/// <summary>
/// Ordering and symbolic analysis for a Cholesky factorization.
/// </summary>
/// <param name="A">Matrix to factorize.</param>
/// <param name="p">Permutation.</param>
private void SymbolicAnalysis(CompressedColumnStorage <Complex> A, int[] p)
{
int n = A.ColumnCount;
var sym = this.S = new SymbolicFactorization();
// Find inverse permutation.
sym.pinv = Permutation.Invert(p);
// C = spones(triu(A(P,P)))
var C = PermuteSym(A, sym.pinv, false);
// Find etree of C.
sym.parent = GraphHelper.EliminationTree(n, n, C.ColumnPointers, C.RowIndices, false);
// Postorder the etree.
var post = GraphHelper.TreePostorder(sym.parent, n);
// Find column counts of chol(C)
var c = GraphHelper.ColumnCounts(SymbolicColumnStorage.Create(C, false), sym.parent, post, false);
sym.cp = new int[n + 1];
// Find column pointers for L
sym.unz = sym.lnz = Helper.CumulativeSum(sym.cp, c, n);
}
/// <summary>
/// Symbolic ordering and analysis for QR.
/// </summary>
private void SymbolicAnalysis(ColumnOrdering order, CompressedColumnStorage <double> A)
{
int m = A.RowCount;
int n = A.ColumnCount;
var sym = this.symFactor = new SymbolicFactorization();
// Fill-reducing ordering
sym.q = AMD.Generate(A, order);
var C = order > 0 ? Permute(A, null, sym.q) : SymbolicColumnStorage.Create(A);
// etree of C'*C, where C=A(:,q)
sym.parent = GraphHelper.EliminationTree(m, n, C.ColumnPointers, C.RowIndices, true);
int[] post = GraphHelper.TreePostorder(sym.parent, n);
sym.cp = GraphHelper.ColumnCounts(C, sym.parent, post, true); // col counts chol(C'*C)
bool ok = C != null && sym.parent != null && sym.cp != null && CountV(C);
if (ok)
{
sym.unz = 0;
for (int k = 0; k < n; k++)
{
sym.unz += sym.cp[k];
}
}
}
/// <summary>
/// Ordering and symbolic analysis for a Cholesky factorization
/// </summary>
/// <param name="order"></param>
/// <param name="A"></param>
/// <returns></returns>
static SymbolicFactorization SymbolicAnalysis(int order, SparseMatrix A)
{
int n = A.n;
int[] c, post, P;
SymbolicFactorization S = new SymbolicFactorization(); // allocate result S
P = Ordering.AMD(order, A); // P = amd(A+A'), or natural
S.pinv = Common.InversePermutation(P, n); // find inverse permutation
if (order != 0 && S.pinv == null)
{
return(null);
}
SparseMatrix C = A.PermuteSym(S.pinv, false); // C = spones(triu(A(P,P)))
S.parent = Common.EliminationTree(C, false); // find etree of C
post = Common.TreePostorder(S.parent, n); // postorder the etree
c = Common.ColumnCounts(C, S.parent, post, false); // find column counts of chol(C)
S.cp = new int[n + 1]; // allocate result S.cp
S.unz = S.lnz = Common.CumulativeSum(S.cp, c, n); // find column pointers for L
return(S);
}
/// <summary>
/// Symbolic ordering and analysis for LU.
/// </summary>
/// <param name="A"></param>
/// <param name="p">Permutation.</param>
private void SymbolicAnalysis(CompressedColumnStorage <Complex> A, int[] p)
{
var sym = this.S = new SymbolicFactorization();
// Fill-reducing ordering
sym.q = p;
// Guess nnz(L) and nnz(U)
sym.unz = sym.lnz = 4 * (A.ColumnPointers[n]) + n;
}
/// <summary>
/// Symbolic ordering and analysis for LU.
/// </summary>
/// <param name="order"></param>
/// <param name="A"></param>
private void SymbolicAnalysis(ColumnOrdering order, CompressedColumnStorage <double> A)
{
var sym = this.symFactor = new SymbolicFactorization();
// Fill-reducing ordering
sym.q = AMD.Generate(A, order);
// Guess nnz(L) and nnz(U)
sym.unz = sym.lnz = 4 * (A.ColumnPointers[n]) + n;
}
/// <summary>
/// Symbolic ordering and analysis for LU.
/// </summary>
/// <param name="order"></param>
/// <param name="A"></param>
/// <returns></returns>
static SymbolicFactorization SymbolicAnalysis(int order, SparseMatrix A)
{ // ori: cs_sqr
int n = A.n;
SymbolicFactorization S = new SymbolicFactorization(); // allocate result S
S.q = Ordering.AMD(order, A); // fill-reducing ordering
if (order != 0 && S.q == null)
{
return(null);
}
S.unz = 4 * (A.p[n]) + n; // For LU factorization only
S.lnz = S.unz; // Guess nnz(L) and nnz(U)
return(S); // return result S
}
/// <summary>
/// Symbolic ordering and analysis for QR
/// </summary>
static SymbolicFactorization SymbolicAnalysis(int order, SparseMatrix A)
{ // ori: cs_sqr
int n = A.n, k;
bool ok = true;
int[] post;
SymbolicFactorization S = new SymbolicFactorization(); // allocate result S
S.q = Ordering.AMD(order, A); // fill-reducing ordering
if (order != 0 && S.q == null)
{
return(null);
}
SparseMatrix C = order > 0 ? (SparseMatrix)A.Permute(null, S.q, false) : A;
S.parent = Common.EliminationTree(C, true); // etree of C'*C, where C=A(:,q)
post = Common.TreePostorder(S.parent, n);
S.cp = Common.ColumnCounts(C, S.parent, post, true); // col counts chol(C'*C)
ok = C != null && S.parent != null && S.cp != null && CountV(C, S);
if (ok)
{
for (S.unz = 0, k = 0; k < n; k++)
{
S.unz += S.cp[k];
}
}
ok = ok && S.lnz >= 0 && S.unz >= 0; // int overflow guard
return(ok ? S : null); // return result S
}
/// <summary>
/// Ordering and symbolic analysis for a LDL' factorization.
/// </summary>
/// <param name="order">Column ordering.</param>
/// <param name="A">Matrix to factorize.</param>
private void SymbolicAnalysis(ColumnOrdering order, CompressedColumnStorage <double> A)
{
int n = A.ColumnCount;
var sym = this.S = new SymbolicFactorization();
var ap = A.ColumnPointers;
var ai = A.RowIndices;
// P = amd(A+A') or natural
var P = AMD.Generate(A, order);
var Pinv = Permutation.Invert(P);
// Output: column pointers and elimination tree.
var lp = new int[n + 1];
var parent = new int[n];
// Workspace
var lnz = new int[n];
var flag = new int[n];
int i, k, p, kk, p2;
for (k = 0; k < n; k++)
{
// L(k,:) pattern: all nodes reachable in etree from nz in A(0:k-1,k)
parent[k] = -1; // parent of k is not yet known
flag[k] = k; // mark node k as visited
lnz[k] = 0; // count of nonzeros in column k of L
kk = (P != null) ? (P[k]) : (k); // kth original, or permuted, column
p2 = ap[kk + 1];
for (p = ap[kk]; p < p2; p++)
{
// A(i,k) is nonzero (original or permuted A)
i = (Pinv != null) ? (Pinv[ai[p]]) : (ai[p]);
if (i < k)
{
// follow path from i to root of etree, stop at flagged node
for (; flag[i] != k; i = parent[i])
{
// find parent of i if not yet determined
if (parent[i] == -1)
{
parent[i] = k;
}
lnz[i]++; // L(k,i) is nonzero
flag[i] = k; // mark i as visited
}
}
}
}
// construct Lp index array from Lnz column counts
lp[0] = 0;
for (k = 0; k < n; k++)
{
lp[k + 1] = lp[k] + lnz[k];
}
sym.parent = parent;
sym.cp = lp;
sym.q = P;
sym.pinv = Pinv;
}
/// <summary>
/// Compute the Numeric Cholesky factorization, L = chol (A, [pinv parent cp]).
/// </summary>
/// <returns>Numeric Cholesky factorization</returns>
public static NumericFactorization Factorize(SparseMatrix A, SymbolicFactorization S)
{
Complex d, lki;
Complex[] Lx, x, Cx;
int top, i, p, k;
int[] Li, Lp, cp, pinv, s, c, parent, Cp, Ci;
SparseMatrix L, C;
NumericFactorization N = new NumericFactorization(); // allocate result
int n = A.n;
c = new int[n]; // get int workspace
s = new int[n];
x = new Complex[n]; // get double workspace
cp = S.cp;
pinv = S.pinv;
parent = S.parent;
C = pinv != null?A.PermuteSym(pinv, true) : A;
Cp = C.p;
Ci = C.i;
Cx = C.x;
N.L = L = new SparseMatrix(n, n, cp[n], true); // allocate result
Lp = L.p;
Li = L.i;
Lx = L.x;
for (k = 0; k < n; k++)
{
Lp[k] = c[k] = cp[k];
}
for (k = 0; k < n; k++) // compute L(k,:) for L*L' = C
{
// Nonzero pattern of L(k,:)
top = Common.EReach(C, k, parent, s, c); // find pattern of L(k,:)
x[k] = 0; // x (0:k) is now zero
for (p = Cp[k]; p < Cp[k + 1]; p++) // x = full(triu(C(:,k)))
{
if (Ci[p] <= k)
{
x[Ci[p]] = Cx[p];
}
}
d = x[k]; // d = C(k,k)
x[k] = 0; // clear x for k+1st iteration
// Triangular solve
for (; top < n; top++) // solve L(0:k-1,0:k-1) * x = C(:,k)
{
i = s[top]; // s [top..n-1] is pattern of L(k,:)
lki = x[i] / Lx[Lp[i]]; // L(k,i) = x (i) / L(i,i)
x[i] = 0; // clear x for k+1st iteration
for (p = Lp[i] + 1; p < c[i]; p++)
{
x[Li[p]] -= Lx[p] * lki;
}
d -= lki * Complex.Conjugate(lki); // d = d - L(k,i)*L(k,i)
p = c[i]++;
Li[p] = k; // store L(k,i) in column i
Lx[p] = Complex.Conjugate(lki);
}
// Compute L(k,k)
if (d.Real <= 0 || d.Imaginary != 0)
{
return(null); // not pos def
}
p = c[k]++;
Li[p] = k; // store L(k,k) = sqrt (d) in column k
Lx[p] = Complex.Sqrt(d);
}
Lp[n] = cp[n]; // finalize L
return(N); // success
}
/// <summary>
/// Sparse QR factorization [V,beta,pinv,R] = qr (A)
/// </summary>
static NumericFactorization Factorize(SparseMatrix A, SymbolicFactorization S)
{
double[] Beta;
int i, p, p1, top, len, col;
int m = A.m;
int n = A.n;
int[] Ap = A.p;
int[] Ai = A.i;
Complex[] Ax = A.x;
int[] q = S.q;
int[] parent = S.parent;
int[] pinv = S.pinv;
int m2 = S.m2;
int vnz = S.lnz;
int rnz = S.unz;
int[] leftmost = S.leftmost;
int[] w = new int[m2 + n]; // get int workspace
Complex[] x = new Complex[m2]; // get double workspace
int s = m2; // offset into w
SparseMatrix R, V;
NumericFactorization N = new NumericFactorization(); // allocate result
N.L = V = new SparseMatrix(m2, n, vnz, true); // allocate result V
N.U = R = new SparseMatrix(m2, n, rnz, true); // allocate result R
N.B = Beta = new double[n]; // allocate result Beta
int[] Rp = R.p;
int[] Ri = R.i;
Complex[] Rx = R.x;
int[] Vp = V.p;
int[] Vi = V.i;
Complex[] Vx = V.x;
for (i = 0; i < m2; i++)
{
w[i] = -1; // clear w, to mark nodes
}
rnz = 0; vnz = 0;
for (int k = 0; k < n; k++) // compute V and R
{
Rp[k] = rnz; // R(:,k) starts here
Vp[k] = p1 = vnz; // V(:,k) starts here
w[k] = k; // add V(k,k) to pattern of V
Vi[vnz++] = k;
top = n;
col = q != null ? q[k] : k;
for (p = Ap[col]; p < Ap[col + 1]; p++) // find R(:,k) pattern
{
i = leftmost[Ai[p]]; // i = min(find(A(i,q)))
for (len = 0; w[i] != k; i = parent[i]) // traverse up to k
{
//len++;
w[s + len++] = i;
w[i] = k;
}
while (len > 0)
{
--top;
--len;
w[s + top] = w[s + len]; // push path on stack
}
i = pinv[Ai[p]]; // i = permuted row of A(:,col)
x[i] = Ax[p]; // x (i) = A(:,col)
if (i > k && w[i] < k) // pattern of V(:,k) = x (k+1:m)
{
Vi[vnz++] = i; // add i to pattern of V(:,k)
w[i] = k;
}
}
for (p = top; p < n; p++) // for each i in pattern of R(:,k)
{
i = w[s + p]; // R(i,k) is nonzero
ApplyHouseholder(V, i, Beta[i], x); // apply (V(i),Beta(i)) to x
Ri[rnz] = i; // R(i,k) = x(i)
Rx[rnz++] = x[i];
x[i] = 0;
if (parent[i] == k)
{
vnz = V.Scatter(i, 0, w, null, k, V, vnz);
}
}
for (p = p1; p < vnz; p++) // gather V(:,k) = x
{
Vx[p] = x[Vi[p]];
x[Vi[p]] = 0;
}
Ri[rnz] = k; // R(k,k) = norm (x)
Rx[rnz++] = CreateHouseholder(Vx, p1, ref Beta[k], vnz - p1); // [v,beta]=house(x)
}
Rp[n] = rnz; // finalize R
Vp[n] = vnz; // finalize V
return(N); // success
}
/// <summary>
/// Compute nnz(V) = S.lnz, S.pinv, S.leftmost, S.m2 from A and S.parent
/// </summary>
static bool CountV(SparseMatrix A, SymbolicFactorization S)
{
int i, k, p, pa, n = A.n, m = A.m;
int[] Ap = A.p, Ai = A.i, head,
tail, nque, pinv, leftmost, w, parent = S.parent;
S.pinv = pinv = new int[m + n]; // allocate pinv,
S.leftmost = leftmost = new int[m]; // and leftmost
w = new int[m]; // get workspace
head = new int[n];
tail = new int[n];
nque = new int[n]; // Initialized to 0's
for (k = 0; k < n; k++)
{
head[k] = -1; // queue k is empty
}
for (k = 0; k < n; k++)
{
tail[k] = -1;
}
for (i = 0; i < m; i++)
{
leftmost[i] = -1;
}
for (k = n - 1; k >= 0; k--)
{
for (p = Ap[k]; p < Ap[k + 1]; p++)
{
leftmost[Ai[p]] = k; // leftmost[i] = min(find(A(i,:)))
}
}
for (i = m - 1; i >= 0; i--) // scan rows in reverse order
{
pinv[i] = -1; // row i is not yet ordered
k = leftmost[i];
if (k == -1)
{
continue; // row i is empty
}
if (nque[k]++ == 0)
{
tail[k] = i; // first row in queue k
}
w[i] = head[k]; // put i at head of queue k
head[k] = i;
}
S.lnz = 0;
S.m2 = m;
for (k = 0; k < n; k++) // find row permutation and nnz(V)
{
i = head[k]; // remove row i from queue k
S.lnz++; // count V(k,k) as nonzero
if (i < 0)
{
i = S.m2++; // add a fictitious row
}
pinv[i] = k; // associate row i with V(:,k)
if (--nque[k] <= 0)
{
continue; // skip if V(k+1:m,k) is empty
}
S.lnz += nque[k]; // nque [k] is nnz (V(k+1:m,k))
if ((pa = parent[k]) != -1) // move all rows to parent of k
{
if (nque[pa] == 0)
{
tail[pa] = tail[k];
}
w[tail[k]] = head[pa];
head[pa] = w[i];
nque[pa] += nque[k];
}
}
for (i = 0; i < m; i++)
{
if (pinv[i] < 0)
{
pinv[i] = k++;
}
}
return(true);
}
// [L,U,pinv]=lu(A, [q lnz unz]). lnz and unz can be guess
static NumericFactorization Factorize(SparseMatrix A, SymbolicFactorization S, double tol)
{
int i, n = A.n;
int[] q = S.q;
int lnz = S.lnz;
int unz = S.unz;
int[] pinv;
NumericFactorization N = new NumericFactorization(); // allocate result
SparseMatrix L, U;
N.L = L = new SparseMatrix(n, n, lnz, true); // allocate result L
N.U = U = new SparseMatrix(n, n, unz, true); // allocate result U
N.pinv = pinv = new int[n]; // allocate result pinv
double[] x = new double[n]; // get double workspace
int[] xi = new int[2 * n]; // get int workspace
for (i = 0; i < n; i++)
{
pinv[i] = -1; // no rows pivotal yet
}
lnz = unz = 0;
int ipiv, top, p, col;
double pivot, a, t;
int[] Li, Ui, Lp = L.p, Up = U.p;
double[] Lx, Ux;
for (int k = 0; k < n; k++) // compute L(:,k) and U(:,k)
{
// Triangular solve
Lp[k] = lnz; // L(:,k) starts here
Up[k] = unz; // U(:,k) starts here
if ((lnz + n > L.nzmax && !L.Resize(2 * L.nzmax + n)) ||
(unz + n > U.nzmax && !U.Resize(2 * U.nzmax + n)))
{
return(null);
}
Li = L.i;
Lx = L.x;
Ui = U.i;
Ux = U.x;
col = q != null ? (q[k]) : k;
top = Triangular.SolveSp(L, A, col, xi, x, pinv, true); // x = L\A(:,col)
// Find pivot
ipiv = -1;
a = -1;
for (p = top; p < n; p++)
{
i = xi[p]; // x(i) is nonzero
if (pinv[i] < 0) // row i is not yet pivotal
{
if ((t = Math.Abs(x[i])) > a)
{
a = t; // largest pivot candidate so far
ipiv = i;
}
}
else // x(i) is the entry U(pinv[i],k)
{
Ui[unz] = pinv[i];
Ux[unz++] = x[i];
}
}
if (ipiv == -1 || a <= 0)
{
return(null);
}
if (pinv[col] < 0 && Math.Abs(x[col]) >= a * tol)
{
ipiv = col;
}
// Divide by pivot
pivot = x[ipiv]; // the chosen pivot
Ui[unz] = k; // last entry in U(:,k) is U(k,k)
Ux[unz++] = pivot;
pinv[ipiv] = k; // ipiv is the kth pivot row
Li[lnz] = ipiv; // first entry in L(:,k) is L(k,k) = 1
Lx[lnz++] = 1;
for (p = top; p < n; p++) // L(k+1:n,k) = x / pivot
{
i = xi[p];
if (pinv[i] < 0) // x(i) is an entry in L(:,k)
{
Li[lnz] = i; // save unpermuted row in L
Lx[lnz++] = x[i] / pivot; // scale pivot column
}
x[i] = 0; // x [0..n-1] = 0 for next k
}
}
// Finalize L and U
Lp[n] = lnz;
Up[n] = unz;
Li = L.i; // fix row indices of L for final pinv
for (p = 0; p < lnz; p++)
{
Li[p] = pinv[Li[p]];
}
L.Resize(0); // remove extra space from L and U
U.Resize(0);
return(N); // success
}
public static CCS GenerateP_Delta_Mpc(Model target, LoadCase loadCase, IRrefFinder rrefFinder)
{
target.ReIndexNodes();
var n = target.Nodes.Count;
var boundaryConditions = GetModelBoundaryConditions(target, loadCase);
var lastRow = 0;
#region step 1
//step 1: combine all eqs to one system
//var filledCols = new bool[6 * n];
var extraEqCount = 0;
foreach (var mpcElm in target.MpcElements)
{
if (mpcElm.AppliesForLoadCase(loadCase))
{
extraEqCount += mpcElm.GetExtraEquationsCount();
}
}
extraEqCount += boundaryConditions.RowCount;
var allEqsCrd = new CoordinateStorage <double>(extraEqCount, n * 6 + 1, 1);//rows: extra eqs, cols: 6*n+1 (+1 is for right hand side)
foreach (var mpcElm in target.MpcElements)
{
if (mpcElm.AppliesForLoadCase(loadCase))
{
var extras = mpcElm.GetExtraEquations();
extras.EnumerateMembers((row, col, val) =>
{
allEqsCrd.At(row + lastRow, col, val);
});
lastRow += extras.RowCount;
}
}
{
boundaryConditions.EnumerateMembers((row, col, val) =>
{
allEqsCrd.At(row + lastRow, col, val);
});
lastRow += boundaryConditions.RowCount;
}
var allEqs = allEqsCrd.ToCCs();
#endregion
#region comment
/*
#region step 2
* //step 2: create adjacency matrix of variables
*
* //step 2-1: find nonzero pattern
* var allEqsNonzeroPattern = allEqs.Clone();
*
* for (var i = 0; i < allEqsNonzeroPattern.Values.Length; i++)
* allEqsNonzeroPattern.Values[i] = 1;
*
* //https://math.stackexchange.com/questions/2340450/extract-independent-sub-systems-from-a-bigger-linear-eq-system
* var tmp = allEqsNonzeroPattern.Transpose();
*
* var variableAdj = tmp.Multiply(allEqsNonzeroPattern);
#endregion
*
#region step 3
* //extract parts
* var parts = EnumerateGraphParts(variableAdj);
*
#endregion
*
#region step 4
* {
*
* allEqs.EnumerateColumns((colNum, vals) =>
* {
* if (vals.Count == 0)
* Console.WriteLine("Col {0} have {1} nonzeros", colNum, vals.Count);
* });
*
* var order = ColumnOrdering.MinimumDegreeAtPlusA;
*
* // Partial pivoting tolerance (0.0 to 1.0)
* double tolerance = 1.0;
*
* var lu = CSparse.Double.Factorization.SparseLU.Create(allEqs, order, tolerance);
*
* }
*
#endregion
*/
#endregion
rrefFinder.CalculateRref(allEqs);
var q = AMD.Generate(allEqs, ColumnOrdering.MinimumDegreeAtA);
var s = new SymbolicFactorization()
{
q = q
};
var qr = CSparse.Double.Factorization.SparseQR.Create(allEqs, ColumnOrdering.MinimumDegreeAtA);
var r = ((CCS)ReflectionUtils.GetFactorR(qr, "R"));
//var q = ((CCS)ReflectionUtils.GetFactorR(qr, "Q"));
//var s = ((SymbolicFactorization)ReflectionUtils.GetFactorR(qr, "S"));
var idependents = new bool[allEqs.RowCount];
var rd = allEqs.ToDenseMatrix();
r.EnumerateMembers((row, col, val) =>
{
if (row == col)
{
if (Math.Abs(val) < 1e-6)
{
return;
}
}
idependents[row] = true;
}
);
//var t=qr.
throw new NotImplementedException();
//return buf.ToCCs();
}
