jidt/cuda/kraskovCuda.c

391 lines
13 KiB
C

#include <jni.h>
#include <stdlib.h>
#include <stdio.h>
#define check(ans) { _check((ans), __FILE__, __LINE__); }
#include "gpuMILibrary.h"
#include "gpuCMILibrary.h"
#include "ctimer.h"
#ifdef __cplusplus
extern "C" {
#endif
/*
* Class: infodynamics_measures_continuous_kraskov_MutualInfoCalculatorMultiVariateKraskov
* Method: MIKraskov
* Signature: (I[DI[DIIIZZZIZ[I)[D
*/
JNIEXPORT jdoubleArray JNICALL
Java_infodynamics_measures_continuous_kraskov_MutualInfoCalculatorMultiVariateKraskov_MIKraskov(
JNIEnv *env, jobject thisObj, jint j_N,
jobjectArray j_sourceArray, jint j_dimx,
jobjectArray j_destArray, jint j_dimy,
jint j_k, jint j_theiler, jboolean j_returnLocals,
jboolean j_useMaxNorm, jboolean j_isAlgorithm1, jint j_nbSurrogates,
jboolean j_reorderingsGiven, jobjectArray j_orderings) {
// Check that incoming data has correct size
// =====================
jsize sourceLength = (*env)->GetArrayLength(env, j_sourceArray);
jsize destLength = (*env)->GetArrayLength(env, j_destArray);
// if (sourceLength != j_N || destLength != j_N || (j_N%(j_nbSurrogates+1) != 0)) {
if (sourceLength != j_N || destLength != j_N) {
jclass Exception = (*env)->FindClass(env, "java/lang/Exception");
(*env)->ThrowNew(env, Exception, "Data has wrong length.");
}
if (!j_isAlgorithm1) {
jclass Exception = (*env)->FindClass(env, "java/lang/Exception");
(*env)->ThrowNew(env, Exception, "Only algorithm 1 is supported.");
}
if ((j_returnLocals || !j_isAlgorithm1) && (j_nbSurrogates > 0)) {
jclass Exception = (*env)->FindClass(env, "java/lang/Exception");
(*env)->ThrowNew(env, Exception, "Surrogates only supported for average MI with KSG1.");
}
// Copy variables from Java
// =====================
int N = j_N;
int k = j_k;
int dimx = j_dimx;
int dimy = j_dimy;
int theiler = j_theiler;
int returnLocals = j_returnLocals ? 1 : 0;
int useMaxNorm = j_useMaxNorm ? 1 : 0;
int isAlgorithm1 = j_isAlgorithm1 ? 1 : 0;
int nb_surrogates = j_nbSurrogates;
int reorderingsGiven = j_reorderingsGiven ? 1 : 0;
CPerfTimer pt = startTimer("Java array copy");
float *source = (float *) malloc(N * dimx * sizeof(float));
float *dest = (float *) malloc(N * dimy * sizeof(float));
if (NULL == source || NULL == dest) {
jclass Exception = (*env)->FindClass(env, "java/lang/Exception");
(*env)->ThrowNew(env, Exception, "Error allocating data.");
}
for (int i = 0; i < N; i++) {
jdoubleArray j_sourceRow = (jdoubleArray) (*env)->GetObjectArrayElement(env, j_sourceArray, i);
jdoubleArray j_destRow = (jdoubleArray) (*env)->GetObjectArrayElement(env, j_destArray, i);
jdouble *sourceRow = (*env)->GetDoubleArrayElements(env, j_sourceRow, NULL);
jdouble *destRow = (*env)->GetDoubleArrayElements(env, j_destRow, NULL);
// Data in java are doubles, but GPUs need floats.
// We have to cast them manually, so we can't memcopy
// The following two for-loops get two matrices in T-by-D indexing (i.e.
// first dimension is time, second is variable) and return the data in
// column-major form
for (int j = 0; j < dimx; j++) {
source[N*j + i] = (float) sourceRow[j];
}
for (int j = 0; j < dimy; j++) {
dest[N*j + i] = (float) destRow[j];
}
// // The following two for-loops get the same matrices but return arrays
// // in row-major form. Code is left here for future development purposes.
// for (j = 0; j < dimx; j++) {
// register float x = (float) sourceRow[j];
// source[dimx*i + j] = x;
// pointset[dims*i + j] = x;
// }
// for (j = 0; j < dimy; j++) {
// register float x = (float) destRow[j];
// dest[dimy*i + j] = x
// pointset[i*dims + dimx + j] = x;
// }
(*env)->ReleaseDoubleArrayElements(env, j_sourceRow, sourceRow, 0);
(*env)->ReleaseDoubleArrayElements(env, j_destRow, destRow, 0);
(*env)->DeleteLocalRef(env, j_sourceRow);
(*env)->DeleteLocalRef(env, j_destRow);
// FIXME: I'm not entirely sure I'm freeing all the memory here. I should
// check for memory leaks more carefully.
}
int **reorderings = NULL;
if (reorderingsGiven) {
reorderings = (int **) malloc(nb_surrogates * sizeof(int *));
for (int i = 0; i < nb_surrogates; i++) {
jintArray j_order = (jdoubleArray) (*env)->GetObjectArrayElement(env, j_orderings, i);
jint *order = (*env)->GetIntArrayElements(env, j_order, NULL);
reorderings[i] = (int *) malloc(N * sizeof(int));
for (int j = 0; j < N; j++) {
reorderings[i][j] = order[j];
}
(*env)->ReleaseIntArrayElements(env, j_order, order, 0);
(*env)->DeleteLocalRef(env, j_order);
}
}
stopTimer(pt);
// Call C function
// =========================
int resultSize;
if (returnLocals) {
resultSize = N;
} else if (nb_surrogates > 0) {
resultSize = nb_surrogates + 1;
} else {
resultSize = 3;
}
float *result = (float *) malloc(resultSize * sizeof(float));
jidt_error_t ret;
if (!reorderingsGiven) {
ret = MIKraskov_C(N, source, dimx, dest, dimy,
k, theiler, nb_surrogates, returnLocals, useMaxNorm,
isAlgorithm1, result);
} else {
ret = MIKraskovWithReorderings(N, source, dimx, dest, dimy, k, theiler,
nb_surrogates, returnLocals, useMaxNorm,
isAlgorithm1, result, reorderingsGiven,
reorderings);
}
if (JIDT_ERROR == ret) {
jclass Exception = (*env)->FindClass(env, "java/lang/Exception");
(*env)->ThrowNew(env, Exception, "Error in GPU execution.");
}
// Free memory and return
// =========================
if (source) free(source);
if (dest) free(dest);
if (reorderingsGiven) {
for (int i = 0; i < nb_surrogates; i++) {
if (reorderings[i]) free(reorderings[i]);
}
if (reorderings) free (reorderings);
}
jdouble outCArray[resultSize];
for (int i = 0; i < resultSize; i++) {
outCArray[i] = result[i];
}
// Set Java array for return
jdoubleArray outJNIArray = (*env)->NewDoubleArray(env, resultSize); // allocate
if (NULL == outJNIArray) {
jclass Exception = (*env)->FindClass(env, "java/lang/Exception");
(*env)->ThrowNew(env, Exception, "Error creating return array.");
}
(*env)->SetDoubleArrayRegion(env, outJNIArray, 0 , resultSize, outCArray); // copy
if (result) { free(result); }
return outJNIArray;
} // End of function MIKraskov
/*
* Class: infodynamics_measures_continuous_kraskov_ConditionalMutualInfoCalculatorMultiVariateKraskov
* Method: CMIKraskov
* Signature: (I[DI[DI[DIIIZZZIZ[II)[D
*/
JNIEXPORT jdoubleArray JNICALL
Java_infodynamics_measures_continuous_kraskov_ConditionalMutualInfoCalculatorMultiVariateKraskov_CMIKraskov(
JNIEnv *env, jobject thisObj, jint j_N,
jobjectArray j_sourceArray, jint j_dimx,
jobjectArray j_destArray, jint j_dimy,
jobjectArray j_condArray, jint j_dimz,
jint j_k, jint j_theiler, jboolean j_returnLocals,
jboolean j_useMaxNorm, jboolean j_isAlgorithm1, jint j_nbSurrogates,
jboolean j_reorderingsGiven, jobjectArray j_orderings,
jint j_variableToReorder) {
// Check that incoming data has correct size
// =====================
jsize sourceLength = (*env)->GetArrayLength(env, j_sourceArray);
jsize destLength = (*env)->GetArrayLength(env, j_destArray);
jsize condLength = (*env)->GetArrayLength(env, j_condArray);
// if (sourceLength != j_N || destLength != j_N || (j_N%(j_nbSurrogates+1) != 0)) {
if (sourceLength != j_N || destLength != j_N || condLength != j_N) {
jclass Exception = (*env)->FindClass(env, "java/lang/Exception");
(*env)->ThrowNew(env, Exception, "Data has wrong length.");
}
if (!j_isAlgorithm1) {
jclass Exception = (*env)->FindClass(env, "java/lang/Exception");
(*env)->ThrowNew(env, Exception, "Only algorithm 1 is supported.");
}
if ((j_returnLocals || !j_isAlgorithm1) && (j_nbSurrogates > 0)) {
jclass Exception = (*env)->FindClass(env, "java/lang/Exception");
(*env)->ThrowNew(env, Exception, "Surrogates only supported for average MI with KSG1.");
}
// Copy variables from Java
// =====================
int N = j_N;
int k = j_k;
int dimx = j_dimx;
int dimy = j_dimy;
int dimz = j_dimz;
int theiler = j_theiler;
int returnLocals = j_returnLocals ? 1 : 0;
int useMaxNorm = j_useMaxNorm ? 1 : 0;
int isAlgorithm1 = j_isAlgorithm1 ? 1 : 0;
int nb_surrogates = j_nbSurrogates;
int reorderingsGiven = j_reorderingsGiven ? 1 : 0;
int variableToReorder = j_variableToReorder;
CPerfTimer pt = startTimer("Java array copy");
float *source = (float *) malloc(N * dimx * sizeof(float));
float *dest = (float *) malloc(N * dimy * sizeof(float));
float *cond = (float *) malloc(N * dimz * sizeof(float));
if (NULL == source || NULL == dest || NULL == cond) {
jclass Exception = (*env)->FindClass(env, "java/lang/Exception");
(*env)->ThrowNew(env, Exception, "Error allocating data.");
}
for (int i = 0; i < N; i++) {
jdoubleArray j_sourceRow = (jdoubleArray) (*env)->GetObjectArrayElement(env, j_sourceArray, i);
jdoubleArray j_destRow = (jdoubleArray) (*env)->GetObjectArrayElement(env, j_destArray, i);
jdoubleArray j_condRow = (jdoubleArray) (*env)->GetObjectArrayElement(env, j_condArray, i);
jdouble *sourceRow = (*env)->GetDoubleArrayElements(env, j_sourceRow, NULL);
jdouble *destRow = (*env)->GetDoubleArrayElements(env, j_destRow, NULL);
jdouble *condRow = (*env)->GetDoubleArrayElements(env, j_condRow, NULL);
// Data in java are doubles, but GPUs need floats.
// We have to cast them manually, so we can't memcopy
// The following for-loops get two matrices in T-by-D indexing (i.e.
// first dimension is time, second is variable) and return the data in
// column-major form
for (int j = 0; j < dimx; j++) {
source[N*j + i] = (float) sourceRow[j];
}
for (int j = 0; j < dimy; j++) {
dest[N*j + i] = (float) destRow[j];
}
for (int j = 0; j < dimz; j++) {
cond[N*j + i] = (float) condRow[j];
}
(*env)->ReleaseDoubleArrayElements(env, j_sourceRow, sourceRow, 0);
(*env)->ReleaseDoubleArrayElements(env, j_destRow, destRow, 0);
(*env)->ReleaseDoubleArrayElements(env, j_condRow, condRow, 0);
(*env)->DeleteLocalRef(env, j_sourceRow);
(*env)->DeleteLocalRef(env, j_destRow);
(*env)->DeleteLocalRef(env, j_condRow);
// FIXME: I'm not entirely sure I'm freeing all the memory here. I should
// check for memory leaks more carefully.
}
int **reorderings = NULL;
if (reorderingsGiven) {
reorderings = (int **) malloc(nb_surrogates * sizeof(int *));
for (int i = 0; i < nb_surrogates; i++) {
jintArray j_order = (jdoubleArray) (*env)->GetObjectArrayElement(env, j_orderings, i);
jint *order = (*env)->GetIntArrayElements(env, j_order, NULL);
reorderings[i] = (int *) malloc(N * sizeof(int));
for (int j = 0; j < N; j++) {
reorderings[i][j] = order[j];
}
(*env)->ReleaseIntArrayElements(env, j_order, order, 0);
(*env)->DeleteLocalRef(env, j_order);
}
}
stopTimer(pt);
// Call C function
// =========================
int resultSize;
if (returnLocals) {
resultSize = N;
} else if (nb_surrogates > 0) {
resultSize = nb_surrogates + 1;
} else {
resultSize = 6;
}
float *result = (float *) malloc(resultSize * sizeof(float));
jidt_error_t ret;
if (!reorderingsGiven) {
ret = CMIKraskov_C(N, source, dimx, dest, dimy, cond, dimz,
k, theiler, nb_surrogates, returnLocals, useMaxNorm,
isAlgorithm1, result, variableToReorder);
} else {
ret = CMIKraskovWithReorderings(N, source, dimx, dest, dimy, cond, dimz, k, theiler,
nb_surrogates, returnLocals, useMaxNorm,
isAlgorithm1, result, reorderingsGiven,
reorderings, variableToReorder);
}
if (JIDT_ERROR == ret) {
jclass Exception = (*env)->FindClass(env, "java/lang/Exception");
(*env)->ThrowNew(env, Exception, "Error in GPU execution.");
}
// Free memory and return
// =========================
if (source) free(source);
if (dest) free(dest);
if (cond) free(cond);
if (reorderingsGiven) {
for (int i = 0; i < nb_surrogates; i++) {
if (reorderings[i]) free(reorderings[i]);
}
if (reorderings) free (reorderings);
}
jdouble outCArray[resultSize];
for (int i = 0; i < resultSize; i++) {
outCArray[i] = result[i];
}
// Set Java array for return
jdoubleArray outJNIArray = (*env)->NewDoubleArray(env, resultSize); // allocate
if (NULL == outJNIArray) {
jclass Exception = (*env)->FindClass(env, "java/lang/Exception");
(*env)->ThrowNew(env, Exception, "Error creating return array.");
}
(*env)->SetDoubleArrayRegion(env, outJNIArray, 0 , resultSize, outCArray); // copy
if (result) { free(result); }
return outJNIArray;
} // End of function CMIKraskov
#ifdef __cplusplus
}
#endif