Skip to content

I
imagej-elphel
Commit 3cedfc7a authored by Andrey Filippov's avatar Andrey Filippov
Browse files
Options

Found multithreaded Cholesky almost 10 times faster than single-threaded 

for n=2300, 16 threads
parent 663a964a
Hide whitespace changes
Inline Side-by-side
Showing with 1064 additions and 169 deletions
src/main/java/com/elphel/imagej/common/CholeskyBlock.java 0 → 100644
View file @ 3cedfc7a
package com.elphel.imagej.common;
import java.util.Arrays;
import java.util.concurrent.atomic.AtomicInteger;
import com.elphel.imagej.tileprocessor.ImageDtt;
import Jama.Matrix;
public class CholeskyBlock {
public final int m; // tile size
public final int np; // number of elements in row/col
public final int n; // number of tiles in row/col
public final int nf; // number of full rows/cols
public final double [] A;
public final double [] L;
public CholeskyBlock (
double [][] A_in,
int size) {
m = size;
np = A_in.length;
nf = np / m;
n = ((nf * m) < np)? (nf + 1) : nf;
A = new double [n*n*m*m];
L = new double [A.length];
setup_ATriangle(A_in);
}
private void setup_ATriangle(double [][] A_in) {
for (int tile_row = 0; tile_row < nf; tile_row++) {
for (int tile_col = 0; tile_col < tile_row; tile_col++) {
int indx = indx_IJ(tile_row, tile_col);
for (int k = 0; k < m; k++) {
System.arraycopy(
A_in[tile_row*m +k],
tile_col * m,
A,
indx + m * k,
m);
}
}
// copy diagonal
int indx = indx_IJ(tile_row, tile_row);
for (int k = 0; k < m; k++) {
System.arraycopy(
A_in[tile_row*m +k],
tile_row * m,
A,
indx + m * k,
k+1);
}
}
if (n > nf) { // if there are small tiles below and to the right
int tile_row = nf;
int m1 = np - m * nf;
for (int tile_col = 0; tile_col < nf; tile_col++) {
int indx = indx_IJ(tile_row, tile_col);
for (int k = 0; k < m1; k++) {
System.arraycopy(
A_in[tile_row*m +k],
tile_col * m,
A,
indx + m * k,
m);
}
}
// copy diagonal
int indx = indx_IJ(tile_row, tile_row);
for (int k = 0; k < m1; k++) {
System.arraycopy(
A_in[tile_row*m +k],
tile_row * m,
A,
indx + m1 * k,
k+1);
}
}
}
public Matrix getL() {
return new Matrix(get_LTriangle(),np,np);
}
public double [][] get_LTriangle() {
double [][] L_out = new double[np][np];
for (int tile_row = 0; tile_row < nf; tile_row++) {
for (int tile_col = 0; tile_col < tile_row; tile_col++) {
int indx = indx_IJ(tile_row, tile_col);
for (int k = 0; k < m; k++) {
System.arraycopy(
L,
indx + m * k,
L_out[tile_row*m +k],
tile_col * m,
m);
}
}
// copy diagonal
int indx = indx_IJ(tile_row, tile_row);
for (int k = 0; k < m; k++) {
System.arraycopy(
L,
indx + m * k,
L_out[tile_row*m +k],
tile_row * m,
k+1);
}
}
if (n > nf) { // if there are small tiles below and to the right
int tile_row = nf;
int m1 = np - m * nf;
for (int tile_col = 0; tile_col < nf; tile_col++) {
int indx = indx_IJ(tile_row, tile_col);
for (int k = 0; k < m1; k++) {
System.arraycopy(
L,
indx + m * k,
L_out[tile_row*m +k],
tile_col * m,
m);
}
}
// copy diagonal
int indx = indx_IJ(tile_row, tile_row);
for (int k = 0; k < m1; k++) {
System.arraycopy(
L,
indx + m1 * k,
L_out[tile_row*m +k],
tile_row * m,
k+1);
}
}
return L_out;
}
/**
* Get index of the top-left tile corner
* @param i tile row
* @param j tile column
* @return index in A and L arrays
*/
public int indx_IJ(int i, int j) {
return j * (m * np) + ((j >= nf) ? (np-nf*m): m) * m * i;
}
/**
* Get a new diagonal square submatrix
* @param arr A or L packed array (line-scan order for each tile column)
* @param i tile index on the diagonal
* @return new array with tile data (m x m, or smaller for the bottom right)
*/
public double [][] getDiagSquare(
double [] arr,
int i){
int l = (i >= nf) ? (np-nf*m): m;
double [][] a_diag = new double[l][l];
return getDiagSquare (arr, a_diag, i);
}
/**
* Get a new diagonal square submatrix
* @param arr A or L packed array (line-scan order for each tile column)
* @param a_diag - preallocated array (should be smaller if needed)
* @param i tile index on the diagonal
* @return new array with tile data (m x m, or smaller for the bottom right)
*/
public double [][] getDiagSquare(
double [] arr,
double [][] a_diag,
int i){
int indx = indx_IJ(i,i);
int l = a_diag.length;
for (int k = 0; k < l; k++) {
int kl = k*l;
System.arraycopy( arr, indx+kl, a_diag[k], 0, l);
}
return a_diag;
}
/**
* Get a new diagonal square submatrix, copy only lower triangle including diagonal
* @param arr A or L packed array (line-scan order for each tile column)
* @param i tile index on the diagonal
* @return new array with tile data (m x m, or smaller for the bottom right)
*/
public double [][] getDiagTriangle(
double [] arr,
int i){
int l = (i >= nf) ? (np-nf*m): m;
double [][] a_diag = new double[l][l];
return getDiagLTriangle (arr, a_diag, i);
}
/**
* Get a new diagonal square submatrix, copy only lower triangle including diagonal
* @param arr A or L packed array (line-scan order for each tile column)
* @param a_diag - preallocated array (should be smaller if needed)
* @param i tile index on the diagonal
* @return new array with tile data (m x m, or smaller for the bottom right)
*/
public double [][] getDiagLTriangle(
double [] arr,
double [][] a_diag,
int i){
int indx = indx_IJ(i,i);
int l = a_diag.length;
for (int k = 0; k < l; k++) {
int kl = k*l;
System.arraycopy( arr, indx+kl, a_diag[k], 0, k+1);
}
return a_diag;
}
/**
* Save calculated tile L lower diagonal matrix to a packed array
* @param arr A or L packed array (line-scan order for each tile column)
* @param l_diag lower triangular array with Cholesky decomposition
* @param i tile index on a diagonal
*/
public void putDiagLTriangle(
double [] arr,
double [][] l_diag,
int i) {
int indx = indx_IJ(i,i);
int l = l_diag.length;
for (int k = 0; k < l; k++) {
int kl = k*l;
System.arraycopy(l_diag[k], 0, arr, indx+kl, k+1);
}
}
public void setL21(
int i, // i > j,
int j) { // j <nf
int indx_diag = indx_IJ(j,j);
int indx_ij = indx_IJ(i,j);
int h = (i < nf) ? m : (np-nf*m);
// prepare solving Lx = b, copy tile A -> L
System.arraycopy(A, indx_ij, L, indx_ij, m * h);
for (int l_row = 0; l_row < m; l_row++) {
for (int x_col= 0; x_col < h; x_col++) { // b-vector
int lindx = indx_ij + m * x_col + l_row;
double ls = L[lindx];
for (int l_col = 0; l_col < l_row; l_col++) {
ls -= L[indx_ij + m * x_col + l_col] * L[indx_diag + m* l_row+l_col];
}
L[lindx] = ls/L[indx_diag + (m + 1)* l_row];
}
}
return;
}
public void setA22(
int diag,// < col, < row
int row, // >= col
int col) {
int h = (row < nf) ? m : (np-nf*m);
int indx_a = indx_IJ(row,col);
int indx_lrow = indx_IJ(row,diag);
if (row == col) {
for (int i = 0; i < h; i++) {
for (int j = 0; j <= i; j++) {
for (int k = 0; k < h; k++) {
A[indx_a + i * m + j] -= L[indx_lrow + i * h + k] * L[indx_lrow + j * h + k];
}
}
}
} else {
int indx_lcol = indx_IJ(col,diag);
for (int i = 0; i < h; i++) {
for (int j = 0; j < m; j++) {
for (int k = 0; k < m; k++) {
A[indx_a + i * m + j] -= L[indx_lrow + i * m + k] * L[indx_lcol + j * m + k];
}
}
}
}
return;
}
public void choleskyBlockMulti() {
final Thread[] threads = ImageDtt.newThreadArray();
final AtomicInteger ai = new AtomicInteger(0);
final double [][] Am = new double[m][m];
final double [][] Ah = (n > nf) ? new double[(np-nf*m)][(np-nf*m)] : null;
final double [][] A1 = (np < m) ? Ah : Am; // smaller than a tile
final double [][] Lm = new double[m][m];
final double [][] Lh = (n > nf) ? new double[(np-nf*m)][(np-nf*m)] : null;
final double [][] L1 = (np < m) ? Lh : Lm; // smaller than a tile
// Extract top-left tile (only lower triangle)
getDiagLTriangle(
A, // double [] arr,
A1, // double [][] a_diag,
0); // int i)
// Decompose, get L (lower triangle)
cholesky_single (
A1, // double [][] a,
L1); // double [][] l)
// Save it to L-array:
putDiagLTriangle(
L, // double [] arr,
L1, // double [][] l_diag,
0); // int i)
// Calculate first column under diagonal (L21) - maybe use m
ai.set(1); // start from the second tile row
for (int ithread = 0; ithread < threads.length; ithread++) { // first sum for pairs
threads[ithread] = new Thread() {
public void run() {
for (int tile_row = ai.getAndIncrement(); tile_row < n; tile_row = ai.getAndIncrement()) {
setL21(tile_row, 0);
}
}
};
}
ImageDtt.startAndJoin(threads);
for (int tile_diag = 1; tile_diag < n; tile_diag++) {
final int ftile_diag = tile_diag;
// Calculate A in one tile column of the remaining A2'
// start with diagonal (top) tile, and calculate its Cholesky
// In parallel, calculate A for all tiles in that column below diagonal
ai.set(ftile_diag);
for (int ithread = 0; ithread < threads.length; ithread++) { // first sum for pairs
threads[ithread] = new Thread() {
public void run() {
for (int nRow= ai.getAndIncrement(); nRow < n; nRow = ai.getAndIncrement()) {
setA22(
ftile_diag-1, // int diag,// < col, < row
nRow, // int row, // >= col
ftile_diag); // int col)
if (nRow == ftile_diag) {
double [][] At = (ftile_diag < nf) ? Am: Ah; // full or reduced array
double [][] Lt = (ftile_diag < nf) ? Lm: Lh; // full or reduced array
// Extract top-left tile (only lower triangle)
getDiagLTriangle(
A, // double [] arr,
At, // double [][] a_diag,
ftile_diag); // int i)
// Decompose, get L (lower triangle)
cholesky_single (
At, // double [][] a,
Lt); // double [][] l)
// Save it to L-array:
putDiagLTriangle(
L, // double [] arr,
Lt, // double [][] l_diag,
ftile_diag); // int i)
/*
// Calculate first column under diagonal (L21) - maybe use m
for (int tr = ftile_diag; tr < n; tr++) {
setL21(tr, ftile_diag);
}
*/
}
}
}
};
}
ImageDtt.startAndJoin(threads);
if (ftile_diag < (n-1)) {
// Now in parallel calculate L in column ftile_diag under diagonal and
// finish A2' to the right of ftile_diag column
final int left_rows = n - ftile_diag - 1;
final int num_tiles = left_rows * (left_rows + 1) / 2 + 1;
ai.set(0);
for (int ithread = 0; ithread < threads.length; ithread++) { // first sum for pairs
threads[ithread] = new Thread() {
public void run() {
for (int ntile = ai.getAndIncrement(); ntile < num_tiles; ntile = ai.getAndIncrement()) {
if (ntile == 0) {
// Calculate first column of L under diagonal (L21) - maybe use m
for (int tr = ftile_diag + 1; tr < n; tr++) {
setL21(
tr, // row > column
ftile_diag); // column
}
} else {
int nrow = (int) Math.floor(-0.5 + 0.5* Math.sqrt(1 + 8 * (ntile-1)));
int ncol = (ntile-1) - (nrow * (nrow + 1) /2);
int row = ftile_diag + nrow + 1;
int col = ftile_diag + ncol + 1;
setA22(
ftile_diag-1, // int diag,// < col, < row
row, // int row, // >= col
col); // int col)
}
}
}
};
}
ImageDtt.startAndJoin(threads);
}
}
}
// Cholesky-Banachiewicz Algorithm ?
// Single-threaded
public static double [][] cholesky_single (
double [][] a,
double [][] l) {
int n = a.length;
// double [][] l = new double[n][n];
for (int j = 0; j < n; j++) {
Arrays.fill(l[j], 0.0);
}
// Main loop.
for (int j = 0; j < n; j++) {
double[] Lrowj = l[j];
double d = 0.0;
for (int k = 0; k < j; k++) {
double[] Lrowk = l[k];
double s = 0.0;
for (int i = 0; i < k; i++) {
s += Lrowk[i]*Lrowj[i];
}
Lrowj[k] = s = (a[j][k] - s)/l[k][k];
d = d + s*s;
}
d = a[j][j] - d;
l[j][j] = Math.sqrt(Math.max(d,0.0));
}
return l;
}
}
src/main/java/com/elphel/imagej/common/CholeskyLDLTMulti.java
View file @ 3cedfc7a
...@@ -7,6 +7,7 @@ import com.elphel.imagej.tileprocessor.ImageDtt; ...@@ -7,6 +7,7 @@ import com.elphel.imagej.tileprocessor.ImageDtt;
import Jama.CholeskyDecomposition; import Jama.CholeskyDecomposition;
import Jama.Matrix; import Jama.Matrix;
import ij.ImagePlus;
public class CholeskyLDLTMulti { public class CholeskyLDLTMulti {
public final int n; public final int n;
...@@ -14,178 +15,149 @@ public class CholeskyLDLTMulti { ...@@ -14,178 +15,149 @@ public class CholeskyLDLTMulti {
public int solve_step = 10; public int solve_step = 10;
public double thread_scale = 2.5; // split jobs in thread_scale * num_treads chunks public double thread_scale = 2.5; // split jobs in thread_scale * num_treads chunks
// public double [][] L;
public CholeskyLDLTMulti(Matrix matA, boolean b) {
boolean debug = true; public CholeskyLDLTMulti(Matrix matA) {
int dbg_j = 100; // 1520;
n = matA.getRowDimension(); n = matA.getRowDimension();
L = matA.getArray(); // will be modified, copy externally if needed L = CholeskyLDLTMulti_multi(matA);
final Thread[] threads = ImageDtt.newThreadArray();
final AtomicInteger ai = new AtomicInteger(0);
for (int j = 0; j < n; j++) {
if (j == dbg_j) {
System.out.println("CholeskyLDLTMulti(): j=dbg_j="+j);
}
final int fj=j;
for (int k = 0; k < j; k++) {
final int fk=k;
L[j][j] -= L[j][k]*L[j][k]; // diagonal
final double Ljk = L[j][k];
// for (int i = j+1; i < n; i++) {
// l[i][j] -= l[i][k]*l[j][k]; // not including diagonal
// }
// parallel
ai.set(j+1);
for (int ithread = 0; ithread < threads.length; ithread++) {
threads[ithread] = new Thread() {
public void run() {
for (int i = ai.getAndIncrement(); i < n; i = ai.getAndIncrement()) {
L[i][fj] -= L[i][fk]*Ljk;
}
}
};
}
ImageDtt.startAndJoin(threads);
}
// parallel
// for (int i = j+1; i < n; i++) {
// l[i][j] /= l[j][j];
// }
final double Ljj = L[j][j];
ai.set(j+1);
for (int ithread = 0; ithread < threads.length; ithread++) {
threads[ithread] = new Thread() {
public void run() {
for (int i = ai.getAndIncrement(); i < n; i = ai.getAndIncrement()) {
L[i][fj] /= Ljj;
}
}
};
}
ImageDtt.startAndJoin(threads);
if (debug) {
System.out.println("CholeskyLDLTMulti(): j="+j);
}
}
return;
} }
// L overwites strict lower A, d overwrites diagonal A public CholeskyLDLTMulti(Matrix matA, int mode) {
public CholeskyLDLTMulti(Matrix matA, int[] ii) {
// boolean debug = true;
int dbg_j = 100; // 1520;
n = matA.getRowDimension(); n = matA.getRowDimension();
// not needed switch (mode) {
L = matA.getArray(); // will be modified, copy externally if needed case 0: L = CholeskyLDLTMulti_single(matA); break;
final double [][] a = matA.getArray(); case 1: L = CholeskyLDLTMulti_multi(matA); break;
final Thread[] threads = ImageDtt.newThreadArray(); case 2: L = CholeskyLDLTMulti_fast(matA); break;
final AtomicInteger ai = new AtomicInteger(0); default:L = CholeskyLDLTMulti_single(matA);
}
}
public double[][] CholeskyLDLTMulti_multi(Matrix matA) {
final double [][] a = matA.getArray();
final Thread[] threads = ImageDtt.newThreadArray();
final AtomicInteger ai = new AtomicInteger(0);
final AtomicInteger ati = new AtomicInteger(0);
final double [] ajj_threaded = new double[threads.length];
final int threads_chunk = (int) (n / (thread_scale * threads.length));
for (int j = 0; j < n; j++) { for (int j = 0; j < n; j++) {
final int fj = j; final int fj = j;
if (j == dbg_j) { final double [] aj = a[j];
System.out.println("CholeskyLDLTMulti(): j=dbg_j="+j); double ajj = aj[j];
} if (j > threads_chunk) { // use multithreaded
/* Arrays.fill(ajj_threaded, 0.0);
if (j > 0) {
ai.set(0); ai.set(0);
for (int ithread = 0; ithread < threads.length; ithread++) { ati.set(0);
for (int ithread = 0; ithread < threads.length; ithread++) { // first sum for pairs
threads[ithread] = new Thread() { threads[ithread] = new Thread() {
public void run() { public void run() {
for (int i = ai.getAndIncrement(); i < fj; i = ai.getAndIncrement()) { int nthread = ati.getAndIncrement();
a[fj][fj] -= a[i][i]*a[fj][i]*a[fj][i]; // squared for (int i0 = ai.getAndAdd(threads_chunk); i0 < fj; i0 = ai.getAndAdd(threads_chunk)) {
int i1 = Math.min(fj, i0+threads_chunk);
for (int i = i0; i < i1; i++) {
double aji = aj[i];
ajj_threaded[nthread] += a[i][i]*aji*aji; // squared
}
} }
} }
}; };
} }
ImageDtt.startAndJoin(threads); ImageDtt.startAndJoin(threads);
for (int nthread = 0; nthread < ajj_threaded.length; nthread++) {
ajj -= ajj_threaded[nthread];
}
} else {
for (int i = 0; i < j; i++) { // use aj, a, i
double aji = aj[i];
ajj -= a[i][i]*aji*aji; // squared
}
} }
*/ aj[j] = ajj;
for (int i = 0; i < j; i++) { // TODO: above run single-threaded for small, then multithreaded with post-accumulation of ajj
a[j][j] -= a[i][i]*a[j][i]*a[j][i]; // squared if ((n - j+1) > threads_chunk) { // use multithreaded
ai.set(j+1);
for (int ithread = 0; ithread < threads.length; ithread++) {
threads[ithread] = new Thread() {
public void run() {
for (int k0 = ai.getAndAdd(threads_chunk); k0 < n; k0 = ai.getAndAdd(threads_chunk)) {
int k1 = Math.min(n, k0+threads_chunk);
for (int k = k0; k < k1; k++) {
double [] ak = a[k];
for (int i = 0; i < fj; i++) { // ak, aj, i,
ak[fj] -= a[i][i] * ak[i] * aj[i];
}
ak[fj] /= aj[fj];
}
}
}
};
}
ImageDtt.startAndJoin(threads);
} else { // use single-threaded
for (int k = j+1; k < n; k++) {
final double [] ak = a[k];
for (int i = 0; i < j; i++) { // ak, aj, i,
ak[j] -= a[i][i]* ak[i] * aj[i];
}
ak[j] /= ajj;
}
} }
/* for (int i = 0; i < j; i++) {
a[j][j] -= a[i][i]*a[j][i]*a[j][i]; // squared
} */
for (int k = j+1; k < n; k++) {
final int fk = k;
/*
if (j > 0) {
ai.set(0);
for (int ithread = 0; ithread < threads.length; ithread++) {
threads[ithread] = new Thread() {
public void run() {
for (int i = ai.getAndIncrement(); i < fj; i = ai.getAndIncrement()) {
a[fk][fj] -= a[i][i]* a[fk][i] * a[fj][i];
}
}
};
}
ImageDtt.startAndJoin(threads);
}
*/
for (int i = 0; i < j; i++) {
a[k][j] -= a[i][i]* a[k][i] * a[j][i];
}
a[k][j] /= a[j][j];
}
/* for (int k = j+1; k < n; k++) {
for (int i = 0; i < j; i++) {
a[k][j] -= a[i][i]* a[k][i] * a[j][i];
}
a[k][j] /= a[j][j];
} */
} }
return; // L = a;
return a;
} }
public CholeskyLDLTMulti(Matrix matA, int ii) { // single-threaded public double[][] CholeskyLDLTMulti_fast(Matrix matA) {
n = matA.getRowDimension();
// not needed
// L = matA.getArray(); // will be modified, copy externally if needed
final double [][] a = matA.getArray();
for (int j = 0; j < n; j++) {
for (int i = 0; i < j; i++) {
a[j][j] -= a[i][i]*a[j][i]*a[j][i]; // squared
}
for (int k = j+1; k < n; k++) {
for (int i = 0; i < j; i++) {
a[k][j] -= a[i][i]* a[k][i] * a[j][i];
}
a[k][j] /= a[j][j];
}
}
L = a;
return;
}
public CholeskyLDLTMulti(Matrix matA) {
n = matA.getRowDimension();
final double [][] a = matA.getArray(); final double [][] a = matA.getArray();
final Thread[] threads = ImageDtt.newThreadArray(); final Thread[] threads = ImageDtt.newThreadArray();
final AtomicInteger ai = new AtomicInteger(0); final AtomicInteger ai = new AtomicInteger(0);
final AtomicInteger ati = new AtomicInteger(0); final AtomicInteger ati = new AtomicInteger(0);
final double [] ajj_threaded = new double[threads.length]; final double [] ajj_threaded = new double[threads.length];
final int threads_chunk = (int) (n / (thread_scale * threads.length)); final int threads_chunk = (int) (n / (thread_scale * threads.length));
final Integer [] slots_order = new Integer [threads.length]; // per-thread slot index
final int [] slot_j = new int [threads.length]; // slot synchronized to this j (including)
final double [][][] slots_a = new double [threads.length][n][n]; // copies of L in progress
Arrays.fill(slot_j, -1);
// Arrays.setAll(slots_order, i->i);
// for (int i = 0; i < slots_order.length; i++) {
// slots_order[i] = i;
// }
for (int j = 0; j < n; j++) { for (int j = 0; j < n; j++) {
final int fj = j; final int fj = j;
Arrays.setAll(slots_order, i->i); //https://stackoverflow.com/questions/68331291/how-do-i-reorder-another-array-based-on-a-sorted-array-java
Arrays.sort(slots_order, (la, lb) -> slot_j[lb] - slot_j[la]); // decreasing order
final double [] aj = a[j]; final double [] aj = a[j];
double ajj = aj[j]; double ajj = aj[j];
if (j > threads_chunk) { // use multithreaded if (j > threads_chunk) { // use multithreaded TEMPORARY MAKE IT ALWAYS SINGLE
Arrays.fill(ajj_threaded, 0.0); Arrays.fill(ajj_threaded, 0.0);
ai.set(0); ai.set(0);
ati.set(0); ati.set(0);
for (int ithread = 0; ithread < threads.length; ithread++) { // first sum for pairs for (int ithread = 0; ithread < threads.length; ithread++) { // first sum for pairs
threads[ithread] = new Thread() { threads[ithread] = new Thread() {
public void run() { public void run() {
int nthread = ati.getAndIncrement(); int ithread = ati.getAndIncrement();
double [][] local_a;
for (int i0 = ai.getAndAdd(threads_chunk); i0 < fj; i0 = ai.getAndAdd(threads_chunk)) { for (int i0 = ai.getAndAdd(threads_chunk); i0 < fj; i0 = ai.getAndAdd(threads_chunk)) {
int slot = slots_order[ithread];
local_a = slots_a[slot];
int local_j = slot_j[slot];
if (local_j < fj-1) { // catch up if needed
for (int col = local_j+1; col < fj; col++){
for (int row = local_j; row < n; row++) {
local_a[row][col] = a[row][col]; // copying from global to local memory
}
}
slot_j[slot] = fj-1;
}
int i1 = Math.min(fj, i0+threads_chunk); int i1 = Math.min(fj, i0+threads_chunk);
double [] local_aj = local_a[fj];
for (int i = i0; i < i1; i++) { for (int i = i0; i < i1; i++) {
double aji = aj[i]; double aji = local_aj[i];
ajj_threaded[nthread] += a[i][i]*aji*aji; // squared ajj_threaded[slot] += local_a[i][i]*aji*aji; // squared
} }
} }
} }
...@@ -202,20 +174,45 @@ public class CholeskyLDLTMulti { ...@@ -202,20 +174,45 @@ public class CholeskyLDLTMulti {
} }
} }
aj[j] = ajj; aj[j] = ajj;
final double fajj = ajj; // use in all threads, a[j][j] is also valid
// TODO: above run single-threaded for small, then multithreaded with post-accumulation of ajj // TODO: above run single-threaded for small, then multithreaded with post-accumulation of ajj
if ((n - j+1) > threads_chunk) { // use multithreaded if ((n - j+1) > threads_chunk) { // use multithreaded
// re-evaluate slots, they could change after diagonal calculations
Arrays.setAll(slots_order, i->i); //https://stackoverflow.com/questions/68331291/how-do-i-reorder-another-array-based-on-a-sorted-array-java
Arrays.sort(slots_order, (la, lb) -> slot_j[lb] - slot_j[la]); // decreasing order
ai.set(j+1); ai.set(j+1);
ati.set(0);
for (int ithread = 0; ithread < threads.length; ithread++) { for (int ithread = 0; ithread < threads.length; ithread++) {
threads[ithread] = new Thread() { threads[ithread] = new Thread() {
public void run() { public void run() {
int ithread = ati.getAndIncrement();
double [][] local_a;
for (int k0 = ai.getAndAdd(threads_chunk); k0 < n; k0 = ai.getAndAdd(threads_chunk)) { for (int k0 = ai.getAndAdd(threads_chunk); k0 < n; k0 = ai.getAndAdd(threads_chunk)) {
int slot = slots_order[ithread];
local_a = slots_a[slot];
int local_j = slot_j[slot];
if (local_j < fj-1) { // catch up if needed
for (int col = local_j+1; col < fj; col++){
for (int row =col; row < n; row++) {
local_a[row][col] = a[row][col]; // copying from global to local memory
}
}
slot_j[slot] = fj-1;
}
double [] local_aj = local_a[fj];
int k1 = Math.min(n, k0+threads_chunk); int k1 = Math.min(n, k0+threads_chunk);
for (int k = k0; k < k1; k++) { for (int k = k0; k < k1; k++) {
double [] ak = a[k]; double [] local_ak = local_a[k];
double akj = local_ak[fj];
for (int i = 0; i < fj; i++) { // ak, aj, i, for (int i = 0; i < fj; i++) { // ak, aj, i,
ak[fj] -= a[i][i] * ak[i] * aj[i]; // ak[fj] -= a[i][i] * ak[i] * aj[i];
// local_ak[fj] -= local_a[i][i] * local_ak[i] * local_aj[i];
akj -= local_a[i][i] * local_ak[i] * local_aj[i];
} }
ak[fj] /= aj[fj]; // ak[fj] /= aj[fj];
a[k][fj] = local_ak[fj] / fajj; // write to global memory
a[k][fj] = akj / fajj; // write to global memory
} }
} }
} }
...@@ -232,44 +229,34 @@ public class CholeskyLDLTMulti { ...@@ -232,44 +229,34 @@ public class CholeskyLDLTMulti {
} }
} }
} }
L = a; // L = a;
return; return a;
} }
public double [][] CholeskyLDLTMulti_single(Matrix matA) { // single-threaded
public CholeskyLDLTMulti(Matrix matA, long ll) { // single-threaded // n = matA.getRowDimension();
n = matA.getRowDimension();
// not needed // not needed
// L = matA.getArray(); // will be modified, copy externally if needed
final double [][] a = matA.getArray(); final double [][] a = matA.getArray();
for (int j = 0; j < n; j++) { for (int j = 0; j < n; j++) {
final double [] aj = a[j]; for (int i = 0; i < j; i++) {
double ajj = aj[j]; a[j][j] -= a[i][i]*a[j][i]*a[j][i]; // squared
for (int i = 0; i < j; i++) { // use aj, a, i
double aji = aj[i];
// a[j][j] -= a[i][i]*a[j][i]*a[j][i]; // squared
ajj -= a[i][i]*aji*aji; // squared
aj[j] -= a[i][i]*aji*aji; // squared
} }
aj[j] = ajj;
for (int k = j+1; k < n; k++) { for (int k = j+1; k < n; k++) {
final double [] ak = a[k]; for (int i = 0; i < j; i++) {
for (int i = 0; i < j; i++) { // ak, aj, i, a[k][j] -= a[i][i]* a[k][i] * a[j][i];
// a[k][j] -= a[i][i]* a[k][i] * a[j][i]; // use digonal
ak[j] -= a[i][i]* ak[i] * aj[i];
} }
// a[k][j] /= a[j][j]; a[k][j] /= a[j][j];
// ak[j] /= aj[j];
ak[j] /= ajj;
} }
} }
L = a; // L = a;
return; return a;
} }
public Matrix solve0 (Matrix B) { // multithreaded
public Matrix solve0 (Matrix B) { // multithreased
if (B.getRowDimension() != n) { if (B.getRowDimension() != n) {
throw new IllegalArgumentException("Matrix row dimensions must agree."); throw new IllegalArgumentException("Matrix row dimensions must agree.");
} }
...@@ -407,18 +394,56 @@ public class CholeskyLDLTMulti { ...@@ -407,18 +394,56 @@ public class CholeskyLDLTMulti {
return new Matrix(x,n); return new Matrix(x,n);
} }
public static void testCholesky(ImagePlus imp_src) {
float [] fpixels = (float[]) imp_src.getProcessor().getPixels();
int width = imp_src.getWidth();
int height = imp_src.getHeight();
int n = height;
double [][] a = new double [n][n];
double [][] b = new double [n][1];
for (int i = 0; i < n; i++) {
for (int j= 0; j < n; j++) {
a[i][j] = fpixels[i*width+j];
}
b[i][0] = fpixels[i*width+n];
}
testCholesky(
new Matrix(a), // Matrix wjtjlambda,
new Matrix(b)); // Matrix jty)
}
public static Matrix[] testCholesky( public static Matrix[] testCholesky(
Matrix wjtjlambda, Matrix wjtjlambda,
Matrix jty) { Matrix jty) {
Matrix wjtjlambda_copy0 = new Matrix(wjtjlambda.getArrayCopy()); Matrix wjtjlambda_copy0 = new Matrix(wjtjlambda.getArrayCopy());
CholeskyDecomposition choleskyDecomposition = new CholeskyDecomposition(wjtjlambda_copy0); Matrix wjtjlambda_copy1 = new Matrix(wjtjlambda.getArrayCopy());
Matrix wjtjlambda_copy2 = new Matrix(wjtjlambda.getArrayCopy());
Matrix wjtjlambda_copy = new Matrix(wjtjlambda.getArrayCopy()); Matrix wjtjlambda_copy = new Matrix(wjtjlambda.getArrayCopy());
CholeskyLDLTMulti choleskyLDLTMulti = new CholeskyLDLTMulti(wjtjlambda_copy); double [] starts = new double[7];
double start_time = (((double) System.nanoTime()) * 1E-9);
CholeskyDecomposition choleskyDecomposition = new CholeskyDecomposition(wjtjlambda_copy);
starts[0] = (((double) System.nanoTime()) * 1E-9) - start_time;
CholeskyLDLTMulti choleskyLDLTMulti_single = new CholeskyLDLTMulti(wjtjlambda_copy0,0);
starts[1] = (((double) System.nanoTime()) * 1E-9) - start_time;
CholeskyLDLTMulti choleskyLDLTMulti_multi = new CholeskyLDLTMulti(wjtjlambda_copy1,1);
starts[2] = (((double) System.nanoTime()) * 1E-9) - start_time;
CholeskyLDLTMulti choleskyLDLTMulti_fast = new CholeskyLDLTMulti(wjtjlambda_copy2,2);
starts[3] = (((double) System.nanoTime()) * 1E-9) - start_time;
CholeskyLDLTMulti choleskyLDLTMulti = choleskyLDLTMulti_fast;
Matrix mdelta_cholesky = choleskyDecomposition.solve(jty); Matrix mdelta_cholesky = choleskyDecomposition.solve(jty);
starts[4] = (((double) System.nanoTime()) * 1E-9) - start_time;
Matrix mdelta_cholesky_multi0 = choleskyLDLTMulti.solve0(jty); Matrix mdelta_cholesky_multi0 = choleskyLDLTMulti.solve0(jty);
starts[5] = (((double) System.nanoTime()) * 1E-9) - start_time;
Matrix mdelta_cholesky_multi = choleskyLDLTMulti.solve(jty); Matrix mdelta_cholesky_multi = choleskyLDLTMulti.solve(jty);
starts[6] = (((double) System.nanoTime()) * 1E-9) - start_time;
System.out.println("testCholesky(): choleskyDecomposition: "+(starts[0])+" sec");
System.out.println("testCholesky(): choleskyLDLTMulti_single: "+(starts[1]-starts[0])+" sec");
System.out.println("testCholesky(): choleskyLDLTMulti_multi: "+(starts[2]-starts[1])+" sec");
System.out.println("testCholesky(): choleskyLDLTMulti_fast: "+(starts[3]-starts[2])+" sec");
System.out.println("testCholesky(): mdelta_cholesky: "+(starts[4]-starts[3])+" sec");
System.out.println("testCholesky(): mdelta_cholesky_multi0: "+(starts[5]-starts[4])+" sec");
System.out.println("testCholesky(): mdelta_cholesky_multi: "+(starts[6]-starts[5])+" sec");
return new Matrix[] {mdelta_cholesky, mdelta_cholesky_multi0, mdelta_cholesky_multi}; return new Matrix[] {mdelta_cholesky, mdelta_cholesky_multi0, mdelta_cholesky_multi};
} }
......
src/main/java/com/elphel/imagej/common/CholeskyLLTMulti.java 0 → 100644
View file @ 3cedfc7a
package com.elphel.imagej.common;
import java.util.Arrays;
import java.util.concurrent.atomic.AtomicInteger;
import com.elphel.imagej.tileprocessor.ImageDtt;
import Jama.CholeskyDecomposition;
import Jama.Matrix;
import ij.ImagePlus;
public class CholeskyLLTMulti {
public final int n;
public final double [][] L;
public int solve_step = 1; //0;
public double thread_scale = 10; // 2.5; // split jobs in thread_scale * num_treads chunks
public Matrix getL () {
return new Matrix(L,n,n);
}
public CholeskyLLTMulti(Matrix matA, int mode) {
n = matA.getRowDimension();
L = new double[n][n];
switch (mode) {
case 0: CholeskyLLTMulti_single(matA); break;
case 1: CholeskyLLTMulti_multi(matA); break;
case 2: CholeskyLLTMulti_fast(matA); break;
default:CholeskyLLTMulti_single(matA);
}
}
// Cholesky-Banachiewicz Algorithm ?
public double [][] CholeskyLLTMulti_single (Matrix mA) {
double[][] a = mA.getArray();
// Main loop.
for (int j = 0; j < n; j++) {
double[] Lrowj = L[j];
double d = 0.0;
for (int k = 0; k < j; k++) {
double[] Lrowk = L[k];
double s = 0.0;
for (int i = 0; i < k; i++) {
s += Lrowk[i]*Lrowj[i];
}
Lrowj[k] = s = (a[j][k] - s)/L[k][k];
d = d + s*s;
}
d = a[j][j] - d;
L[j][j] = Math.sqrt(Math.max(d,0.0));
// for (int k = j+1; k < n; k++) {
// L[j][k] = 0.0;
// }
}
return L;
}
public double [][] CholeskyLLTMulti_multi (Matrix mA) {
double[][] a = mA.getArray();
final Thread[] threads = ImageDtt.newThreadArray();
final AtomicInteger ai = new AtomicInteger(0);
final AtomicInteger ati = new AtomicInteger(0);
final int threads_chunk = (int) (n / (thread_scale * threads.length));
final double [] threads_d = new double [threads.length];
// Main loop.
for (int j = 0; j < n; j++) {
final int fj = j;
final double[] Lrowj = L[j];
double d = 0.0;
if (j > threads_chunk) {
ai.set(0);
ati.set(0);
Arrays.fill(threads_d, 0.0);
for (int ithread = 0; ithread < threads.length; ithread++) { // first sum for pairs
threads[ithread] = new Thread() {
public void run() {
int nthread = ati.getAndIncrement();
for (int k0 = ai.getAndAdd(threads_chunk); k0 < fj; k0 = ai.getAndAdd(threads_chunk)) {
int k1 = Math.min(fj, k0+threads_chunk);
for (int k = k0; k < k1; k++) {
double[] Lrowk = L[k];
double s = 0.0;
for (int i = 0; i < k; i++) {
s += Lrowk[i]*Lrowj[i];
}
s = (a[fj][k] - s)/L[k][k];
Lrowj[k] = s;
threads_d[nthread] += s * s;
}
}
}
};
}
ImageDtt.startAndJoin(threads);
for (double dt:threads_d) d+=dt;
} else {
for (int k = 0; k < j; k++) {
double[] Lrowk = L[k];
double s = 0.0;
for (int i = 0; i < k; i++) {
s += Lrowk[i]*Lrowj[i];
}
Lrowj[k] = s = (a[j][k] - s)/L[k][k];
d = d + s*s;
}
}
d = a[j][j] - d;
L[j][j] = Math.sqrt(Math.max(d,0.0));
// for (int k = j+1; k < n; k++) {
// L[j][k] = 0.0;
// }
}
return a;
}
public double [][] CholeskyLLTMulti_fast (Matrix mA) {
double[][] a = mA.getArray();
final Thread[] threads = ImageDtt.newThreadArray();
final AtomicInteger ai = new AtomicInteger(0);
final AtomicInteger ati = new AtomicInteger(0);
final int threads_chunk = (int) (n / (thread_scale * threads.length));
final double [] threads_d = new double [threads.length];
// Main loop.
for (int j = 0; j < n; j++) {
final int fj = j;
final double[] Lrowj = L[j];
double d = 0.0;
// WRONG! parallel only for the inner-most (i), because k should progress consequently
if (j > threads_chunk) {
ai.set(0);
ati.set(0);
Arrays.fill(threads_d, 0.0);
for (int ithread = 0; ithread < threads.length; ithread++) { // first sum for pairs
threads[ithread] = new Thread() {
public void run() {
int nthread = ati.getAndIncrement();
// double[] localLrowj = Lrowj.clone();
for (int k0 = ai.getAndAdd(threads_chunk); k0 < fj; k0 = ai.getAndAdd(threads_chunk)) {
int k1 = Math.min(fj, k0+threads_chunk);
for (int k = k0; k < k1; k++) {
double[] Lrowk = L[k];
double s = 0.0;
for (int i = 0; i < k; i++) {
s += Lrowk[i]*Lrowj[i];
}
s = (a[fj][k] - s)/L[k][k];
Lrowj[k] = s; // write ***
threads_d[nthread] += s * s;
}
}
}
};
}
ImageDtt.startAndJoin(threads);
for (double dt:threads_d) d+=dt;
} else {
for (int k = 0; k < j; k++) {
double[] Lrowk = L[k];
double s = 0.0;
for (int i = 0; i < k; i++) {
s += Lrowk[i]*Lrowj[i];
}
Lrowj[k] = s = (a[j][k] - s)/L[k][k];
d = d + s*s;
}
}
d = a[j][j] - d;
L[j][j] = Math.sqrt(Math.max(d,0.0)); // write ***
// for (int k = j+1; k < n; k++) {
// L[j][k] = 0.0;
// }
}
return a;
}
public Matrix solve (Matrix B, int mode) {
switch (mode) {
case 0: return solve_single(B);
// case 1: return solve_multi (B);
// case 2: return solve_fast (B);
default:return solve_single(B);
}
}
/** Solve A*X = B
@param B A Matrix with as many rows as A and any number of columns.
@return X so that L*L'*X = B
@exception IllegalArgumentException Matrix row dimensions must agree.
@exception RuntimeException Matrix is not symmetric positive definite.
*/
public Matrix solve_single (Matrix B) {
if (B.getRowDimension() != n) {
throw new IllegalArgumentException("Matrix row dimensions must agree.");
}
// Copy right hand side.
double[] x = B.getColumnPackedCopy (); // (for single-column)
// Solve L*Y = B;
for (int k = 0; k < n; k++) {
for (int i = 0; i < k ; i++) {
x[k] -= x[i]*L[k][i];
}
x[k] /= L[k][k];
}
// Solve L'*X = Y;
for (int k = n-1; k >= 0; k--) {
for (int i = k+1; i < n ; i++) {
x[k] -= x[i]*L[i][k];
}
x[k] /= L[k][k];
}
return new Matrix(x,n);
}
public static Matrix solve_single (Matrix B, Matrix L_mat) {
double [][] L = L_mat.getArray();
return solve_single (B, L);
}
public static Matrix solve_single (Matrix B, double [][] L) {
int n = L.length;
if (B.getRowDimension() != n) {
throw new IllegalArgumentException("Matrix row dimensions must agree.");
}
// Copy right hand side.
double[] x = B.getColumnPackedCopy (); // (for single-column)
// Solve L*Y = B;
for (int k = 0; k < n; k++) {
for (int i = 0; i < k ; i++) {
x[k] -= x[i]*L[k][i];
}
x[k] /= L[k][k];
}
// Solve L'*X = Y;
for (int k = n-1; k >= 0; k--) {
for (int i = k+1; i < n ; i++) {
x[k] -= x[i]*L[i][k];
}
x[k] /= L[k][k];
}
return new Matrix(x,n);
}
public static void testCholesky(ImagePlus imp_src) {
float [] fpixels = (float[]) imp_src.getProcessor().getPixels();
int width = imp_src.getWidth();
int height = imp_src.getHeight();
int n = height;
double [][] a = new double [n][n];
double [][] b = new double [n][1];
for (int i = 0; i < n; i++) {
for (int j= 0; j < n; j++) {
a[i][j] = fpixels[i*width+j];
}
b[i][0] = fpixels[i*width+n];
}
testCholesky(
new Matrix(a), // Matrix wjtjlambda,
new Matrix(b), // Matrix jty)
imp_src.getTitle()); // String title);
}
public static Matrix[] testCholesky(
Matrix wjtjlambda,
Matrix jty,
String title) {
int block_size = 70; // 64; // 60; // 70; // 80; // 120; // 150; // 200; // 100; // 4; // 10; // 100; // 28;
String dbg_title=title+"ch_diff_choleskyBlock-choleskyDecomposition-"+block_size;
Matrix wjtjlambda_copy0 = new Matrix(wjtjlambda.getArrayCopy());
Matrix wjtjlambda_copy1 = new Matrix(wjtjlambda.getArrayCopy());
Matrix wjtjlambda_copy2 = new Matrix(wjtjlambda.getArrayCopy());
Matrix wjtjlambda_copy = new Matrix(wjtjlambda.getArrayCopy());
double [] starts = new double[7];
double start_time = (((double) System.nanoTime()) * 1E-9);
CholeskyDecomposition choleskyDecomposition = new CholeskyDecomposition(wjtjlambda_copy);
starts[0] = (((double) System.nanoTime()) * 1E-9) - start_time;
CholeskyLLTMulti choleskyLLTMulti_single = new CholeskyLLTMulti(wjtjlambda_copy0,0);
starts[1] = (((double) System.nanoTime()) * 1E-9) - start_time;
CholeskyBlock choleskyBlock = new CholeskyBlock(wjtjlambda_copy1.getArray(),block_size);
starts[2] = (((double) System.nanoTime()) * 1E-9) - start_time;
choleskyBlock.choleskyBlockMulti();
starts[3] = (((double) System.nanoTime()) * 1E-9) - start_time;
double [][] LTriangle = choleskyBlock.get_LTriangle();
Matrix mdelta_cholesky = choleskyDecomposition.solve(jty);
starts[4] = (((double) System.nanoTime()) * 1E-9) - start_time;
Matrix mdelta_cholesky_single = CholeskyLLTMulti.solve_single(jty, choleskyLLTMulti_single.getL());
starts[5] = (((double) System.nanoTime()) * 1E-9) - start_time;
Matrix mdelta_cholesky_multi = CholeskyLLTMulti.solve_single(jty, LTriangle);
starts[6] = (((double) System.nanoTime()) * 1E-9) - start_time;
System.out.println("testCholesky(): block_size= "+block_size);
System.out.println("testCholesky(): choleskyDecomposition: "+(starts[0])+" sec");
System.out.println("testCholesky(): choleskyLLTMulti_single:"+(starts[1]-starts[0])+" sec");
System.out.println("testCholesky(): CholeskyBlock(): "+ (starts[2]-starts[1])+" sec");
System.out.println("testCholesky(): choleskyBlockMulti(): "+(starts[3]-starts[2])+" sec");
System.out.println("testCholesky(): get_LTriangle(): "+(starts[4]-starts[3])+" sec");
System.out.println("testCholesky(): solve_single(,single): "+(starts[5]-starts[4])+" sec");
System.out.println("testCholesky(): solve_single(,block): "+(starts[6]-starts[5])+" sec");
System.out.println("testCholesky(): title= "+block_size);
System.out.println("testCholesky(): dbg_title= "+block_size);
Matrix ch_diff = choleskyLLTMulti_single.getL().minus(choleskyDecomposition.getL());
Matrix ch_diff_fast = choleskyBlock.getL().minus(choleskyDecomposition.getL());
double [][] dbg_img = {
choleskyLLTMulti_single.getL().getRowPackedCopy(),
choleskyBlock.getL().getRowPackedCopy(),
choleskyDecomposition.getL().getRowPackedCopy(),
ch_diff.getRowPackedCopy(),
ch_diff_fast.getRowPackedCopy()};
String[] dbg_titles = {"choleskyLLTMulti_single","choleskyBlock","choleskyDecomposition",
"choleskyLLTMulti_single-choleskyDecomposition","choleskyBlock-choleskyDecomposition"};
ShowDoubleFloatArrays.showArrays(
dbg_img, // double[] pixels,
ch_diff.getRowDimension(), // int width,
ch_diff.getRowDimension(), // int height,
true,
dbg_title, // String title)
dbg_titles);
return new Matrix[] {mdelta_cholesky, mdelta_cholesky_single, mdelta_cholesky_multi};
}
public static Matrix[] testCholesky0(
Matrix wjtjlambda,
Matrix jty) {
int block_size = 128;
Matrix wjtjlambda_copy0 = new Matrix(wjtjlambda.getArrayCopy());
Matrix wjtjlambda_copy1 = new Matrix(wjtjlambda.getArrayCopy());
Matrix wjtjlambda_copy2 = new Matrix(wjtjlambda.getArrayCopy());
Matrix wjtjlambda_copy = new Matrix(wjtjlambda.getArrayCopy());
double [] starts = new double[8];
double start_time = (((double) System.nanoTime()) * 1E-9);
CholeskyDecomposition choleskyDecomposition = new CholeskyDecomposition(wjtjlambda_copy);
starts[0] = (((double) System.nanoTime()) * 1E-9) - start_time;
CholeskyLLTMulti choleskyLLTMulti_single = new CholeskyLLTMulti(wjtjlambda_copy0,0);
starts[1] = (((double) System.nanoTime()) * 1E-9) - start_time;
CholeskyLLTMulti choleskyLLTMulti_multi = new CholeskyLLTMulti(wjtjlambda_copy1,1);
starts[2] = (((double) System.nanoTime()) * 1E-9) - start_time;
CholeskyLLTMulti choleskyLLTMulti_fast = new CholeskyLLTMulti(wjtjlambda_copy2,2);
starts[3] = (((double) System.nanoTime()) * 1E-9) - start_time;
CholeskyLLTMulti choleskyLLTMulti = choleskyLLTMulti_multi; // fast;
Matrix mdelta_cholesky = choleskyDecomposition.solve(jty);
starts[4] = (((double) System.nanoTime()) * 1E-9) - start_time;
Matrix mdelta_cholesky_single = choleskyLLTMulti.solve(jty, 0);
starts[5] = (((double) System.nanoTime()) * 1E-9) - start_time;
Matrix mdelta_cholesky_multi = choleskyLLTMulti.solve(jty, 1);
starts[6] = (((double) System.nanoTime()) * 1E-9) - start_time;
Matrix mdelta_cholesky_fast = choleskyLLTMulti.solve(jty, 2);
starts[7] = (((double) System.nanoTime()) * 1E-9) - start_time;
System.out.println("testCholesky(): choleskyDecomposition: "+(starts[0])+" sec");
System.out.println("testCholesky(): choleskyLLTMulti_single:"+(starts[1]-starts[0])+" sec");
System.out.println("testCholesky(): choleskyLLTMulti_multi: "+(starts[2]-starts[1])+" sec");
System.out.println("testCholesky(): choleskyLLTMulti_fast: "+(starts[3]-starts[2])+" sec");
System.out.println("testCholesky(): mdelta_cholesky: "+(starts[4]-starts[3])+" sec");
System.out.println("testCholesky(): mdelta_cholesky_single: "+(starts[5]-starts[4])+" sec");
System.out.println("testCholesky(): mdelta_cholesky_multi: "+(starts[6]-starts[5])+" sec");
System.out.println("testCholesky(): mdelta_cholesky_fast: "+(starts[7]-starts[6])+" sec");
Matrix ch_diff = choleskyLLTMulti.getL().minus(choleskyDecomposition.getL());
Matrix ch_diff_fast = choleskyLLTMulti_fast.getL().minus(choleskyDecomposition.getL());
double [][] dbg_img = {
choleskyLLTMulti.getL().getRowPackedCopy(),
choleskyLLTMulti_fast.getL().getRowPackedCopy(),
choleskyDecomposition.getL().getRowPackedCopy(),
ch_diff.getRowPackedCopy(),
ch_diff_fast.getRowPackedCopy()};
String[] dbg_titles = {"choleskyLLTMulti","choleskyLLTMulti_fast","choleskyDecomposition",
"choleskyLLTMulti-choleskyDecomposition","choleskyLLTMulti_fast-choleskyDecomposition"};
ShowDoubleFloatArrays.showArrays(
dbg_img, // double[] pixels,
ch_diff.getRowDimension(), // int width,
ch_diff.getRowDimension(), // int height,
true,
"ch_diff_choleskyLLTMulti-choleskyDecomposition", // String title)
dbg_titles);
return new Matrix[] {mdelta_cholesky, mdelta_cholesky_single, mdelta_cholesky_multi, mdelta_cholesky_fast};
}
}
src/main/java/com/elphel/imagej/correction/Eyesis_Correction.java
View file @ 3cedfc7a
...@@ -96,6 +96,8 @@ import com.elphel.imagej.cameras.CLTParameters; ...@@ -96,6 +96,8 @@ import com.elphel.imagej.cameras.CLTParameters;
import com.elphel.imagej.cameras.ColorProcParameters; import com.elphel.imagej.cameras.ColorProcParameters;
import com.elphel.imagej.cameras.EyesisCorrectionParameters; import com.elphel.imagej.cameras.EyesisCorrectionParameters;
import com.elphel.imagej.cameras.ThermalColor; import com.elphel.imagej.cameras.ThermalColor;
import com.elphel.imagej.common.CholeskyLDLTMulti;
import com.elphel.imagej.common.CholeskyLLTMulti;
import com.elphel.imagej.common.DoubleFHT; import com.elphel.imagej.common.DoubleFHT;
import com.elphel.imagej.common.DoubleGaussianBlur; import com.elphel.imagej.common.DoubleGaussianBlur;
import com.elphel.imagej.common.GenericJTabbedDialog; import com.elphel.imagej.common.GenericJTabbedDialog;
...@@ -865,6 +867,8 @@ public class Eyesis_Correction implements PlugIn, ActionListener { ...@@ -865,6 +867,8 @@ public class Eyesis_Correction implements PlugIn, ActionListener {
addButton("Vegetation LMA", panelOrange, color_process); addButton("Vegetation LMA", panelOrange, color_process);
addButton("Combine LMA Segments", panelOrange, color_process); addButton("Combine LMA Segments", panelOrange, color_process);
addButton("Generate LWIR target", panelOrange, color_process); addButton("Generate LWIR target", panelOrange, color_process);
addButton("Test LDLT Cholesky", panelOrange, color_process);
addButton("Test LLT Cholesky", panelOrange, color_process);
plugInFrame.add(panelOrange); plugInFrame.add(panelOrange);
} }
plugInFrame.pack(); plugInFrame.pack();
...@@ -5807,8 +5811,20 @@ public class Eyesis_Correction implements PlugIn, ActionListener { ...@@ -5807,8 +5811,20 @@ public class Eyesis_Correction implements PlugIn, ActionListener {
SYNC_COMMAND, // // SyncCommand SYNC_COMMAND, SYNC_COMMAND, // // SyncCommand SYNC_COMMAND,
CLT_PARAMETERS, //CLTParameters clt_parameters, CLT_PARAMETERS, //CLTParameters clt_parameters,
true); //boolean combine_segments); true); //boolean combine_segments);
} else if (label.equals("Generate LWIR target")) { } else if (label.equals("Test LDLT Cholesky")) {
VegetationSegment.generateSineTarget(); ImagePlus imp_sel = WindowManager.getCurrentImage();
if (imp_sel == null) {
IJ.showMessage("Error", "No images selected");
return;
}
CholeskyLDLTMulti.testCholesky(imp_sel); // ImagePlus imp.);
} else if (label.equals("Test LLT Cholesky")) {
ImagePlus imp_sel = WindowManager.getCurrentImage();
if (imp_sel == null) {
IJ.showMessage("Error", "No images selected");
return;
}
CholeskyLLTMulti.testCholesky(imp_sel); // ImagePlus imp.);
} }
// //
} }
......
src/main/java/com/elphel/imagej/vegetation/VegetationLMA.java
View file @ 3cedfc7a
...@@ -1907,9 +1907,32 @@ public class VegetationLMA { ...@@ -1907,9 +1907,32 @@ public class VegetationLMA {
Matrix mdelta_cholesky = choleskyDecomposition.solve(jty); Matrix mdelta_cholesky = choleskyDecomposition.solve(jty);
Matrix mdelta_cholesky_multi = choleskyLDLTMulti.solve(jty); Matrix mdelta_cholesky_multi = choleskyLDLTMulti.solve(jty);
*/ */
CholeskyLDLTMulti.testCholesky( boolean save_cholesky = true;
wjtjlambda, // Matrix wjtjlambda, boolean debug_cholesky = true;
jty); if (save_cholesky) {
int dbg_n = wjtjlambda.getRowDimension();
double [] dbg_a0 = wjtjlambda.getRowPackedCopy();
double [] dbg_a = new double[dbg_n*(dbg_n+1)];
for (int i = 0; i < dbg_n; i++) {
System.arraycopy(
dbg_a0,
i*dbg_n,
dbg_a,
i*(dbg_n+1),
dbg_n);
dbg_a[(i+1)*(dbg_n+1)-1] = jty.get(i,0);
}
ShowDoubleFloatArrays.showArrays(
dbg_a, // double[] pixels,
dbg_n+1, // int width,
dbg_n, // int height,
"spd_a_"+debug_iter); // String title)
}
if (debug_cholesky) {
CholeskyLDLTMulti.testCholesky(
wjtjlambda, // Matrix wjtjlambda,
jty);
}
if (debug_level>2) { if (debug_level>2) {
System.out.println("mdelta"); System.out.println("mdelta");
mdelta.print(18, 6); mdelta.print(18, 6);
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment