actlib_dataflow_neuro/dataflow_neuro/primitives.act

/*************************************************************************
 *
 *  This file is part of ACT dataflow neuro library
 *
 *  Copyright (c) 2022 University of Groningen - Ole Richter
 *  Copyright (c) 2022 University of Groningen - Michele Mastella
 *  Copyright (c) 2022 University of Groningen - Hugh Greatorex
 *  Copyright (c) 2022 University of Groningen - Madison Cotteret
 * 
 *
 *  This source describes Open Hardware and is licensed under the CERN-OHL-W v2 or later
 *
 *  You may redistribute and modify this documentation and make products
 *  using it under the terms of the CERN-OHL-W v2 (https:/cern.ch/cern-ohl).
 *  This documentation is distributed WITHOUT ANY EXPRESS OR IMPLIED
 *  WARRANTY, INCLUDING OF MERCHANTABILITY, SATISFACTORY QUALITY
 *  AND FITNESS FOR A PARTICULAR PURPOSE. Please see the CERN-OHL-W v2
 *  for applicable conditions.
 *
 *  Source location: https://git.web.rug.nl/bics/actlib_dataflow_neuro
 *
 *  As per CERN-OHL-W v2 section 4.1, should You produce hardware based on
 *  these sources, You must maintain the Source Location visible in its
 *  documentation.
 *
 **************************************************************************
 */
import "../../dataflow_neuro/cell_lib_async.act";
import "../../dataflow_neuro/cell_lib_std.act";
import "../../dataflow_neuro/treegates.act";
// import tmpl::dataflow_neuro;
// import tmpl::dataflow_neuro;
import std::channel;
open std::channel;

// import std::func;


namespace tmpl {
    namespace dataflow_neuro {

        // @ole talk to rajit, we use valid the wrong way arround according to stdlib
        template<pbool reset; pint V; pint M>
        defchan gen_avMx1of2 <: chan(int<M>) (std::data::Mx1of2?!<M> d; bool!? a; bool!? v)
        {
          { 0 <= V & std::ceil_log2(V) < M : "Initial token value out of range" };
          
          methods {
            /*-- initialize channel, sender end --*/
            send_init {
              [ reset -> (,i:M: [ ((V >> i) & 1) = 0 -> d.d[i].f+ [] else -> d.d[i].t+ ]);[v]
              [] else -> (,i:M: d.d[i].t-,d.d[i].f-);[~v]
              ]
            }
        
            /*-- set output data --*/
            set {
              (,i:M: [((self >> i) & 1) = 0 -> d.d[i].f+ [] else -> d.d[i].t+ ]);[v]
            }
        
            /*-- finish synchronization --*/
            send_up {
              [a]
            }
        
            /*-- reset part of the protocol --*/
            send_rest {
              (,i:M: d.d[i].t-,d.d[i].f-);[~v],[~a]
            }
        
            /*-- initialize channel, receiver end --*/
            recv_init {
              v-;a-
            }
        
            /*-- get value --*/
            get {
              [(&i:M: d.d[i].t | d.d[i].f)];
              self := 0;
              (;i:M: [ d.d[i].t -> self := self | (1 << i)
               [] else -> skip
               ]
              )
            }
        
            /*-- finish synchronization action --*/
            recv_up {
              v+,a+
            }
        
            /*-- reset part of the protocol --*/
            recv_rest {
              [(&i:M:~d.d[i].t & ~d.d[i].f)];v-,a-
            }
        
            /*-- probe expression on receiver --*/
            // i think this deadlocks with recv_up
            recv_probe = v;
        
            // no sender probe
          }
        }
        export defchan avMx1of2 <: gen_avMx1of2<false,0> () { }
        export defchan avrMx1of2 <: gen_avMx1of2<true,0> () { }
        

        /**
         * the buffer template gives you a standart buffer of bitwidth N
         *
         */
        export template<pint N>
        defproc buffer (avMx1of2<N> in; avMx1of2<N> out; bool? reset_B; power supply) {
            //control
            bool _en, _reset_BX,_reset_BXX[N*2];
            A_3C_RB_X4 inack_ctl(.c1=_en,.c2=in.v,.c3=out.v,.y=in.a,.pr_B=_reset_BX,.sr_B=_reset_BX,.vdd=supply.vdd,.vss=supply.vss);
            A_1C1P_X1 en_ctl(.c1=in.a,.p1=out.v,.y=_en,.vdd=supply.vdd,.vss=supply.vss);


            BUF_X1 reset_buf(.a=reset_B, .y=_reset_BX,.vdd=supply.vdd,.vss=supply.vss);
            sigbuf<N*2> reset_bufarray(.in=_reset_BX, .out=_reset_BXX);
      
            //validity
            bool _in_v;
            vtree<N> vc(.in=in.d,.out=_in_v,.supply=supply);
            BUF_X4 in_v_buf(.a=_in_v, .y=in.v,.vdd=supply.vdd,.vss=supply.vss);
            
            //function
            bool _out_a_BX_t[N],_out_a_BX_f[N],_out_a_B,_en_X_t[N],_en_X_f[N];
            A_2C1N_RB_X4 f_buf_func[N];
            A_2C1N_RB_X4 t_buf_func[N];
            sigbuf<N> en_buf_t(.in=_en, .out=_en_X_t, .supply=supply);
            sigbuf<N> en_buf_f(.in=_en, .out=_en_X_f, .supply=supply);
            INV_X1 out_a_inv(.a=out.a,.y=_out_a_B, .vss = supply.vss, .vdd = supply.vdd);
            sigbuf<N> out_a_B_buf_f(.in=_out_a_B,.out=_out_a_BX_t, .supply = supply);
            sigbuf<N> out_a_B_buf_t(.in=_out_a_B,.out=_out_a_BX_f, .supply = supply);
            // check if you can also do single var to array connect a=b[N]
            // and remove them from the loop
            (i:N: 
                f_buf_func[i].y=out.d.d[i].f;
                t_buf_func[i].y=out.d.d[i].t;
                f_buf_func[i].c1=_en_X_f[i];
                t_buf_func[i].c1=_en_X_t[i];
                f_buf_func[i].c2=_out_a_BX_f[i];
                t_buf_func[i].c2=_out_a_BX_t[i];
                f_buf_func[i].n1=in.d.d[i].f;
                t_buf_func[i].n1=in.d.d[i].t;
                f_buf_func[i].vdd=supply.vdd;
                t_buf_func[i].vdd=supply.vdd;
                f_buf_func[i].vss=supply.vss;
                t_buf_func[i].vss=supply.vss;
                t_buf_func[i].pr_B = _reset_BXX[i];
                t_buf_func[i].sr_B = _reset_BXX[i];
                f_buf_func[i].pr_B = _reset_BXX[i+N];
                f_buf_func[i].sr_B = _reset_BXX[i+N];
            )
        }
        // A template creating a FIFO of M buffers with N bits each
        export template<pint N;pint M>
        defproc fifo(avMx1of2<N> in; avMx1of2<N> out; bool? reset_B; power supply)
        {
          buffer<N> fifo_element[M];
          bool _reset_BXX[M];
          fifo_element[0].in = in ;
          fifo_element[0].supply = supply;
          fifo_element[0].reset_B = _reset_BXX[0];
          (i:1..M-1:
            fifo_element[i].in = fifo_element[i-1].out;
            fifo_element[i].supply = supply;
            fifo_element[i].reset_B = _reset_BXX[i];
          )
          fifo_element[M-1].out = out;

          // reset buffers
          bool _reset_BX;
          BUF_X1 reset_buf(.a=reset_B, .y=_reset_BX,.vdd=supply.vdd,.vss=supply.vss);
          sigbuf<M> reset_bufarray(.in=_reset_BX, .out=_reset_BXX,.supply = supply);
        }

        /**
         * Buffer_S template.
         * S maybe stands for special.
         * Like a buffer, except that the output function block does not load the data in
         * until the input data is valid.
         * Not entirely sure what the point of it is, 
         * Ole says is useful for funky timing scenarios.
         */
        export template<pint N>
        defproc buffer_s (avMx1of2<N> in; avMx1of2<N> out; bool? reset_B; power supply) {
            //control
            bool _en, _reset_BX,_reset_BXX[N];
            A_3C_RB_X4 inack_ctl(.c1=_en,.c2=in.v,.c3=out.v,.y=in.a,.pr_B=_reset_BX,.sr_B=_reset_BX,.vdd=supply.vdd,.vss=supply.vss);
            A_1C1P_X1 en_ctl(.c1=in.a,.p1=out.v,.y=_en,.vdd=supply.vdd,.vss=supply.vss);


            BUF_X1 reset_buf(.a=reset_B, .y=_reset_BX,.vdd=supply.vdd,.vss=supply.vss);
            sigbuf<N> reset_bufarray(.in=_reset_BX, .out=_reset_BXX, .supply=supply);
        
            //validity
            bool _in_v, _in_vX[N];
            vtree<N> vc(.in=in.d,.out=_in_v,.supply=supply);
            BUF_X4 in_v_buf4(.a=_in_v, .y=in.v,.vdd=supply.vdd,.vss=supply.vss);
            sigbuf<N> in_v_bufN(.in = in.v, .out = _in_vX, .supply = supply);

            //function
            bool _out_a_BX_t[N],_out_a_BX_f[N],_out_a_B,_en_X_t[N],_en_X_f[N];
            A_2C2N_RB_X4 f_buf_func[N];
            A_2C2N_RB_X4 t_buf_func[N];
            sigbuf<N> en_buf_t(.in=_en, .out=_en_X_t, .supply=supply);
            sigbuf<N> en_buf_f(.in=_en, .out=_en_X_f, .supply=supply);
            INV_X1 out_a_inv(.a=out.a,.y=_out_a_B, .vss = supply.vss, .vdd = supply.vdd);
            sigbuf<N> out_a_B_buf_f(.in=_out_a_B,.out=_out_a_BX_t, .supply=supply);
            sigbuf<N> out_a_B_buf_t(.in=_out_a_B,.out=_out_a_BX_f, .supply=supply);
            // check if you can also do single var to array connect a=b[N]
            // and remove them from the loop
            (i:N: 
                f_buf_func[i].y=out.d.d[i].f;
                t_buf_func[i].y=out.d.d[i].t;
                f_buf_func[i].c1=_en_X_f[i];
                t_buf_func[i].c1=_en_X_t[i];
                f_buf_func[i].c2=_out_a_BX_f[i];
                t_buf_func[i].c2=_out_a_BX_t[i];
                f_buf_func[i].n1=in.d.d[i].f;
                t_buf_func[i].n1=in.d.d[i].t;
                f_buf_func[i].n2=_in_vX[i];
                t_buf_func[i].n2=_in_vX[i];
                f_buf_func[i].vdd=supply.vdd;
                t_buf_func[i].vdd=supply.vdd;
                f_buf_func[i].vss=supply.vss;
                t_buf_func[i].vss=supply.vss;
                t_buf_func[i].pr_B = _reset_BXX[i];
                t_buf_func[i].sr_B = _reset_BXX[i];
                f_buf_func[i].pr_B = _reset_BXX[i];
                f_buf_func[i].sr_B = _reset_BXX[i];
            )
        }
    
        export template<pint N>
        defproc demux (avMx1of2<N> in; avMx1of2<N> out1; avMx1of2<N> out2; bool? reset_B; avMx1of2<1> cond; power supply) {
            //control
            bool _en, _reset_BX,_reset_BXX[2*N], _out_v, _in_c_v_;
            
            OR2_X1 out_or(.a=out1.v, .b=out2.v, .y=_out_v,.vdd=supply.vdd,.vss=supply.vss);
            A_3C_RB_X4 inack_ctl(.c1=_en,.c2=_in_c_v_,.c3=_out_v,.y=in.a,.pr_B=_reset_BX,.sr_B=_reset_BX,.vdd=supply.vdd,.vss=supply.vss);
            cond.a = in.a; // THIS SHOULD BE IMPROVED UPON IN FUTURE VERSIONS
            cond.v = _in_c_v_;
            A_1C1P_X1 en_ctl(.c1=in.a,.p1=_out_v,.y=_en,.vdd=supply.vdd,.vss=supply.vss);
            BUF_X1 reset_buf(.a=reset_B, .y=_reset_BX,.vdd=supply.vdd,.vss=supply.vss);
            sigbuf<2*N> reset_bufarray(.in=_reset_BX, .out=_reset_BXX);
            //validity
            bool _in_v, _c_f_buf[N], _c_t_buf[N], _c_v;
            
            sigbuf<N> c_buf_t(.in=cond.d.d[0].t, .out=_c_t_buf, .supply=supply);
            sigbuf<N> c_buf_f(.in=cond.d.d[0].f, .out=_c_f_buf, .supply=supply);
            
            OR2_X1 c_f_c_t_or(.a=cond.d.d[0].t, .b=cond.d.d[0].f, .y=_c_v,.vdd=supply.vdd,.vss=supply.vss);
            vtree<N> vc(.in=in.d,.out=_in_v,.supply=supply);
            
            A_2C_B_X1 c_el(.c1=_c_v, .c2=_in_v, .y=_in_c_v_,.vdd=supply.vdd,.vss=supply.vss);
            BUF_X4 in_v_buf(.a=_in_v, .y=in.v,.vdd=supply.vdd,.vss=supply.vss);
            
            
            //function
            //func buffer out1
            bool _out1_a_BX_t[N],_out1_a_BX_f[N],_out1_a_B,_en1_X_t[N],_en1_X_f[N];
            A_2C2N_RB_X4 out1_f_buf_func[N];
            A_2C2N_RB_X4 out1_t_buf_func[N];
            sigbuf<N> out1_en_buf_t(.in=_en, .out=_en1_X_t, .supply=supply);
            sigbuf<N> out1_en_buf_f(.in=_en, .out=_en1_X_f, .supply=supply);
            INV_X1 out1_a_inv(.a=out1.a,.y=_out1_a_B, .vdd = supply.vdd, .vss = supply.vss);
            sigbuf<N> out1_a_B_buf_f(.in=_out1_a_B,.out=_out1_a_BX_t, .supply=supply);
            sigbuf<N> out1_a_B_buf_t(.in=_out1_a_B,.out=_out1_a_BX_f, .supply=supply);
            (i:N: 
                out1_f_buf_func[i].y=out1.d.d[i].f;
                out1_t_buf_func[i].y=out1.d.d[i].t;
                out1_f_buf_func[i].c1=_en1_X_f[i];
                out1_t_buf_func[i].c1=_en1_X_t[i];
                out1_f_buf_func[i].c2=_out1_a_BX_f[i];
                out1_t_buf_func[i].c2=_out1_a_BX_t[i];
                out1_f_buf_func[i].n1=in.d.d[i].f;
                out1_t_buf_func[i].n1=in.d.d[i].t;
                out1_f_buf_func[i].vdd=supply.vdd;
                out1_t_buf_func[i].vdd=supply.vdd;
                out1_f_buf_func[i].vss=supply.vss;
                out1_t_buf_func[i].vss=supply.vss;
                out1_t_buf_func[i].pr_B = _reset_BXX[i];
                out1_t_buf_func[i].sr_B = _reset_BXX[i];
                out1_f_buf_func[i].pr_B = _reset_BXX[i];
                out1_f_buf_func[i].sr_B = _reset_BXX[i];
                out1_f_buf_func[i].n2=_c_f_buf[i];
                out1_t_buf_func[i].n2=_c_f_buf[i];
            )

            //func buffer out2
            bool _out2_a_BX_t[N],_out2_a_BX_f[N],_out2_a_B,_en2_X_t[N],_en2_X_f[N];
            A_2C2N_RB_X4 out2_f_buf_func[N];
            A_2C2N_RB_X4 out2_t_buf_func[N];
            sigbuf<N> out2_en_buf_t(.in=_en, .out=_en2_X_t, .supply=supply);
            sigbuf<N> out2_en_buf_f(.in=_en, .out=_en2_X_f, .supply=supply);
            INV_X1 out2_a_inv(.a=out2.a,.y=_out2_a_B, .vdd = supply.vdd, .vss = supply.vss);
            sigbuf<N> out2_a_B_buf_f(.in=_out2_a_B,.out=_out2_a_BX_t);
            sigbuf<N> out2_a_B_buf_t(.in=_out2_a_B,.out=_out2_a_BX_f);
            (i:N: 
                out2_f_buf_func[i].y=out2.d.d[i].f;
                out2_t_buf_func[i].y=out2.d.d[i].t;
                out2_f_buf_func[i].c1=_en2_X_f[i];
                out2_t_buf_func[i].c1=_en2_X_t[i];
                out2_f_buf_func[i].c2=_out2_a_BX_f[i];
                out2_t_buf_func[i].c2=_out2_a_BX_t[i];
                out2_f_buf_func[i].n1=in.d.d[i].f;
                out2_t_buf_func[i].n1=in.d.d[i].t;
                out2_f_buf_func[i].vdd=supply.vdd;
                out2_t_buf_func[i].vdd=supply.vdd;
                out2_f_buf_func[i].vss=supply.vss;
                out2_t_buf_func[i].vss=supply.vss;
                out2_t_buf_func[i].pr_B = _reset_BXX[i+N];
                out2_t_buf_func[i].sr_B = _reset_BXX[i+N];
                out2_f_buf_func[i].pr_B = _reset_BXX[i+N];
                out2_f_buf_func[i].sr_B = _reset_BXX[i+N];
                out2_f_buf_func[i].n2=_c_t_buf[i];
                out2_t_buf_func[i].n2=_c_t_buf[i];
            )
        }

        export template<pint N>
        defproc fork (avMx1of2<N> in; avMx1of2<N> out1; avMx1of2<N> out2 ; bool? reset_B; power supply) {
          
          // control 
          bool _en, _reset_BX,_reset_BXX[N*2];
          A_4C_RB_X4 inack_ctl(.c1=_en,.c2=in.v,.c3=out1.v,.c4=out2.v,.y=in.a,.pr_B=_reset_BX,.sr_B=_reset_BX,.vdd=supply.vdd,.vss=supply.vss);
          A_1C2P_X1 en_ctl(.c1=in.a,.p1=out1.v,.p2=out2.v,.y=_en,.vdd=supply.vdd,.vss=supply.vss);
          

          //reset_buffers
          BUF_X1 reset_buf(.a=reset_B, .y=_reset_BX,.vdd=supply.vdd,.vss=supply.vss);
          sigbuf<N*2> reset_bufarray(.in=_reset_BX, .out=_reset_BXX);


          //validity
          bool _in_v;
          vtree<N> vc(.in=in.d,.out=_in_v,.supply=supply);
          BUF_X4 in_v_buf(.a=_in_v, .y=in.v,.vdd=supply.vdd,.vss=supply.vss);

          //function
            //func buffer out1
            bool _out1_a_BX_t[N],_out1_a_BX_f[N],_out1_a_B,_en1_X_t[N],_en1_X_f[N];
            A_2C1N_RB_X4 out1_f_buf_func[N];
            A_2C1N_RB_X4 out1_t_buf_func[N];
            sigbuf<N> out1_en_buf_t(.in=_en, .out=_en1_X_t, .supply=supply);
            sigbuf<N> out1_en_buf_f(.in=_en, .out=_en1_X_f, .supply=supply);
            INV_X1 out1_a_inv(.a=out1.a,.y=_out1_a_B);
            sigbuf<N> out1_a_B_buf_f(.in=_out1_a_B,.out=_out1_a_BX_t);
            sigbuf<N> out1_a_B_buf_t(.in=_out1_a_B,.out=_out1_a_BX_f);
            (i:N: 
                out1_f_buf_func[i].y=out1.d.d[i].f;
                out1_t_buf_func[i].y=out1.d.d[i].t;
                out1_f_buf_func[i].c1=_en1_X_f[i];
                out1_t_buf_func[i].c1=_en1_X_t[i];
                out1_f_buf_func[i].c2=_out1_a_BX_f[i];
                out1_t_buf_func[i].c2=_out1_a_BX_t[i];
                out1_f_buf_func[i].n1=in.d.d[i].f;
                out1_t_buf_func[i].n1=in.d.d[i].t;
                out1_f_buf_func[i].vdd=supply.vdd;
                out1_t_buf_func[i].vdd=supply.vdd;
                out1_f_buf_func[i].vss=supply.vss;
                out1_t_buf_func[i].vss=supply.vss;
                out1_t_buf_func[i].pr_B = _reset_BXX[i];
                out1_t_buf_func[i].sr_B = _reset_BXX[i];
                out1_f_buf_func[i].pr_B = _reset_BXX[i];
                out1_f_buf_func[i].sr_B = _reset_BXX[i];
            )
            //func buffer out2
            bool _out2_a_BX_t[N],_out2_a_BX_f[N],_out2_a_B,_en2_X_t[N],_en2_X_f[N];
            A_2C1N_RB_X4 out2_f_buf_func[N];
            A_2C1N_RB_X4 out2_t_buf_func[N];
            sigbuf<N> out2_en_buf_t(.in=_en, .out=_en2_X_t, .supply=supply);
            sigbuf<N> out2_en_buf_f(.in=_en, .out=_en2_X_f, .supply=supply);
            INV_X1 out2_a_inv(.a=out2.a,.y=_out2_a_B);
            sigbuf<N> out2_a_B_buf_f(.in=_out2_a_B,.out=_out2_a_BX_t);
            sigbuf<N> out2_a_B_buf_t(.in=_out2_a_B,.out=_out2_a_BX_f);
            (i:N: 
                out2_f_buf_func[i].y=out2.d.d[i].f;
                out2_t_buf_func[i].y=out2.d.d[i].t;
                out2_f_buf_func[i].c1=_en2_X_f[i];
                out2_t_buf_func[i].c1=_en2_X_t[i];
                out2_f_buf_func[i].c2=_out2_a_BX_f[i];
                out2_t_buf_func[i].c2=_out2_a_BX_t[i];
                out2_f_buf_func[i].n1=in.d.d[i].f;
                out2_t_buf_func[i].n1=in.d.d[i].t;
                out2_f_buf_func[i].vdd=supply.vdd;
                out2_t_buf_func[i].vdd=supply.vdd;
                out2_f_buf_func[i].vss=supply.vss;
                out2_t_buf_func[i].vss=supply.vss;
                out2_t_buf_func[i].pr_B = _reset_BXX[i+N];
                out2_t_buf_func[i].sr_B = _reset_BXX[i+N];
                out2_f_buf_func[i].pr_B = _reset_BXX[i+N];
                out2_f_buf_func[i].sr_B = _reset_BXX[i+N];
            )
        }
        
        // Demux 
        export template<pint N; pbool CONDITION_SIGN>
        defproc demux_td (avMx1of2<N> in; avMx1of2<N> out1; a1of1 token; bool? reset_B; avMx1of2<1> cond; power supply) {
            //control
            bool _en, _reset_BX,_reset_BXX[N], _out_v, _in_c_v_, _reset_BXt;

            
            OR2_X1 out_or(.a=out1.v, .b=token.r, .y=_out_v,.vdd=supply.vdd,.vss=supply.vss);
            A_3C_RB_X4 inack_ctl(.c1=_en,.c2=_in_c_v_,.c3= _out_v,.y=in.a,.pr_B=_reset_BX,.sr_B=_reset_BX,.vdd=supply.vdd,.vss=supply.vss);
            A_1C1P_X1 en_ctl(.c1=in.a,.p1=_out_v,.y=_en,.vdd=supply.vdd,.vss=supply.vss);
            BUF_X1 reset_buf(.a=reset_B, .y=_reset_BX,.vdd=supply.vdd,.vss=supply.vss);
            BUF_X1 reset_buf_token(.a=_reset_BX, .y=_reset_BXt,.vdd=supply.vdd,.vss=supply.vss);
            sigbuf<N> reset_bufarray(.in=_reset_BX, .out=_reset_BXX, .supply=supply);
            //validity
            bool _in_v, _c_tk_buf, _c_d_buf[N], _c_v, cond_inv_t, cond_inv_f;


            cond.a = in.a;
            cond.v = _c_v;
            
            OR2_X1 c_f_c_t_or(.a=cond.d.d[0].t, .b=cond.d.d[0].f, .y=_c_v,.vdd=supply.vdd,.vss=supply.vss);
            
            //orientation of condition
            [ CONDITION_SIGN ->
            BUF_X1 c_buf_tk(.a=cond.d.d[0].t, .y=_c_tk_buf, .vss = supply.vss, .vdd = supply.vdd);
            sigbuf<N> c_buf_d(.in=cond.d.d[0].f, .out=_c_d_buf, .supply=supply);
            [] else ->
            INV_X1 invout_t(.a = cond.d.d[0].t,.y=cond_inv_t,.vdd = supply.vdd,.vss=supply.vss);
            INV_X1 invout_f(.a = cond.d.d[0].f,.y=cond_inv_f,.vdd = supply.vdd,.vss=supply.vss);
            BUF_X1 c_buf_tk_inv(.a=cond_inv_t, .y=_c_tk_buf, .vss = supply.vss, .vdd = supply.vdd);
            sigbuf<N> c_buf_d_inv(.in=cond_inv_f, .out=_c_d_buf, .supply=supply);
            ]

            
            vtree<N> vc(.in=in.d,.out=_in_v,.supply=supply);
            
            A_2C_B_X1 c_el(.c1=_c_v, .c2=_in_v, .y=_in_c_v_,.vdd=supply.vdd,.vss=supply.vss);
            BUF_X4 in_v_buf(.a=_in_v, .y=in.v,.vdd=supply.vdd,.vss=supply.vss);
            
            
            //function
            //func buffer out1
            bool _out1_a_BX_t[N],_out1_a_BX_f[N],_out1_a_B,_en1_X_t[N],_en1_X_f[N];
            A_2C2N_RB_X4 out1_f_buf_func[N];
            A_2C2N_RB_X4 out1_t_buf_func[N];
            sigbuf<N> out1_en_buf_t(.in=_en, .out=_en1_X_t, .supply=supply);
            sigbuf<N> out1_en_buf_f(.in=_en, .out=_en1_X_f, .supply=supply);
            INV_X1 out1_a_inv(.a=out1.a,.y=_out1_a_B, .vss = supply.vss, .vdd = supply.vdd);
            sigbuf<N> out1_a_B_buf_f(.in=_out1_a_B,.out=_out1_a_BX_t, .supply=supply);
            sigbuf<N> out1_a_B_buf_t(.in=_out1_a_B,.out=_out1_a_BX_f, .supply=supply);
            (i:N: 
                out1_f_buf_func[i].y=out1.d.d[i].f;
                out1_t_buf_func[i].y=out1.d.d[i].t;
                out1_f_buf_func[i].c1=_en1_X_f[i];
                out1_t_buf_func[i].c1=_en1_X_t[i];
                out1_f_buf_func[i].c2=_out1_a_BX_f[i];
                out1_t_buf_func[i].c2=_out1_a_BX_t[i];
                out1_f_buf_func[i].n1=in.d.d[i].f;
                out1_t_buf_func[i].n1=in.d.d[i].t;
                out1_f_buf_func[i].vdd=supply.vdd;
                out1_t_buf_func[i].vdd=supply.vdd;
                out1_f_buf_func[i].vss=supply.vss;
                out1_t_buf_func[i].vss=supply.vss;
                out1_t_buf_func[i].pr_B = _reset_BXX[i];
                out1_t_buf_func[i].sr_B = _reset_BXX[i];
                out1_f_buf_func[i].pr_B = _reset_BXX[i];
                out1_f_buf_func[i].sr_B = _reset_BXX[i];
                out1_f_buf_func[i].n2=_c_d_buf[i];
                out1_t_buf_func[i].n2=_c_d_buf[i];
            )

            //token out
           
          bool token_a_out;
          A_2C2N_RB_X4 token_buf;
          INV_X1 outt_a_inv(.a=token.a,.y=token_a_out, .vss = supply.vss, .vdd = supply.vdd);

          
          token_buf.y = token.r;
          token_buf.c1 = _en;
          token_buf.c2 = token_a_out; 
          token_buf.n1 = _c_tk_buf;
          token_buf.n2 = _in_v;
          
          token_buf.vdd = supply.vdd;
          token_buf.vss = supply.vss;
          token_buf.pr_B = _reset_BXt;
          token_buf.sr_B = _reset_BXt;
        }

        /** 
         * Drops a packet if condition is met, otherwise passes it on.
         * This is a very lazy implementation, where the cond MUST NOT CHANGE DURING OPERATION.
         * Means that this should be used in a very small set of circumstances.
         * 
         * params:
         * N: size of packet
         * CONDITION_DROP: value of cond when packets are dropped.
         */
        export template<pint N; pbool CONDITION_DROP>
        defproc dropper_static (avMx1of2<N> in; avMx1of2<N> out; bool? cond; power supply) {
            bool _drop, _dropB;
            INV_X1 inv(.a = cond, .vss = supply.vss, .vdd = supply.vdd);

            [~CONDITION_DROP ->
                _dropB = cond;
                _drop = inv.y;
            [] CONDITION_DROP ->
                _drop = cond;
                _dropB = inv.y;
            ]

            bool _in_vX;
            vtree<N> vt(.in = in.d, .supply = supply);
            BUF_X4 in_v_buf(.a = vt.out, .y = _in_vX, .vss = supply.vss, .vdd = supply.vdd);
            AND2_X1 and2(.a = _drop, .b = _in_vX, .vss = supply.vss, .vdd = supply.vdd);
            OR2_X1 or2(.a = out.a, .b = and2.y, .vss = supply.vss, .vdd = supply.vdd);
            A_2C_B_X1 ack_Cel(.c1 = or2.y, .c2 = _in_vX, .y = in.a);
            _in_vX = in.v;

            // Sigbufs
            sigbuf<N*2> sb_dropB(.in = _dropB, .supply = supply);
            sigbuf<N*2> sb_in_v(.in = _in_vX, .supply = supply);

            AND3_X1 and_t[N];
            AND3_X1 and_f[N];
            (i:N:
                and_t[i].a = in.d.d[i].t;
                and_f[i].a = in.d.d[i].f;

                and_t[i].y = out.d.d[i].t;
                and_f[i].y = out.d.d[i].f;
                
                and_t[i].b = sb_dropB.out[i];
                and_f[i].b = sb_dropB.out[i+N];

                and_t[i].c = sb_in_v.out[i];
                and_f[i].c = sb_in_v.out[i+N];

                and_t[i].vss = supply.vss;
                and_t[i].vdd = supply.vdd;
            )



        }
        
        
        export 
        defproc arbiter_handshake(a1of1 in1; a1of1 in2; a1of1 out; power supply)
        {
          bool _y1_arb,_y2_arb;

          A_2C_B_X1 ack_cell1(.c1 = out.a,.c2 = _y1_arb,.y = in1.a,.vdd = supply.vdd, .vss = supply.vss);
          A_2C_B_X1 ack_cell2(.c1 = out.a,.c2 = _y2_arb,.y = in2.a,.vdd = supply.vdd, .vss = supply.vss);
          OR2_X1 or_cell(.a = _y1_arb, .b = _y2_arb, .y = out.r,.vdd = supply.vdd, .vss = supply.vss);
          ARBITER arbiter(.a = in1.r, .b = in2.r, .c = in2.a, .d = in1.a, .y1 = _y1_arb, .y2 = _y2_arb, .vdd = supply.vdd, .vss = supply.vss);
          
        }
        //The buffer_t_valid doesn't work
        // export 
        // defproc buffer_t_valid(a1of1 in; a1of1 out; bool? reset_B; power supply)
        // {
        //     //control
        //     bool _en, _reset_BX;
        //     A_3C_RB_X4 inack_ctl(.c1=_en,.c2=in.r,.c3=out.r,.y=in.a,.pr_B=_reset_BX,.sr_B=_reset_BX,.vdd=supply.vdd,.vss=supply.vss);
        //     A_1C1P_X1 en_ctl(.c1=in.a,.p1=out.r,.y=_en,.vdd=supply.vdd,.vss=supply.vss);
            
        //     //function
        //     bool _out_a_B;
        //     INV_X1 inv_outa(.a = out.a,.y=_out_a_B,.vdd = supply.vdd,.vss=supply.vss);
        //     A_2C1N_RB_X4 buf_func(.c1 = _en,.c2 = _out_a_B, .n1 = in.r,.y = out.r, .pr_B = _reset_BX, .sr_B = _reset_BX,.vdd = supply.vdd,.vss=supply.vss);

            
        //     //reset buffers
        //     BUF_X1 reset_buf(.a=reset_B, .y=_reset_BX,.vdd=supply.vdd,.vss=supply.vss);
        // }

        

        export template<pint N>
        defproc merge (avMx1of2<N> in1; avMx1of2<N> in2; avMx1of2<N> out ; bool? reset_B; power supply) {

          //out acknowledge sigbuffer and inverter
          bool _out_a_B,_out_a_BX[2*N];
          INV_X1 out_a_inverter(.a = out.a, .y = _out_a_B);
          sigbuf<2*N> out_a_buffer(.in = _out_a_B,.out = _out_a_BX,.supply=supply);

          //control
          bool _in1_a_B,_in2_a_B,_en,_en_X[2*N], _reset_BX,_reset_BXX[2*N];
          bool _in1_arb,_in2_arb,_in1_arb_X[2*N],_in2_arb_X[2*N];
          A_4C_RB_X4 in1ack_ctl(.c1=_in1_arb,.c2=_en,.c3=in1.v,.c4=out.v,.y=in1.a,.pr_B=_reset_BX,.sr_B=_reset_BX,.vdd=supply.vdd,.vss=supply.vss);
          A_4C_RB_X4 in2ack_ctl(.c1=_in2_arb,.c2=_en,.c3=in2.v,.c4=out.v,.y=in2.a,.pr_B=_reset_BX,.sr_B=_reset_BX,.vdd=supply.vdd,.vss=supply.vss);
          A_4P1N1N_X1 en_ctl(.p1 = in1.a,.p2=in2.a,.p3=out.a,.p4 = out.v, .n1 = in1.a,.n2 = in2.a,.y = _en,.vdd=supply.vdd,.vss=supply.vss);
          sigbuf<2*N> en_buffer(.in = _en,.out = _en_X,.supply=supply);
          INV_X1 in1ack_ctl_inv(.a=in1.a,.y=_in1_a_B,.vdd=supply.vdd,.vss=supply.vss);
          INV_X1 in2ack_ctl_inv(.a=in2.a,.y=_in2_a_B,.vdd=supply.vdd,.vss=supply.vss);

          //reset_buffers
          BUF_X1 reset_buf(.a=reset_B, .y=_reset_BX,.vdd=supply.vdd,.vss=supply.vss);
          sigbuf<N*2> reset_bufarray(.in=_reset_BX, .out=_reset_BXX);

          //validity
          a1of1 _in1_temp,_in2_temp,_out_temp;
          bool _in1_arb_temp,_in2_arb_temp;
          vtree<N> vc1(.in=in1.d,.out=in1.v,.supply=supply);
          vtree<N> vc2(.in=in2.d,.out=in2.v,.supply=supply);
          arbiter_handshake validity_arb(.in1 = _in1_temp,.in2 = _in2_temp,.out =_out_temp, .supply = supply);
          _in1_temp.r = in1.v;
          _in2_temp.r = in2.v;
          _in1_temp.a = _in1_arb_temp;
          _in2_temp.a = _in2_arb_temp;
          _out_temp.r = _out_temp.a;
          AND2_X1 AND_arb1(.a = _in2_a_B,.b = _in1_arb_temp, .y = _in1_arb);
          AND2_X1 AND_arb2(.a = _in1_a_B,.b = _in2_arb_temp, .y = _in2_arb);
          sigbuf<2*N> arb2function1(.in = _in1_arb,.out = _in1_arb_X,.supply=supply);
          sigbuf<2*N> arb2function2(.in = _in2_arb,.out = _in2_arb_X,.supply=supply);

          //function
          A_2C2N2N_RB_X1 merge_func_t[N];
          A_2C2N2N_RB_X1 merge_func_f[N];
          (i:N: 
            merge_func_t[i].c1 = _en_X[i];
            merge_func_t[i].c2 = _out_a_BX[i];
            merge_func_t[i].n1 = _in1_arb_X[i];
            merge_func_t[i].n2 = in1.d.d[i].t;
            merge_func_t[i].n3 = _in2_arb_X[i];
            merge_func_t[i].n4 = in2.d.d[i].t;
            merge_func_t[i].y = out.d.d[i].t;
            merge_func_t[i].vdd=supply.vdd;
            merge_func_t[i].vss=supply.vss;
            merge_func_t[i].pr_B = _reset_BXX[i];
            merge_func_t[i].sr_B = _reset_BXX[i];

            merge_func_f[i].c1 = _en_X[i+N];
            merge_func_f[i].c2 = _out_a_BX[i+N];
            merge_func_f[i].n1 = _in1_arb_X[i+N];
            merge_func_f[i].n2 = in1.d.d[i].f;
            merge_func_f[i].n3 = _in2_arb_X[i+N];
            merge_func_f[i].n4 = in2.d.d[i].f;
            merge_func_f[i].y = out.d.d[i].f;
            merge_func_f[i].vdd=supply.vdd;
            merge_func_f[i].vss=supply.vss;
            merge_func_f[i].pr_B = _reset_BXX[i+N];
            merge_func_f[i].sr_B = _reset_BXX[i+N];
            )
        }
 
        export 
        defproc buffer_t(a1of1 in; a1of1 out; bool? reset_B; power supply)
        {
            //control
            bool _en, _reset_BX;
            A_2C1N_RB_X4 inack_ctl(.c1=_en,.c2=in.r,.n1=out.r,.y=in.a,.pr_B=_reset_BX,.sr_B=_reset_BX,.vdd=supply.vdd,.vss=supply.vss);
            A_1C1P_X1 en_ctl(.c1=in.a,.p1=out.r,.y=_en,.vdd=supply.vdd,.vss=supply.vss);
            
            //function
            bool _out_a_B;
            INV_X1 inv_outa(.a = out.a,.y=_out_a_B,.vdd = supply.vdd,.vss=supply.vss);
            A_2C1N_RB_X4 buf_func(.c1 = _en,.c2 = _out_a_B, .n1 = in.r,.y = out.r, .pr_B = _reset_BX, .sr_B = _reset_BX,.vdd = supply.vdd,.vss=supply.vss);

            
            //reset buffers
            BUF_X1 reset_buf(.a=reset_B, .y=_reset_BX,.vdd=supply.vdd,.vss=supply.vss);
        }
        // A template creating a FIFO of N buffers tokens
        export template<pint N>
        defproc fifo_t(a1of1 in; a1of1 out; bool? reset_B; power supply)
        {
          buffer_t fifo_element[N];
          bool _reset_BXX[N];
          fifo_element[0].in.r = in.r;
          fifo_element[0].in.a = in.a;
          fifo_element[0].supply = supply;
          fifo_element[0].reset_B = _reset_BXX[0];
          (i:1..N-1:
            fifo_element[i].in.r = fifo_element[i-1].out.r;
            fifo_element[i].in.a = fifo_element[i-1].out.a;
            fifo_element[i].supply = supply;
            fifo_element[i].reset_B = _reset_BXX[i];
          )
          fifo_element[N-1].out.r = out.r;
          fifo_element[N-1].out.a = out.a;

          // reset buffers
          bool _reset_BX;
          BUF_X1 reset_buf(.a=reset_B, .y=_reset_BX,.vdd=supply.vdd,.vss=supply.vss);
          sigbuf<N> reset_bufarray(.in=_reset_BX, .out=_reset_BXX, .supply = supply);
        }

        // Programmable delay line.
        // N is the number of layers,
        // the longest layer having 2**N DLY elements
        // Circuit for creating delays, there are N delay layers.  
        // The block has the parameters:
        // N -> the number is the number of layers with the longest being 2**N elements
        // wl -> word length, length of each word 
        // N_dly_cfg -> the number of config bits in the ACK delay line
        // The block has the pins:
        // in -> input data
        // out -> output data
        // s -> bit word with size N that sets delay configuration. int(s) = number of delays
        export template<pint N>
        defproc delayprog (bool! out; bool? in, s[N]; power supply)
        {
            
            { N >= 0 : "What?" };
            { N < 10 : "Delay prog size is given in 2**N. Given N is ridiculous." };


            AND2_X1 and2[N];
            MUX2_X1 mu2[N];
            DLY4_X1 dly[(1<<N) -1];

            bool _a[N+1]; // Holds the input to each row
         
            _a[0] = in;

            pint i_delay;
            i_delay = 0; // Index of the last connected delay element

            (i:0..N-1:
                // For each row
                and2[i].a = _a[i];
                and2[i].b = s[i];

                // Delays
                dly[i_delay].a = and2[i].y;
                i_delay = i_delay + 1;
                (j:1..(1<<i)-1:
                    dly[i_delay].a = dly[i_delay-1].y;
                    i_delay = i_delay +1;
                )

                // Mux
                mu2[i].a = _a[i];
                mu2[i].s = s[i];
                dly[i_delay-1].y = mu2[i].b;
                _a[i+1] = mu2[i].y;
            )

            out = mu2[N-1].y;


            // Connect everything to vdd/gnd
            (i:N:and2[i].vdd = supply.vdd;)
            (i:N:mu2[i].vdd = supply.vdd;)
            (i:((1<<N)-1):dly[i].vdd = supply.vdd;)

            (i:N:and2[i].vss = supply.vss;)
            (i:N:mu2[i].vss = supply.vss;)
            (i:((1<<N)-1):dly[i].vss = supply.vss;)
        }

        // Non programmable delays
        // N is number of delays to have in series (not log!!).
        // Is useful for testing purposes.
        // But should probably remove before running innovus etc.
        export template<pint N>
        defproc delay_chain (bool out; bool in; power supply) {
            { N >= 0 : "What?" };
            [N >= 1 ->
            DLY4_X1 dly[N];

            dly[0].vdd = supply.vdd;
            dly[0].vss = supply.vss;
            dly[0].a = in;

            (i:1..N-1:
                dly[i].vdd = supply.vdd;
                dly[i].vss = supply.vss;
                dly[i].a = dly[i-1].y;
            )

            dly[N-1].vdd = supply.vdd;
            dly[N-1].vss = supply.vss;
            dly[N-1].y = out;
            [] N = 0 ->
            in = out;
            ]
            
        }


        /**
         * Appends a hard-coded word "VAL" to an input.
         * Works by piping through all sigs, but adding
         * some extra sigs when the input is valid.
         * N is size of channel to pipe through.
         * NVAL is size of word to be put on output.
         * VAL is word to be put on output.
         * Output looks like
         * 0..............N........N+NVAL-1
         * --input_data----LSB....MSB
         *
         */
        export template<pint N, NVAL, VAL>
        defproc append (avMx1of2<N> in; avMx1of2<N+NVAL> out; power supply)
        {
            { N >= 0 : "What?" };
            { NVAL >= 0 : "What?" };
            { VAL < 1<<NVAL : "VAL too big!" };

            // valid tree
            vtree<N> in_val(.supply = supply);
            (i:N:
                in_val.in.d[i].t = in.d.d[i].t;
                in_val.in.d[i].f = in.d.d[i].f;
            )

            // wire through most signals
            (i:N:
                in.d.d[i].t = out.d.d[i].t;
                in.d.d[i].f = out.d.d[i].f;
            )
            in.a = out.a;
            in.v = out.v;

            // appender
            pint bitval;
            sigbuf<NVAL> sb(.in = in_val.out, .supply = supply);
            TIELO_X1 tielows[NVAL];
            (i:NVAL:tielows[i].vss = supply.vss; tielows[i].vdd = supply.vdd;)
            (i:0..NVAL-1:
                bitval = (VAL & ( 1 << i )) >> i;
                [  bitval = 1 ->
                    out.d.d[i+N].t = sb.out[i];
                    out.d.d[i+N].f = tielows[i].y;
                [] bitval = 0 ->
                    out.d.d[i+N].f = sb.out[i];
                    out.d.d[i+N].t = tielows[i].y;
                [] bitval >= 2 -> {false : "fuck"};        
                ]
            )

        }
/**
* Drops bits. Slices lines. Crop in. Enhance.
* Useful if say, have an 8 bit packet coming in, but 
* receiver only needs 3 of them.
* KEEPS all bits between the two bounds.
* e.g. drop_lines(8, 0, 3) would keep lines [0,1,2]
**/
export template<pint N, N0, N1>
defproc slice_data(avMx1of2<N> in; avMx1of2<std::min(N1,N)-std::max(N0,0)> out; power supply) {
    // {N0 >= 0 : "N0 can be minimum 0!"};
    // {N1 <= N : "N1 can be maximum N"};

    pint _N1, _N0;

    _N1 = std::min(N1,N);
    _N0 = std::max(N0,0);


    // BUF_X1 ack_buf(.a = out.a, .y = in.a, .vss = supply.vss, .vdd = supply.vdd);

    vtree<N> in_vt(.in = in.d, .out = in.v, .supply = supply);
    (i:_N1-_N0:
        in.d.d[i + _N0] = out.d.d[i];
    )

    // in.a = out.a;

    A_2C_B_X1 Cel(.c1 = out.a, .c2 =in.v, .y = in.a, .vss = supply.vss, .vdd = supply.vdd);

}


  export template<pint N; pint CONDITION_BIT>
        defproc demux_bit (avMx1of2<N+1> in; avMx1of2<N> out1; avMx1of2<N> out2; bool? reset_B; power supply)
        {
            demux<N> demux(.reset_B = reset_B, .out1=out1, .out2=out2);

            in.d.d[CONDITION_BIT].f = demux.cond.d.d[0].f;
            in.d.d[CONDITION_BIT].t = demux.cond.d.d[0].t;

            A_2C_B_X1 val_Cel(.c1 = demux.in.v, .c2 = demux.cond.v, .y = in.v,
                .vdd = supply.vdd, .vss = supply.vss);

            // Not actually needed bc the current version of demux
            // Something like below should be added once the handshakes are properly decoupled.
            // wires the data and cond ack lines together anyway.
            // A_2C_B_X1 ack_Cel(.c1 = demux.in.a, .c2 = demux.cond.a, .y = in.a,
            //     .vdd = supply.vdd, .vss = supply.vss);

            // in.v = demux.in.v;
            in.a = demux.in.a;

            (i:0..CONDITION_BIT-1:
                in.d.d[i].f = demux.in.d.d[i].f;
                in.d.d[i].t = demux.in.d.d[i].t;)

            (i:CONDITION_BIT+1..N:
                in.d.d[i].f = demux.in.d.d[i-1].f;
                in.d.d[i].t = demux.in.d.d[i-1].t;)

        }

      export template<pint N>
        defproc demux_bit_msb (avMx1of2<N+1> in; avMx1of2<N> out1; avMx1of2<N> out2; bool? reset_B; power supply)
        {
            demux_bit<N,N> demux(.in = in, .out1 = out1, .out2 = out2, .reset_B = reset_B, .out1=out1, .out2=out2);
        }

/**
* Create M sigbufs to buffer an M bool array to N strength.
* Done lazily.
**/
export template<pint M, N>
defproc sigbuf_boolarray(bool? in[M]; bool! out[M]; power supply) {
    sigbuf<N> sb[M];
    (i:M:
        sb[i].in = in[i];
        sb[i].out[0] = out[i];
        sb[i].supply = supply;
    )

}


}}