// ************************************************************************** // // // // eses eses // // eses eses // // eses eseses esesese eses Embedded Systems Group // // ese ese ese ese ese // // ese eseseses eseseses ese Department of Computer Science // // eses eses ese eses // // eses eseses eseseses eses University of Kaiserslautern // // eses eses // // // // ************************************************************************** // // This module implements a carry-lookahead adder for radix-2 numbers // // With a depth of log(N) and a size of O(N), the circuit is therefore optimal// // both concerning the depth and the size. For explanations on the macros to // // define the prefix tree, see the modules on general radix-B arithmetic. // // ************************************************************************** // macro N = 6; // number of digits macro K = log(N); // depth of the carry lookahead tree macro natval(x,m) = sum(i=0..m-1) ((x[i]?1:0) * exp(2,i)); // WidthOfLevel determines the number of wires at level i of the lookahead tree macro WidthOfLevel(i,n) = (i==0 ? n : WidthOfLevel(i-1,(n+1)/2)); macro SumOfWidths(j) = sum(i=0..j) WidthOfLevel(i,N); // index mapping to map the two-dimensional arrays p,g to a one-dimensional ones macro idm(i,j) = (i==0 ? j : j+SumOfWidths(i-1)); // recur and indices are used to determine the carry signals // if i=indices(k) holds, then // - i.0 is the largest exponent s.t. exp(2,i.0) divides k // - i.1 := exp(2,i.0) // - i.2 is determined such that k=i.1*(i.2+1) macro recur(x) = (x.0+1,2*x.1,x.2); macro indices(k) = (k%2==1 ? (0,1,k-1) : recur(indices(k/2))); module NatAddCLAOpt([N]bool ?x,?y,[N+1]bool s){ event [N]bool c; event [SumOfWidths(K)]bool p,g; // first row of generate and propagate signals for(i=0..N-1) { g[idm(0,i)] = x[i] & y[i]; p[idm(0,i)] = x[i] | y[i]; } // deeper levels of generate and propagate signals as parallel prefix tree for(i=0..K-1) let(s = exp(2,i)) let(w1 = WidthOfLevel(i,N)) let(w2 = (w1+1)/2) for(j=0..w2-1) { g[idm(i+1,j)] = (2*j+1==w1 ? g[idm(i,2*j)] : g[idm(i,2*j+1)] | g[idm(i,2*j)] & p[idm(i,2*j+1)]); p[idm(i+1,j)] = (2*j+1==w1 ? p[idm(i,2*j)] : p[idm(i,2*j+1)] & p[idm(i,2*j)]); } // determine the carry signals in terms of g[idm(i,j] and p[idm(i,j] c[0] = false; for(k=1..N-1) let(i = indices(k)) c[k] = g[idm(i.0,i.2)] | c[i.1*i.2] & p[idm(i.0,i.2)]; // compute the sum digits for(i=0..N-1) { s[i] = x[i] xor y[i] xor c[i]; } s[N] = g[idm(K,0)]; // alltogether, the addition operation is performed assert(natval(s,N+1) == natval(x,N) + natval(y,N)); }