memory is busy, or we are single

                        //      stepping.  This also includes our stalls for

                        //      op_break and op_lock, so we don't need to

                        //      include those as well here.

                        // This also includes whether or not the divide or

                        // floating point units are busy.

                        (alu_stall)

                        ||(((op_valid_div)||(op_valid_fpu))

                                &&(!adf_ce_unconditional))

                        //

                        // Stall if we are going into memory with an operation

                        //      that cannot be pipelined, and the memory is

                        //      already busy

                        ||(mem_stalled) // &&(op_valid_mem) part of mem_stalled

                        ||(op_Rcc)

                        )

                        ||(dcd_valid)&&(

                                // Stall if we need to wait for an operand A

                                // to be ready to read

                                (dcd_A_stall)

                                // Likewise for B, also includes logic

                                // regarding immediate offset (register must

                                // be in register file if we need to add to

                                // an immediate)

                                ||(dcd_B_stall)

                                // Or if we need to wait on flags to work on the

                                // CC register

                                ||(dcd_F_stall)

                        );

        assign  op_ce = ((dcd_valid)||(dcd_illegal)||(dcd_early_branch))&&(!op_stall);

`else

        assign  op_stall = (alu_busy)||(div_busy)||(fpu_busy)||(wr_reg_ce)

                        ||(mem_busy)||(op_valid)||(!master_ce)||(wr_flags_ce);

        assign  op_ce = ((dcd_valid)||(dcd_illegal)||(dcd_early_branch))&&(!op_stall);

`endif

        // BUT ... op_ce is too complex for many of the data operations.  So

        // let's make their circuit enable code simpler.  In particular, if

        // op_ doesn't need to be preserved, we can change it all we want

        // ... right?  The clear_pipeline code, for example, really only needs

        // to determine whether op_valid is true.

        assign  op_change_data_ce = (!op_stall);

        //

        //      PIPELINE STAGE #4 :: ALU / Memory

        //              Calculate stall conditions

        //

        // 1. Basic stall is if the previous stage is valid and the next is

        //      busy.

        // 2. Also stall if the prior stage is valid and the master clock enable

        //      is de-selected

        // 3. Stall if someone on the other end is writing the CC register,

        //      since we don't know if it'll put us to sleep or not.

        // 4. Last case: Stall if we would otherwise move a break instruction

        //      through the ALU.  Break instructions are not allowed through

        //      the ALU.

`ifdef  OPT_PIPELINED

        assign  alu_stall = (((!master_ce)||(mem_rdbusy)||(alu_busy))&&(op_valid_alu)) //Case 1&2

                        ||(prelock_stall)

                        ||((op_valid)&&(op_break))

                        ||(wr_reg_ce)&&(wr_write_cc)

                        ||(div_busy)||(fpu_busy);

        assign  alu_ce = (master_ce)&&(op_valid_alu)&&(!alu_stall)

                                &&(!clear_pipeline);

`else

        assign  alu_stall = (op_valid_alu)&&((!master_ce)||(op_break));

        assign  alu_ce = (master_ce)&&(op_valid_alu)&&(!alu_stall)&&(!clear_pipeline);

`endif

        //

        //

        // Note: if you change the conditions for mem_ce, you must also change

        // alu_pc_valid.

        //

        assign  mem_ce = (master_ce)&&(op_valid_mem)&&(!mem_stalled)

                        &&(!clear_pipeline);

`ifdef  OPT_PIPELINED_BUS_ACCESS

        assign  mem_stalled = (!master_ce)||(alu_busy)||((op_valid_mem)&&(

                                (mem_pipe_stalled)

                                ||(prelock_stall)

                                ||((!op_pipe)&&(mem_busy))

                                ||(div_busy)

                                ||(fpu_busy)

                                // Stall waiting for flags to be valid

                                // Or waiting for a write to the PC register

                                // Or CC register, since that can change the

                                //  PC as well

                                ||((wr_reg_ce)&&(wr_reg_id[4] == op_gie)

                                        &&((wr_write_pc)||(wr_write_cc)))));

`else

`ifdef  OPT_PIPELINED

        assign  mem_stalled = (mem_busy)||((op_valid_mem)&&(

                                (!master_ce)

                                // Stall waiting for flags to be valid

                                // Or waiting for a write to the PC register

                                // Or CC register, since that can change the

                                //  PC as well

                                ||((wr_reg_ce)&&(wr_reg_id[4] == op_gie)&&((wr_write_pc)||(wr_write_cc)))));

`else

        assign  mem_stalled = (op_valid_mem)&&(!master_ce);

`endif

`endif

        // ALU, DIV, or FPU CE ... equivalent to the OR of all three of these

        assign  adf_ce_unconditional = (master_ce)&&(!clear_pipeline)&&(op_valid)

                                &&(!op_valid_mem)&&(!mem_rdbusy)

                                &&((!op_valid_alu)||(!alu_stall))&&(!op_break)

                                &&(!div_busy)&&(!fpu_busy)&&(!clear_pipeline);

        //

        //

        //      PIPELINE STAGE #1 :: Prefetch

        //

        //

        wire    pf_stalled;

        assign  pf_stalled = (dcd_stalled)||(dcd_phase);

        wire    pf_new_pc;

        assign  pf_new_pc = (new_pc)||((dcd_early_branch)&&(!clear_pipeline));

        wire    [(AW-1):0]       pf_request_address;

        assign  pf_request_address = ((dcd_early_branch)&&(!clear_pipeline))

                                ? dcd_branch_pc:pf_pc[(AW+1):2];

        assign  pf_gie = gie;

`ifdef  OPT_SINGLE_FETCH

        prefetch        #(ADDRESS_WIDTH)

                        pf(i_clk, (i_rst), pf_new_pc, w_clear_icache,

                                (!pf_stalled),

                                pf_request_address,

                                pf_instruction, pf_instruction_pc,

                                        pf_valid, pf_illegal,

                                pf_cyc, pf_stb, pf_we, pf_addr, pf_data,

                                pf_ack, pf_stall, pf_err, i_wb_data);

`else

`ifdef  OPT_DOUBLE_FETCH

        wire    [1:0]    pf_dbg;

        dblfetch #(ADDRESS_WIDTH)

                pf(i_clk, i_rst, pf_new_pc,

                                w_clear_icache,

                                (!pf_stalled),

                                pf_request_address,

                                pf_instruction, pf_instruction_pc,

                                        pf_valid,

                                pf_cyc, pf_stb, pf_we, pf_addr, pf_data,

                                        pf_ack, pf_stall, pf_err, i_wb_data,

                                pf_illegal);

`else // Not single fetch and not double fetch

`ifdef  OPT_TRADITIONAL_PFCACHE

        pfcache #(LGICACHE, ADDRESS_WIDTH)

                pf(i_clk, i_rst, pf_new_pc, w_clear_icache,

                                // dcd_pc,

                                (!pf_stalled),

                                pf_request_address,

                                pf_instruction, pf_instruction_pc, pf_valid,

                                pf_cyc, pf_stb, pf_we, pf_addr, pf_data,

                                        pf_ack, pf_stall, pf_err, i_wb_data,

                                pf_illegal);

`else

        pipefetch       #(RESET_BUS_ADDRESS, LGICACHE, ADDRESS_WIDTH)

                        pf(i_clk, i_rst, pf_new_pc,

                                        w_clear_icache, (!pf_stalled),

                                        (new_pc)?pf_pc[(AW+1):2]:dcd_branch_pc,

                                        pf_instruction, pf_instruction_pc, pf_valid,

                                pf_cyc, pf_stb, pf_we, pf_addr, pf_data,

                                        pf_ack, pf_stall, pf_err, i_wb_data,

                                (mem_cyc_lcl)||(mem_cyc_gbl),

                                pf_illegal);

`endif  // OPT_TRADITIONAL_CACHE

`endif  // OPT_DOUBLE_FETCH

`endif  // OPT_SINGLE_FETCH

        assign          dcd_ce = (!dcd_valid)||(!dcd_stalled);

        idecode #(AW, IMPLEMENT_MPY, EARLY_BRANCHING, IMPLEMENT_DIVIDE,

                        IMPLEMENT_FPU)

                instruction_decoder(i_clk,

                        (clear_pipeline)||(w_clear_icache),

                        dcd_ce,

                        dcd_stalled, pf_instruction, pf_gie,

                        pf_instruction_pc, pf_valid, pf_illegal,

                        dcd_valid, dcd_phase,

                        dcd_illegal, dcd_pc, dcd_gie,

                        { dcd_Rcc, dcd_Rpc, dcd_R },

                        { dcd_Acc, dcd_Apc, dcd_A },

                        { dcd_Bcc, dcd_Bpc, dcd_B },

                        dcd_I, dcd_zI, dcd_F, dcd_wF, dcd_opn,

                        dcd_ALU, dcd_M, dcd_DIV, dcd_FP, dcd_break, dcd_lock,

                        dcd_wR,dcd_rA, dcd_rB,

                        dcd_early_branch,

                        dcd_branch_pc, dcd_ljmp,

                        dcd_pipe,

                        dcd_sim, dcd_sim_immv);

`ifdef  OPT_PIPELINED_BUS_ACCESS

        reg             r_op_pipe;

        initial r_op_pipe = 1'b0;

        // To be a pipeable operation, there must be

        //      two valid adjacent instructions

        //      Both must be memory instructions

        //      Both must be writes, or both must be reads

        //      Both operations must be to the same identical address,

        //              or at least a single (one) increment above that address

        //

        // However ... we need to know this before this clock, hence this is

        // calculated in the instruction decoder.

        always @(posedge i_clk)

                if (clear_pipeline)

                        r_op_pipe <= 1'b0;

                else if (op_ce)

                        r_op_pipe <= dcd_pipe;

                else if (mem_ce) // Clear us any time an op_ is clocked in

                        r_op_pipe <= 1'b0;

        assign  op_pipe = r_op_pipe;

`else

        assign  op_pipe = 1'b0;

`endif

        //

        //

        //      PIPELINE STAGE #3 :: Read Operands (Registers)

        //