Skip to content

I
imagej-elphel
Commit 84eeaf35 authored by Andrey Filippov's avatar Andrey Filippov
Browse files
Options

Added GPU 2D phase correlation, related and debug functionality

parent dc4e1f60
Expand all Hide whitespace changes
Inline Side-by-side
Showing with 2252 additions and 794 deletions
src/main/java/com/elphel/imagej/cameras/EyesisCorrectionParameters.java
View file @ 84eeaf35
......@@ -231,6 +231,7 @@ public class EyesisCorrectionParameters {
cp.zcorrect= this.zcorrect;
cp.saveSettings= this.saveSettings;
cp.sourceDirectory= this.sourceDirectory;
cp.tile_processor_gpu = this.tile_processor_gpu;
cp.use_set_dirs = this.use_set_dirs;
// cp.sourcePrefix= this.sourcePrefix;
// cp.sourceSuffix= this.sourceSuffix;
......@@ -388,6 +389,8 @@ public class EyesisCorrectionParameters {
properties.setProperty(prefix+"saveSettings",this.saveSettings+"");
properties.setProperty(prefix+"sourceDirectory",this.sourceDirectory);
properties.setProperty(prefix+"tile_processor_gpu",this.tile_processor_gpu);
properties.setProperty(prefix+"use_set_dirs", this.use_set_dirs+"");
properties.setProperty(prefix+"sourcePrefix",this.sourcePrefix);
......@@ -543,6 +546,7 @@ public class EyesisCorrectionParameters {
if (properties.getProperty(prefix+"zcorrect")!=null) this.zcorrect=Boolean.parseBoolean(properties.getProperty(prefix+"zcorrect"));
if (properties.getProperty(prefix+"saveSettings")!=null) this.saveSettings=Boolean.parseBoolean(properties.getProperty(prefix+"saveSettings"));
if (properties.getProperty(prefix+"sourceDirectory")!= null) this.sourceDirectory=properties.getProperty(prefix+"sourceDirectory");
if (properties.getProperty(prefix+"tile_processor_gpu")!= null) this.tile_processor_gpu=properties.getProperty(prefix+"tile_processor_gpu");
if (properties.getProperty(prefix+"firstSubCamera")!= null) this.firstSubCamera=Integer.parseInt(properties.getProperty(prefix+"firstSubCamera"));
if (properties.getProperty(prefix+"firstSubCameraConfig")!= null) this.firstSubCameraConfig=Integer.parseInt(properties.getProperty(prefix+"firstSubCameraConfig"));
if (properties.getProperty(prefix+"numSubCameras")!= null) this.numSubCameras=Integer.parseInt(properties.getProperty(prefix+"numSubCameras"));
......
src/main/java/com/elphel/imagej/correction/EyesisDCT.java
View file @ 84eeaf35
......@@ -1220,7 +1220,7 @@ public class EyesisDCT {
}
if (this.correctionsParameters.deconvolve) { // process with DCT, otherwise use simple debayer
ImageDtt image_dtt = new ImageDtt(false, 1.0); // Bayer( not monochrome), scale correlation strengths
ImageDtt image_dtt = new ImageDtt(dctParameters.dct_size, false, 1.0); // Bayer( not monochrome), scale correlation strengths
double [][][][] dct_data = image_dtt.mdctStack(
stack,
channel,
......
src/main/java/com/elphel/imagej/correction/Eyesis_Correction.java
View file @ 84eeaf35
......@@ -3075,7 +3075,7 @@ private Panel panel1,
}
}
ImageDtt image_dtt = new ImageDtt(false, 1.0); // Bayer( not monochrome), scale correlation strengths
ImageDtt image_dtt = new ImageDtt(DCT_PARAMETERS.dct_size, false, 1.0); // Bayer( not monochrome), scale correlation strengths
double [][][][] dctdc_data = image_dtt.mdctScale(
DBG_IMP.getStack(),
DCT_PARAMETERS.kernel_chn,
......@@ -3173,7 +3173,7 @@ private Panel panel1,
}
}
ImageDtt image_dtt = new ImageDtt(false, 1.0); // Bayer( not monochrome), scale correlation strengths
ImageDtt image_dtt = new ImageDtt(DCT_PARAMETERS.dct_size,false, 1.0); // Bayer( not monochrome), scale correlation strengths
double [][][][] dctdc_data = image_dtt.mdctStack(
DBG_IMP.getStack(),
DCT_PARAMETERS.kernel_chn,
......@@ -5723,6 +5723,7 @@ private Panel panel1,
if (!prepareRigImages()) return false;
String configPath=getSaveCongigPath();
if (configPath.equals("ABORT")) return false;
// if ((CORRECTION_PARAMETERS.tile_processor_gpu != null) &&
if (DEBUG_LEVEL > -2){
System.out.println("++++++++++++++ Calculating combined correlations ++++++++++++++");
......@@ -5740,6 +5741,7 @@ private Panel panel1,
try {
TWO_QUAD_CLT.prepareFilesForGPUDebug(
CORRECTION_PARAMETERS.tile_processor_gpu,// String save_prefix, // absolute path to the cuda project root
QUAD_CLT, // QuadCLT quadCLT_main,
QUAD_CLT_AUX, // QuadCLT quadCLT_aux,
CLT_PARAMETERS, // EyesisCorrectionParameters.DCTParameters dct_parameters,
......@@ -7048,7 +7050,10 @@ private Panel panel1,
}
}
ImageDtt image_dtt = new ImageDtt(false, 1.0); // Bayer( not monochrome), scale correlation strengths
ImageDtt image_dtt = new ImageDtt(
CLT_PARAMETERS.transform_size,
false,
1.0); // Bayer( not monochrome), scale correlation strengths
double [][][][][] clt_data = image_dtt.cltStack(
DBG_IMP.getStack(),
0, // CLT_PARAMETERS.kernel_chn,
......@@ -7082,7 +7087,7 @@ private Panel panel1,
for (int chn = 0; chn < clt_data.length; chn++) {
clt_data[chn] = image_dtt.clt_shiftXY(
clt_data[chn], // final double [][][][] dct_data, // array [tilesY][tilesX][4][dct_size*dct_size]
CLT_PARAMETERS.transform_size, // final int dct_size,
/// CLT_PARAMETERS.transform_size, // final int dct_size,
CLT_PARAMETERS.shift_x, // final double shiftX,
CLT_PARAMETERS.shift_y, // final double shiftY,
(CLT_PARAMETERS.dbg_mode >> 2) & 3, // swap order hor/vert
......@@ -7095,7 +7100,7 @@ private Panel panel1,
for (int chn=0; chn<iclt_data.length;chn++){
iclt_data[chn] = image_dtt.iclt_2d(
clt_data[chn], // scanline representation of dcd data, organized as dct_size x dct_size tiles
CLT_PARAMETERS.transform_size, // final int
/// CLT_PARAMETERS.transform_size, // final int
CLT_PARAMETERS.clt_window, //window_type
CLT_PARAMETERS.iclt_mask, //which of 4 to transform back
CLT_PARAMETERS.dbg_mode, //which of 4 to transform back
......@@ -7178,7 +7183,10 @@ private Panel panel1,
}
String suffix = "-dx_"+(CLT_PARAMETERS.ishift_x+CLT_PARAMETERS.shift_x)+"_dy_"+(CLT_PARAMETERS.ishift_y+CLT_PARAMETERS.shift_y);
ImageDtt image_dtt = new ImageDtt(COLOR_PROC_PARAMETERS.isMonochrome(), CLT_PARAMETERS.getScaleStrength(false)); // Bayer, not monochrome
ImageDtt image_dtt = new ImageDtt(
CLT_PARAMETERS.transform_size,
COLOR_PROC_PARAMETERS.isMonochrome(),
CLT_PARAMETERS.getScaleStrength(false)); // Bayer, not monochrome
String [] titles = {
"redCC", "redSC", "redCS", "redSS",
"blueCC", "blueSC", "blueCS", "blueSS",
......@@ -7235,7 +7243,7 @@ private Panel panel1,
for (int chn = 0; chn < clt_data.length; chn++) {
clt_data2[chn] = image_dtt.clt_shiftXY(
clt_data2[chn], // final double [][][][] dct_data, // array [tilesY][tilesX][4][dct_size*dct_size]
CLT_PARAMETERS.transform_size, // final int dct_size,
/// CLT_PARAMETERS.transform_size, // final int dct_size,
CLT_PARAMETERS.shift_x, // final double shiftX,
CLT_PARAMETERS.shift_y, // final double shiftY,
(CLT_PARAMETERS.dbg_mode >> 2) & 3, // swap order hor/vert
......@@ -7266,7 +7274,7 @@ private Panel panel1,
clt_corr[chn] = image_dtt.clt_correlate(
clt_data[chn], // final double [][][][] data1, // array [tilesY][tilesX][4][dct_size*dct_size]
clt_data2[chn], // final double [][][][] data2, // array [tilesY][tilesX][4][dct_size*dct_size]
CLT_PARAMETERS.transform_size, // final int dct_size,
/// CLT_PARAMETERS.transform_size, // final int dct_size,
CLT_PARAMETERS.getFatZero(image_dtt.isMonochrome()), // final double fat_zero, // add to denominator to modify phase correlation (same units as data1, data2)
CLT_PARAMETERS.tileX, //final int debug_tileX
CLT_PARAMETERS.tileY, //final int debug_tileY
......@@ -7297,7 +7305,7 @@ private Panel panel1,
image_dtt.clt_lpf( // filter in-place
CLT_PARAMETERS.getCorrSigma(image_dtt.isMonochrome()), // final double sigma,
clt_corr[chn], // final double [][][][] clt_data,
CLT_PARAMETERS.transform_size,
/// CLT_PARAMETERS.transform_size,
THREADS_MAX, // maximal number of threads to launch
DEBUG_LEVEL); // globalDebugLevel)
}
......
src/main/java/com/elphel/imagej/tileprocessor/Correlation2d.java
View file @ 84eeaf35
......@@ -194,6 +194,147 @@ public class Correlation2d {
return this.transpose_all_diagonal;
}
/**
* Multiply CLT data of two channels, OK with null inputs (missing colors for monochrome images)
* @param clt_data1 first operand FD CLT data[4][transform_len]
* @param clt_data2 second operand FD CLT data[4][transform_len]
* @return [4][transform_len] FD CLT data
*/
public double[][] correlateSingleColorFD(
double [][] clt_data1,
double [][] clt_data2,
double [][] tcorr){ // null or initialized to [4][transform_len]
if (tcorr == null) tcorr = new double [4][transform_len];
if ((clt_data1 == null) || (clt_data1 == null)) return null; // to work with missing colors for monochrome
for (int i = 0; i < transform_len; i++) {
for (int n = 0; n<4; n++){
tcorr[n][i] = 0;
for (int k=0; k<4; k++){
if (ZI[n][k] < 0)
tcorr[n][i] -=
clt_data1[-ZI[n][k]][i] * clt_data2[k][i];
else
tcorr[n][i] +=
clt_data1[ZI[n][k]][i] * clt_data2[k][i];
}
}
}
return tcorr;
}
/**
* Normalize 2D correlation in FD, LPF (if not null) and convert to pixel domain and trim
* @param tcorr FD representation of the correlation[4][64]
* @param lpf LPF [64] or null
* @param afat_zero2 fat zero to add during normalization, units of squared values
* @param corr_radius if >=0 and < 7 - extract only the central part of the 15x15 square
* @return 2D phase correlation in linescan order
*/
public double[] normalizeConvertCorr(
double [][] tcorr, // null or initialized to [4][transform_len]
double [] lpf,
double afat_zero2, // absolute fat zero, same units as components squared values
int corr_radius,
boolean debug_gpu){
if (tcorr == null) return null;
double afat_zero4 = afat_zero2*afat_zero2;
for (int i = 0; i < transform_len; i++) {
double s = afat_zero4;
for (int n = 0; n< 4; n++){
s += tcorr[n][i]*tcorr[n][i];
}
double k = 1.0/ Math.sqrt(s);
for (int n = 0; n< 4; n++){
tcorr[n][i]*= k;
}
}
if (debug_gpu) {
System.out.println("=== NORMALIZED CORRELATION , afat_zero2="+afat_zero2+", afat_zero4="+afat_zero4+" ===");
for (int dct_mode = 0; dct_mode < 4; dct_mode++) {
System.out.println("------dct_mode="+dct_mode);
for (int i = 0; i < transform_size; i++) {
for (int j = 0; j < transform_size; j++) {
System.out.print(String.format("%10.5f ", tcorr[dct_mode][transform_size * i + j]));
}
System.out.println();
}
}
}
if (lpf != null) {
if (debug_gpu) {
System.out.println("=== LPF for CORRELATION ===");
for (int i = 0; i < transform_size; i++) {
for (int j = 0; j < transform_size; j++) {
System.out.print(String.format("%10.5f ", lpf[transform_size * i + j]));
}
System.out.println();
}
}
for (int n = 0; n<4; n++) {
for (int i = 0; i < transform_len; i++) {
tcorr[n][i] *= lpf[i];
}
}
}
if (debug_gpu) {
System.out.println("=== LPF-ed CORRELATION ===");
for (int dct_mode = 0; dct_mode < 4; dct_mode++) {
System.out.println("------dct_mode="+dct_mode);
for (int i = 0; i < transform_size; i++) {
for (int j = 0; j < transform_size; j++) {
System.out.print(String.format("%10.5f ", tcorr[dct_mode][transform_size * i + j]));
}
System.out.println();
}
}
}
for (int quadrant = 0; quadrant < 4; quadrant++){
int mode = ((quadrant << 1) & 2) | ((quadrant >> 1) & 1); // transpose
tcorr[quadrant] = dtt.dttt_iie(tcorr[quadrant], mode, transform_size, debug_gpu); // not orthogonal, term[0] is NOT *= 1/sqrt(2)
}
if (debug_gpu) {
System.out.println("=== CONVERTED CORRELATION ===");
for (int dct_mode = 0; dct_mode < 4; dct_mode++) {
System.out.println("------dct_mode="+dct_mode);
for (int i = 0; i < transform_size; i++) {
for (int j = 0; j < transform_size; j++) {
System.out.print(String.format("%10.5f ", tcorr[dct_mode][transform_size * i + j]));
}
System.out.println();
}
}
}
// convert from 4 quadrants to 15x15 centered tiles (only composite)
double [] corr_pd = dtt.corr_unfold_tile(tcorr, transform_size);
if (debug_gpu) {
int corr_size = 2* transform_size -1;
System.out.println("=== UNFOLDED CORRELATION ===");
for (int i = 0; i < corr_size; i++) {
for (int j = 0; j < corr_size; j++) {
System.out.print(String.format("%10.5f ", corr_pd[corr_size * i + j]));
}
System.out.println();
}
}
if ((corr_radius <= 0) || (corr_radius >= (transform_size - 1))) {
return corr_pd;
}
int full_size = 2 * transform_size - 1;
int trimmed_size = 2 * corr_radius + 1;
int trim = transform_size - 1 - corr_radius;
double [] trimmed_pd = new double [trimmed_size * trimmed_size];
int ioffs = (full_size + 1)*trim;
for (int orow = 0; orow < trimmed_size; orow++) {
System.arraycopy(corr_pd, orow*full_size + ioffs, trimmed_pd, orow*trimmed_size, trimmed_size);
}
return trimmed_pd;
}
/**
* Multiply CLT data of two channels, normalize amplitude, OK with null inputs (missing colors for monochrome images)
* @param clt_data1 first operand FD CLT data[4][transform_len]
......@@ -343,7 +484,7 @@ public class Correlation2d {
double scale_value, // scale correlation value
double [] col_weights,
double fat_zero) {
double [][][][] clt_data_tile = new double[clt_data.length][][][];
double [][][][] clt_data_tile = new double[clt_data.length][][][]; // [camera][color][quadrant][index]
for (int ncam = 0; ncam < clt_data.length; ncam++) if (clt_data[ncam] != null){
clt_data_tile[ncam] = new double[clt_data[ncam].length][][];
for (int ncol = 0; ncol < clt_data[ncam].length; ncol++) if ((clt_data[ncam][ncol] != null) && (clt_data[ncam][ncol][tileY] != null)){
......
src/main/java/com/elphel/imagej/tileprocessor/DttRad2.java
View file @ 84eeaf35
......@@ -613,7 +613,61 @@ public class DttRad2 {
return y;
}
public double [] dttt_iie(double [] x, int mode, int n, boolean debug_gpu){
double [] y = new double [n*n];
double [] line = new double[n];
// first (horizontal) pass
for (int i = 0; i<n; i++){
System.arraycopy(x, n*i, line, 0, n);
line = ((mode & 1)!=0)? dstiie_direct(line):dctiie_direct(line);
for (int j=0; j < n;j++) y[j*n+i] =line[j]; // transpose
}
if (debug_gpu) {
System.out.println("------after hor, mode="+mode);
for (int i = 0; i < n; i++) {
for (int j = 0; j < n; j++) {
System.out.print(String.format("%10.5f ", y[n * i + j]));
}
System.out.println();
}
}
// second (vertical) pass
for (int i = 0; i<n; i++){
System.arraycopy(y, n*i, line, 0, n);
line = ((mode & 2)!=0)? dstiie_direct(line):dctiie_direct(line);
System.arraycopy(line, 0, y, n*i, n);
}
if (debug_gpu) {
System.out.println("------after vert, mode="+mode);
for (int i = 0; i < n; i++) {
for (int j = 0; j < n; j++) {
System.out.print(String.format("%10.5f ", y[n * i + j]));
}
System.out.println();
}
}
return y;
}
/*
if (debug_gpu) {
System.out.println("=== CONVERTED CORRELATION ===");
for (int dct_mode = 0; dct_mode < 4; dct_mode++) {
System.out.println("------dct_mode="+dct_mode);
for (int i = 0; i < transform_size; i++) {
for (int j = 0; j < transform_size; j++) {
System.out.print(String.format("%10.3f ", tcorr[dct_mode][transform_size * i + j]));
}
System.out.println();
}
}
}
*/
public double [] dttt_iii(double [] x){
......@@ -780,7 +834,7 @@ public class DttRad2 {
}
public double [] dctii_direct(double[] x){
public double [] dctii_direct(double[] x){ // orthogonal, term[0] *= 1/sqrt(2)
int n = x.length;
int t = ilog2(n)-1;
if (CII==null){
......@@ -796,7 +850,7 @@ public class DttRad2 {
return y;
}
public double [] dctiie_direct(double[] x){
public double [] dctiie_direct(double[] x){ // not orthogonal
int n = x.length;
int t = ilog2(n)-1;
if (CIIe==null){
......@@ -928,7 +982,7 @@ public class DttRad2 {
}
}
private void setup_CII(int maxN){
private void setup_CII(int maxN){ // orthogonal, term[0] *= 1/sqrt(2)
if (maxN > N) setup_arrays(maxN);
int l = ilog2(N);
if (!(CII==null) && (CII.length >= l)) return;
......@@ -949,7 +1003,7 @@ public class DttRad2 {
}
}
private void setup_CIIe(int maxN){
private void setup_CIIe(int maxN){ // not orthogonal
if (maxN > N) setup_arrays(maxN);
int l = ilog2(N);
if (!(CIIe==null) && (CIIe.length >= l)) return;
......
src/main/java/com/elphel/imagej/tileprocessor/ImageDtt.java
View file @ 84eeaf35
This source diff could not be displayed because it is too large. You can view the blob instead.
src/main/java/com/elphel/imagej/tileprocessor/MacroCorrelation.java
View file @ 84eeaf35
......@@ -307,6 +307,7 @@ public class MacroCorrelation {
// double [][][][] texture_tiles = save_textures ? new double [tilesY][tilesX][][] : null; // ["RGBA".length()][];
ImageDtt image_dtt = new ImageDtt(
clt_parameters.transform_size,
this.mtp.isMonochrome(),
clt_parameters.getScaleStrength(this.mtp.isAux()));
image_dtt.clt_aberrations_quad_corr(
......@@ -350,7 +351,7 @@ public class MacroCorrelation {
null, // final GeometryCorrection geometryCorrection_main, // if not null correct this camera (aux) to the coordinates of the main
null, // clt_kernels, // final double [][][][][][] clt_kernels, // [channel_in_quad][color][tileY][tileX][band][pixel] , size should match image (have 1 tile around)
clt_parameters.kernel_step,
clt_parameters.transform_size,
/// clt_parameters.transform_size,
clt_parameters.clt_window,
shiftXY, //
0.0, // disparity_corr, // final double disparity_corr, // disparity at infinity
......
src/main/java/com/elphel/imagej/tileprocessor/TileProcessor.java
View file @ 84eeaf35
......@@ -5806,7 +5806,10 @@ public class TileProcessor {
// show testure_tiles
double [][][][] texture_tiles = scan_prev.getTextureTiles();
ImageDtt image_dtt = new ImageDtt(isMonochrome(), clt_parameters.getScaleStrength(is_aux));
ImageDtt image_dtt = new ImageDtt(
clt_parameters.transform_size,
isMonochrome(),
clt_parameters.getScaleStrength(is_aux));
double [][][] dispStrength = st.getDisparityStrengths(
clt_parameters.stMeasSel); // int stMeasSel) // = 1; // Select measurements for supertiles : +1 - combo, +2 - quad +4 - hor +8 - vert)
......@@ -5830,15 +5833,15 @@ public class TileProcessor {
if (!batch_mode && show_nonoverlap){
texture_nonoverlap = image_dtt.combineRBGATiles(
texture_tiles, // array [tp.tilesY][tp.tilesX][4][4*transform_size] or [tp.tilesY][tp.tilesX]{null}
clt_parameters.transform_size,
/// image_dtt.transform_size,
false, // when false - output each tile as 16x16, true - overlap to make 8x8
clt_parameters.sharp_alpha, // combining mode for alpha channel: false - treat as RGB, true - apply center 8x8 only
threadsMax, // maximal number of threads to launch
debugLevel);
sdfa_instance.showArrays(
texture_nonoverlap,
tilesX * (2 * clt_parameters.transform_size),
tilesY * (2 * clt_parameters.transform_size),
tilesX * (2 * image_dtt.transform_size),
tilesY * (2 * image_dtt.transform_size),
true,
name + "-TXTNOL-D",
(clt_parameters.keep_weights?rgba_weights_titles:rgba_titles));
......@@ -5849,7 +5852,7 @@ public class TileProcessor {
int alpha_index = 3;
texture_overlap = image_dtt.combineRBGATiles(
texture_tiles, // array [tp.tilesY][tp.tilesX][4][4*transform_size] or [tp.tilesY][tp.tilesX]{null}
clt_parameters.transform_size,
/// image_dtt.transform_size,
true, // when false - output each tile as 16x16, true - overlap to make 8x8
clt_parameters.sharp_alpha, // combining mode for alpha channel: false - treat as RGB, true - apply center 8x8 only
threadsMax, // maximal number of threads to launch
......@@ -5868,8 +5871,8 @@ public class TileProcessor {
if (show_overlap) {
sdfa_instance.showArrays(
texture_overlap,
tilesX * clt_parameters.transform_size,
tilesY * clt_parameters.transform_size,
tilesX * image_dtt.transform_size,
tilesY * image_dtt.transform_size,
true,
name + "-TXTOL-D",
(clt_parameters.keep_weights?rgba_weights_titles:rgba_titles));
......@@ -7182,7 +7185,9 @@ public class TileProcessor {
CLTPass3d scan_prev = clt_3d_passes.get(clt_3d_passes.size() -1); // get last one
boolean [] these_tiles = scan_prev.getSelected();
DisparityProcessor dp = new DisparityProcessor(this, clt_parameters.transform_size * geometryCorrection.getScaleDzDx());
DisparityProcessor dp = new DisparityProcessor(
this,
clt_parameters.transform_size * geometryCorrection.getScaleDzDx());
boolean [] grown = these_tiles.clone();
growTiles(
2, // grow tile selection by 1 over non-background tiles 1: 4 directions, 2 - 8 directions, 3 - 8 by 1, 4 by 1 more
......
src/main/resources/TileProcessor.cuh
View file @ 84eeaf35
This diff is collapsed.
src/main/resources/dtt8x8.cuh
View file @ 84eeaf35
......@@ -36,10 +36,10 @@
* \brief DCT-II, DST-II, DCT-IV and DST-IV for Complex Lapped Transform of 16x16 (stride 8)
* in GPU
* This file contains building blocks for the 16x16 stride 8 COmplex Lapped Transform (CLT)
* imlementation. DTT-IV are used for forward and inverse 2D CLT, DTT-II - to convert correlation
* implementation. DTT-IV are used for forward and inverse 2D CLT, DTT-II - to convert correlation
* results from the frequency to pixel domain. DTT-III (inverse of DTT-II) is not implemented
* here it is used to convert convolution kernels and LPF to the frequency domain - done in
* softwaer.
* software.
*
* This file is cpompatible with both runtime and driver API, runtime is used for development
* with Nvidia Nsight, driver API when calling these kernels from Java
......@@ -84,23 +84,24 @@ __constant__ float SINN1[] = {0.195090f,0.555570f};
__constant__ float SINN2[] = {0.098017f,0.290285f,0.471397f,0.634393f};
inline __device__ void dttii_shared_mem(float * x0, int inc, int dst_not_dct);
inline __device__ void dttiv_shared_mem(float * x0, int inc, int dst_not_dct);
inline __device__ void dttiv_nodiverg(float * x, int inc, int dst_not_dct);
inline __device__ void dctiv_nodiverg(float * x0, int inc);
inline __device__ void dstiv_nodiverg(float * x0, int inc);
inline __device__ void dttii_shared_mem_nonortho(float * x0, int inc, int dst_not_dct); // does not scale by y[0] (y[7]) by 1/sqrt[0]
inline __device__ void dttii_shared_mem(float * x0, int inc, int dst_not_dct); // used in GPU_DTT24_DRV
inline __device__ void dttiv_shared_mem(float * x0, int inc, int dst_not_dct); // used in GPU_DTT24_DRV
inline __device__ void dttiv_nodiverg (float * x, int inc, int dst_not_dct); // not used
inline __device__ void dctiv_nodiverg (float * x0, int inc); // used in TP
inline __device__ void dstiv_nodiverg (float * x0, int inc); // used in TP
inline __device__ void dct_ii8 ( float x[8], float y[8]); // x,y point to 8-element arrays each
inline __device__ void dct_iv8 ( float x[8], float y[8]); // x,y point to 8-element arrays each
inline __device__ void dst_iv8 ( float x[8], float y[8]); // x,y point to 8-element arrays each
inline __device__ void _dctii_nrecurs8 ( float x[8], float y[8]); // x,y point to 8-element arrays each
inline __device__ void _dctiv_nrecurs8 ( float x[8], float y[8]); // x,y point to 8-element arrays each
inline __device__ void dct_ii8 ( float x[8], float y[8]); // x,y point to 8-element arrays each // not used
inline __device__ void dct_iv8 ( float x[8], float y[8]); // x,y point to 8-element arrays each // not used
inline __device__ void dst_iv8 ( float x[8], float y[8]); // x,y point to 8-element arrays each // not used
inline __device__ void _dctii_nrecurs8 ( float x[8], float y[8]); // x,y point to 8-element arrays each // not used
inline __device__ void _dctiv_nrecurs8 ( float x[8], float y[8]); // x,y point to 8-element arrays each // not used
/**
**************************************************************************
* Converts 2D image (in the GPU memory) using 8x8 DTT 8x8 tiles.
* Mostly for testing and profiling individual converions
* Mostly for testing and profiling individual conversions
*
* \param dst [OUT] - Coefficients as 8x8 tiles
* \param src [IN] - Source image of floats
......@@ -376,6 +377,88 @@ inline __device__ void dttii_shared_mem(float * x0, int inc, int dst_not_dct)
}
}
inline __device__ void dttii_shared_mem_nonortho(float * x0, int inc, int dst_not_dct)
{
float *x1 = x0 + inc;
float *x2 = x1 + inc;
float *x3 = x2 + inc;
float *x4 = x3 + inc;
float *x5 = x4 + inc;
float *x6 = x5 + inc;
float *x7 = x6 + inc;
float u00, u01, u02, u03, u10, u11, u12, u13;
if (dst_not_dct) { // DSTII
// invert odd input samples
u00= ( (*x0) - (*x7));
u10= ( (*x0) + (*x7));
u01= (-(*x1) + (*x6));
u11= (-(*x1) - (*x6));
u02= ( (*x2) - (*x5));
u12= ( (*x2) + (*x5));
u03= (-(*x3) + (*x4));
u13= (-(*x3) - (*x4));
} else { // DCTII
u00= ( (*x0) + (*x7));
u10= ( (*x0) - (*x7));
u01= ( (*x1) + (*x6));
u11= ( (*x1) - (*x6));
u02= ( (*x2) + (*x5));
u12= ( (*x2) - (*x5));
u03= ( (*x3) + (*x4));
u13= ( (*x3) - (*x4));
}
// _dctii_nrecurs4(u00,u01, u02, u03, &v00, &v01, &v02, &v03);
float w00= u00 + u03;
float w10= u00 - u03;
float w01= (u01 + u02);
float w11= (u01 - u02);
float v01= COSPI_1_8_SQRT2 * w10 + COSPI_3_8_SQRT2 * w11;
float v03= COSPI_3_8_SQRT2 * w10 - COSPI_1_8_SQRT2 * w11;
// _dctiv_nrecurs4(u10, u11, u12, u13, &v10, &v11, &v12, &v13);
float w20= ( COSN1[0] * u10 + SINN1[0] * u13);
float w30= (-SINN1[1] * u11 + COSN1[1] * u12);
float w21= ( COSN1[1] * u11 + SINN1[1] * u12);
float w31= -(-SINN1[0] * u10 + COSN1[0] * u13);
float v11 = w20 - w21 - w30 + w31;
float v12 = w20 - w21 + w30 - w31;
if (dst_not_dct) { // DSTII
// Invert output sequence
*x0 = (w30 + w31)* 0.5f; // v13 * SQRT1_8; z10 * 0.5f
*x1 = v03 * SQRT1_8;
*x2 = v12 * SQRT1_8;
*x3 = (w00 - w01) * SQRT1_8; // v02 * SQRT1_8
*x4 = v11 * SQRT1_8;
*x5 = v01 * SQRT1_8;
*x6 = (w20 + w21) * 0.5f; // v10 * SQRT1_8; z00 * 0.5f;
*x7 = (w00 + w01) * 0.5f; // SQRT1_8; // v00 * SQRT1_8 //*** no 1/sqrt(2)!
} else {
*x0 = (w00 + w01) * 0.5f; // SQRT1_8; // v00 * SQRT1_8 //*** no 1/sqrt(2)!
*x1 = (w20 + w21) * 0.5f; // v10 * SQRT1_8; z00 * 0.5f;
*x2 = v01 * SQRT1_8;
*x3 = v11 * SQRT1_8;
*x4 = (w00 - w01) * SQRT1_8; // v02 * SQRT1_8
*x5 = v12 * SQRT1_8;
*x6 = v03 * SQRT1_8;
*x7 = (w30 + w31)* 0.5f; // v13 * SQRT1_8; z10 * 0.5f
}
}
inline __device__ void dttiv_shared_mem(float * x0, int inc, int dst_not_dct)
{
......
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