Finding incorrect implementation of JudyArray - c++11

I'm trying to give a better error report (possible bug) for this case (about judySArray give incorrect result, but I don't know which key that give incorrect result).
The code here from this folder, note on this blog. Dependencies: judySArray.h and cedar.h
// judy.cpp
#include "deps/judySArray.h"
#include <string>
#include <iostream>
#include <cstdlib>
#include <cstring>
using namespace std;
typedef judySArray<double> MSD;
const int MAX_DATA = 12000000;
const char i2ch[] = {'0','1','2','3','4','5','6','7','8','9','a','B','c','D','e','F'};
int get_first_digit(double d) {
while(d > 10) d /= 10;
return d;
}
string to_rhex(int v) {
char hex[32];
int start = 0;
while(v>0) {
hex[start] = i2ch[v%16];
v /= 16;
++start;
}
hex[start] = 0;
return hex;
}
void add_or_inc(MSD &m, const string& key,double set, double inc, int& ctr) {
const char* cstr = key.c_str();
double it = m.find(cstr);
if(!it) {
m.insert(cstr,set);
return;
}
m.insert(cstr,it+inc);
++ctr;
}
int main() {
MSD m(64);
int dup1 = 0, dup2 = 0, dup3 = 0;
for(int z=MAX_DATA;z>0;--z) {
int val2 = MAX_DATA-z;
int val3 = MAX_DATA*2-z;
string key1 = to_string(z);
string key2 = to_string(val2);
string key3 = to_rhex(val3);
add_or_inc(m,key1,z,val2,dup1);
add_or_inc(m,key2,val2,val3,dup2);
add_or_inc(m,key3,val3,z,dup3);
}
cout << dup1 << ' ' << dup2 << ' ' << dup3 << endl;
int total = 0, verify = 0, count = 0;
for(auto &it = m.begin();m.success(); m.next()) {
total += get_first_digit(it.value);
verify += strlen((const char *) it.key);
count += 1;
}
cout << total << ' ' << verify << ' ' << count << endl;
}
other implementation (map, unordered_map, hat-trie and cedar) give correct result:
6009354 6009348 611297
36186112 159701682 23370001
but judy didn't:
6009354 6009348 611297
36186112 159701681 23370000
The problem is, which key that have incorrect result?
I've tried to build a code that insert those keys on another data structure (that is cedar), but that incorrect keys still not detected:
// judy.cpp
#include "deps/judySArray.h"
#include <string>
#include <iostream>
#include <cstdlib>
#include <cstring>
#include <vector>
using namespace std;
typedef judySArray<double> MSD;
const int MAX_DATA = 12000000;
const char i2ch[] = {'0','1','2','3','4','5','6','7','8','9','a','B','c','D','e','F'};
int get_first_digit(double d) {
while(d > 10) d /= 10;
return d;
}
string to_rhex(int v) {
char hex[32];
int start = 0;
while(v>0) {
hex[start] = i2ch[v%16];
v /= 16;
++start;
}
hex[start] = 0;
return hex;
}
void add_or_inc(MSD &m, const string& key,double set, double inc, int& ctr) {
const char* cstr = key.c_str();
double it = m.find(cstr);
if(!it) {
m.insert(cstr,set);
return;
}
m.insert(cstr,it+inc);
++ctr;
}
#include "deps/cedar.h"
class MSD2 {
public:
vector<double> data;
typedef cedar::da<int> CI;
CI da;
bool exists(const string& key,double &old) {
int idx = -1;
bool found = da.exactMatchExists(key.c_str(),key.size(),&idx);
if(found) old = data[idx];
return found;
}
void insert(const string& key,double val) {
da.update(key.c_str(),key.size(),data.size());
data.push_back(val);
}
void update(const string& key,double val) {
int idx = -1;
bool found = da.exactMatchExists(key.c_str(),key.size(),&idx);
if(found) {
data[idx] = val;
return;
}
insert(key,val);
}
};
void add_or_inc(MSD2 &m, const string& key,double set, double inc, int& ctr) {
double old;
if(!m.exists(key,old)) {
m.insert(key,set);
return;
}
m.update(key,old+inc);
++ctr;
}
int main() {
MSD m(64);
MSD2 m2;
int dup1 = 0, dup2 = 0, dup3 = 0;
int vup1 = 0, vup2 = 0, vup3 = 0;
for(int z=MAX_DATA;z>0;--z) {
int val2 = MAX_DATA-z;
int val3 = MAX_DATA*2-z;
string key1 = to_string(z);
string key2 = to_string(val2);
string key3 = to_rhex(val3);
add_or_inc(m,key1,z,val2,dup1);
add_or_inc(m,key2,val2,val3,dup2);
add_or_inc(m,key3,val3,z,dup3);
add_or_inc(m2,key1,z,val2,vup1);
add_or_inc(m2,key2,val2,val3,vup2);
add_or_inc(m2,key3,val3,z,vup3);
}
cout << dup1 << ' ' << dup2 << ' ' << dup3 << endl;
cout << vup1 << ' ' << vup2 << ' ' << vup3 << endl;
int total = 0, verify = 0, count = 0;
int xotal = 0, xerify = 0, xount = 0;
union { int i; int x; } b;
size_t from = 0, p = 0;
char key[256] = {0};
for (b.i = m2.da.begin(from, p); b.i != MSD2::CI::CEDAR_NO_PATH; b.i = m2.da.next(from, p)) {
double it2 = m2.data[b.x]; // <-- find cedar's
xotal += get_first_digit(it2);
m2.da.suffix(key,p,from);
xerify += strlen(key);
xount += 1;
double it = m.find(key); // <-- find judy's
if(it != it2) { // if value doesn't match, print:
cout << "mismatch value for " << key << " : " << it2 << " vs " << it << endl;
}
}
for(auto &it = m.begin();m.success(); m.next()) {
total += get_first_digit(it.value);
verify += strlen((const char *) it.key);
count += 1;
}
cout << total << ' ' << verify << ' ' << count << endl;
cout << xotal << ' ' << xerify << ' ' << xount << endl;
}
compile with: clang++ -std=c++11 judy-findbug.cpp (or g++ -std=c++11)
the output would be:
6009354 6009348 611297
6009354 6009348 611297
36186112 159701681 23370000 <-- judy's
36186112 159701682 23370001 <-- cedar's
cedar has one more value than judy's (that is correct), but it didn't detected by the code above..
How to find that incorrect key(s)?

The bug on the code is someone (me) uncomment the assert(value != 0).
The bug was Karl's Judy implementation should not store null values (0 value).
Solution: use Doug Baskins' Judy implementation.

Related

How to read chunk of the data from a hdf5 file in c++?

I want to read a chunk of data which is just one frame of many frames stored in one dataset. The shape of the whole dataset is (10, 11214,3), 10 frames each frame has 11214 rows and 4 columns. Here is the file. The chunk I want to read would have the shape (11214,3). I can print the predefined array using, but I'm not sure how can I read data from a hdf5 file. Here is my code,
#include <h5xx/h5xx.hpp>
#include <boost/multi_array.hpp>
#include <iostream>
#include <vector>
#include <cstdio>
typedef boost::multi_array<int, 2> array_2d_t;
const int NI=10;
const int NJ=NI;
void print_array(array_2d_t const& array)
{
for (unsigned int j = 0; j < array.shape()[1]; j++)
{
for (unsigned int i = 0; i < array.shape()[0]; i++)
{
printf("%2d ", array[j][i]);
}
printf("\n");
}
}
void write_int_data(std::string const& filename, array_2d_t const& array)
{
h5xx::file file(filename, h5xx::file::trunc);
std::string name;
{
// --- create dataset and fill it with the default array data (positive values)
name = "integer array";
h5xx::create_dataset(file, name, array);
h5xx::write_dataset(file, name, array);
// --- create a slice object (aka hyperslab) to specify the location in the dataset to be overwritten
std::vector<int> offset; int offset_raw[2] = {4,4}; offset.assign(offset_raw, offset_raw + 2);
std::vector<int> count; int count_raw[2] = {2,2}; count.assign(count_raw, count_raw + 2);
h5xx::slice slice(offset, count);
}
}
void read_int_data(std::string const& filename)
{
h5xx::file file(filename, h5xx::file::in);
std::string name = "integer array";
// read and print the full dataset
{
array_2d_t array;
// --- read the complete dataset into array, the array is resized and overwritten internally
h5xx::read_dataset(file, name, array);
printf("original integer array read from file, negative number patch was written using a slice\n");
print_array(array);
printf("\n");
}
}
int main(int argc, char** argv)
{
std::string filename = argv[0];
filename.append(".h5");
// --- do a few demos/tests using integers
{
array_2d_t array(boost::extents[NJ][NI]);
{
const int nelem = NI*NJ;
int data[nelem];
for (int i = 0; i < nelem; i++)
data[i] = i;
array.assign(data, data + nelem);
}
write_int_data(filename, array);
read_int_data(filename);
}
return 0;
}
I'm using the h5xx — a template-based C++ wrapper for the HDF5 library link and boost library.
The datasets are stored in particles/lipids/box/positions path. The dataset name value holds the frames.
argv[0] is not what you want (arguments start at 1, 0 is the program name). Consider bounds checking as well:
std::vector<std::string> const args(argv, argv + argc);
std::string const filename = args.at(1) + ".h5";
the initialization can be done directly, without a temporary array (what is multi_array for, otherwise?)
for (size_t i = 0; i < array.num_elements(); i++)
array.data()[i] = i;
Or indeed, make it an algorithm:
std::iota(array.data(), array.data() + array.num_elements(), 0);
same with vectors:
std::vector<int> offset; int offset_raw[2] = {4,4}; offset.assign(offset_raw, offset_raw + 2);
std::vector<int> count; int count_raw[2] = {2,2}; count.assign(count_raw, count_raw + 2);
besides being a formatting mess can be simply
std::vector offset{4,4}, count{2,2};
h5xx::slice slice(offset, count);
On To The Real Question
The code has no relevance to the file. At all. I created some debug/tracing code to dump the file contents:
void dump(h5xx::group const& g, std::string indent = "") {
auto dd = g.datasets();
auto gg = g.groups();
for (auto it = dd.begin(); it != dd.end(); ++it) {
std::cout << indent << " ds:" << it.get_name() << "\n";
}
for (auto it = gg.begin(); it != gg.end(); ++it) {
dump(*it, indent + "/" + it.get_name());
}
}
int main()
{
h5xx::file xaa("xaa.h5", h5xx::file::mode::in);
dump(xaa);
}
Prints
/particles/lipids/box/edges ds:box_size
/particles/lipids/box/edges ds:step
/particles/lipids/box/edges ds:time
/particles/lipids/box/edges ds:value
/particles/lipids/box/positions ds:step
/particles/lipids/box/positions ds:time
/particles/lipids/box/positions ds:value
Now we can drill down to the dataset. Let's see whether we can figure out the correct type. It certainly is NOT array_2d_t:
h5xx::dataset ds(xaa, "particles/lipids/box/positions/value");
array_2d_t a;
h5xx::datatype detect(a);
std::cout << "type: " << std::hex << ds.get_type() << std::dec << "\n";
std::cout << "detect: " << std::hex << detect.get_type_id() << std::dec << "\n";
Prints
type: 30000000000013b
detect: 30000000000000c
That's a type mismatch. I guess I'll have to learn to read that gibberish as well...
Let's add some diagnostics:
void diag_type(hid_t type)
{
std::cout << " Class " << ::H5Tget_class(type) << std::endl;
std::cout << " Size " << ::H5Tget_size(type) << std::endl;
std::cout << " Sign " << ::H5Tget_sign(type) << std::endl;
std::cout << " Order " << ::H5Tget_order(type) << std::endl;
std::cout << " Precision " << ::H5Tget_precision(type) << std::endl;
std::cout << " NDims " << ::H5Tget_array_ndims(type) << std::endl;
std::cout << " NMembers " << ::H5Tget_nmembers(type) << std::endl;
}
int main()
{
h5xx::file xaa("xaa.h5", h5xx::file::mode::in);
// dump(xaa);
{
h5xx::group g(xaa, "particles/lipids/box/positions");
h5xx::dataset ds(g, "value");
std::cout << "dataset: " << std::hex << ds.get_type() << std::dec << std::endl;
diag_type(ds.get_type());
}
{
array_2d_t a(boost::extents[NJ][NI]);
h5xx::datatype detect(a);
std::cout << "detect: " << std::hex << detect.get_type_id() << std::dec << std::endl;
diag_type(detect.get_type_id());
}
}
Prints
dataset: 30000000000013b
Class 1
Size 4
Sign -1
Order 0
Precision 32
NDims -1
NMembers -1
detect: 30000000000000c
Class 0
Size 4
Sign 1
Order 0
Precision 32
NDims -1
NMembers -1
At least we know that HST_FLOAT (class 1) is required. Let's modify array_2d_t:
using array_2d_t = boost::multi_array<float, 2>;
array_2d_t a(boost::extents[11214][3]);
This at least makes the data appear similarly. Let's ... naively try to read:
h5xx::read_dataset(ds, a);
Oops, that predictably throws
terminate called after throwing an instance of 'h5xx::error'
what(): /home/sehe/Projects/stackoverflow/deps/h5xx/h5xx/dataset/boost_multi_array.hpp:176:read_dataset(): dataset "/particles/lipi
ds/box/positions/value" and target array have mismatching dimensions
No worries, we can guess:
using array_3d_t = boost::multi_array<float, 3>;
array_3d_t a(boost::extents[10][11214][3]);
h5xx::read_dataset(ds, a);
At least this does work. Adapting the print function:
template <typename T> void print_array(T const& array) {
for (auto const& row : array) {
for (auto v : row) printf("%5f ", v);
printf("\n");
}
}
Now we can print the first frame:
h5xx::read_dataset(ds, a);
print_array(*a.begin()); // print the first frame
This prints:
80.480003 35.360001 4.250000
37.450001 3.920000 3.960000
18.530001 -9.690000 4.680000
55.389999 74.339996 4.600000
22.110001 68.709999 3.850000
-4.130000 24.040001 3.730000
40.160000 6.390000 4.730000
-5.400000 35.730000 4.850000
36.669998 22.450001 4.080000
-3.680000 -10.660000 4.180000
(...)
That checks out with h5ls -r -d xaa.h5/particles/lipids/box/positions/value:
particles/lipids/box/positions/value Dataset {75/Inf, 11214, 3}
Data:
(0,0,0) 80.48, 35.36, 4.25, 37.45, 3.92, 3.96, 18.53, -9.69, 4.68,
(0,3,0) 55.39, 74.34, 4.6, 22.11, 68.71, 3.85, -4.13, 24.04, 3.73,
(0,6,0) 40.16, 6.39, 4.73, -5.4, 35.73, 4.85, 36.67, 22.45, 4.08, -3.68,
(0,9,1) -10.66, 4.18, 35.95, 36.43, 5.15, 57.17, 3.88, 5.08, -23.64,
(0,12,1) 50.44, 4.32, 6.78, 8.24, 4.36, 21.34, 50.63, 5.21, 16.29,
(0,15,1) -1.34, 5.28, 22.26, 71.25, 5.4, 19.76, 10.38, 5.34, 78.62,
(0,18,1) 11.13, 5.69, 22.14, 59.7, 4.92, 15.65, 47.28, 5.22, 82.41,
(0,21,1) 2.09, 5.24, 16.87, -11.68, 5.35, 15.54, -0.63, 5.2, 81.25,
(...)
The Home Stretch: Adding The Slice
array_2d_t read_frame(int frame_no) {
h5xx::file xaa("xaa.h5", h5xx::file::mode::in);
h5xx::group g(xaa, "particles/lipids/box/positions");
h5xx::dataset ds(g, "value");
array_2d_t a(boost::extents[11214][3]);
std::vector offsets{frame_no, 0, 0}, counts{1, 11214, 3};
h5xx::slice slice(offsets, counts);
h5xx::read_dataset(ds, a, slice);
return a;
}
There you have it. Now we can print any frame:
print_array(read_frame(0));
Printing the same as before. Let's try the last frame:
print_array(read_frame(9));
Prints
79.040001 36.349998 3.990000
37.250000 3.470000 4.140000
18.600000 -9.270000 4.900000
55.669998 75.070000 5.370000
21.920000 67.709999 3.790000
-4.670000 24.770000 3.690000
40.000000 6.060000 5.240000
-5.340000 36.320000 5.410000
36.369999 22.490000 4.130000
-3.520000 -10.430000 4.280000
(...)
Checking again with h5ls -r -d xaa.h5/particles/lipids/box/positions/value |& grep '(9' | head confirms:
(9,0,0) 79.04, 36.35, 3.99, 37.25, 3.47, 4.14, 18.6, -9.27, 4.9, 55.67,
(9,3,1) 75.07, 5.37, 21.92, 67.71, 3.79, -4.67, 24.77, 3.69, 40, 6.06,
(9,6,2) 5.24, -5.34, 36.32, 5.41, 36.37, 22.49, 4.13, -3.52, -10.43,
(9,9,2) 4.28, 35.8, 36.43, 4.99, 56.6, 4.09, 5.04, -23.37, 49.42, 3.81,
(9,13,0) 6.31, 8.83, 4.56, 22.01, 50.38, 5.43, 16.3, -2.92, 5.4, 22.02,
(9,16,1) 70.09, 5.36, 20.23, 11.12, 5.66, 78.48, 11.34, 6.09, 20.26,
(9,19,1) 61.45, 5.35, 14.25, 48.32, 5.35, 79.95, 1.71, 5.38, 17.56,
(9,22,1) -11.61, 5.39, 15.64, -0.19, 5.06, 80.43, 71.77, 5.29, 75.54,
(9,25,1) 35.14, 5.26, 22.45, 56.86, 5.56, 16.47, 52.97, 6.16, 20.62,
(9,28,1) 65.12, 5.26, 19.68, 71.2, 5.52, 23.39, 49.84, 5.28, 22.7,
Full Listing
#include <boost/multi_array.hpp>
#include <h5xx/h5xx.hpp>
#include <iostream>
using array_2d_t = boost::multi_array<float, 2>;
template <typename T> void print_array(T const& array)
{
for (auto const& row : array) { for (auto v : row)
printf("%5f ", v);
printf("\n");
}
}
void dump(h5xx::group const& g, std::string indent = "") {
auto dd = g.datasets();
auto gg = g.groups();
for (auto it = dd.begin(); it != dd.end(); ++it) {
std::cout << indent << " ds:" << it.get_name() << std::endl;
}
for (auto it = gg.begin(); it != gg.end(); ++it) {
dump(*it, indent + "/" + it.get_name());
}
}
array_2d_t read_frame(int frame_no) {
h5xx::file xaa("xaa.h5", h5xx::file::mode::in);
h5xx::group g(xaa, "particles/lipids/box/positions");
h5xx::dataset ds(g, "value");
array_2d_t arr(boost::extents[11214][3]);
std::vector offsets{frame_no, 0, 0}, counts{1, 11214, 3};
h5xx::slice slice(offsets, counts);
h5xx::read_dataset(ds, arr, slice);
return arr;
}
int main()
{
print_array(read_frame(9));
}

CUDA which is faster? Memory coalescing vs caching?

I have encountered this exercise which asks for which code is faster between the following two.
First code.
int sum = 0;
for(int i = 0; i < n; i++) {
sum += array[i*n + thread_id];
}
Second code.
int sum = 0;
for(int i = 0; i < n; i++) {
sum += array[n*thread_id + i];
}
I would try the code myself I will not have a Nvidia GPU in the following days.
I think that the first code takes advantage of memory coalescing see here, while the second one would take advantage of caching.
Many thanks to #RobertCrovella for clarifying the issues regarding memory coalescing. This is my attempt to benchmark the two codes as asked for. It can be clearly noticed from the output (run on a NVS5400M GPU laptop) that the first code is twice more efficient as compared to the second one. This is because of the memory coalescing taking place in the first one (kernel1).
#include <cuda.h>
#include <ctime>
#include <iostream>
#include <stdio.h>
using namespace std;
#define BLOCK_SIZE 1024
#define GRID_SIZE 1024
// Error Handling
#define gpuErrchk(ans) { gpuAssert((ans), __FILE__, __LINE__); }
inline void gpuAssert(cudaError_t code, const char *file, int line, bool abort=true)
{
if (code != cudaSuccess)
{
fprintf(stderr,"GPUassert: %s %s %d\n", cudaGetErrorString(code), file, line);
if (abort) exit(code);
}
}
//kernel1<<<8,8>>>(d_array,d_sum1,n);
__global__ void kernel1(int *array, long *sum, int n) {
long result=0;
int thread_id=threadIdx.x+blockIdx.x*blockDim.x;
for(int i=0;i<n;i++) {
result += array[i*n + thread_id];
}
//__syncthreads();
sum[thread_id]=result;
}
__global__ void kernel2(int *array, long *sum, int n) {
long result=0;
int thread_id=threadIdx.x+blockIdx.x*blockDim.x;
for(int i=0;i<n;i++) {
result += array[n*thread_id+i];
}
__syncthreads();
sum[thread_id]=result;
}
int main() {
srand((unsigned)time(0));
long *h_sum1,*d_sum1;
long *h_sum2,*d_sum2;
int n=10;
int size1=n*BLOCK_SIZE*GRID_SIZE+n;
int *h_array;
h_array=new int[size1];
h_sum1=new long[size1];
h_sum2=new long[size1];
//random number range
int min =1, max =10;
for(int i=0;i<size1;i++) {
h_array[i]= min + (rand() % static_cast<int>(max - min + 1));
h_sum1[i]=0;
h_sum2[i]=0;
}
int *d_array;
gpuErrchk(cudaMalloc((void**)&d_array,size1*sizeof(int)));
gpuErrchk(cudaMalloc((void**)&d_sum1,size1*sizeof(long)));
cudaEvent_t start, stop;
cudaEventCreate(&start);
cudaEventCreate(&stop);
gpuErrchk(cudaMemcpy(d_array,h_array,size1*sizeof(int),cudaMemcpyHostToDevice));
gpuErrchk(cudaMemcpy(d_sum1,h_sum1,size1*sizeof(long),cudaMemcpyHostToDevice));
cudaEventRecord(start);
kernel1<<<GRID_SIZE,BLOCK_SIZE>>>(d_array,d_sum1,n);
cudaEventRecord(stop);
gpuErrchk(cudaMemcpy(h_sum1,d_sum1,size1*sizeof(long),cudaMemcpyDeviceToHost));
float milliSeconds1=0;
cudaEventElapsedTime(&milliSeconds1,start,stop);
gpuErrchk(cudaMalloc((void**)&d_sum2,size1*sizeof(long)));
gpuErrchk(cudaMemcpy(d_sum2,h_sum2,size1*sizeof(long),cudaMemcpyHostToDevice));
cudaEventRecord(start);
kernel2<<<GRID_SIZE,BLOCK_SIZE>>>(d_array,d_sum2,10);
cudaEventRecord(stop);
gpuErrchk(cudaMemcpy(h_sum2,d_sum2,size1*sizeof(long),cudaMemcpyDeviceToHost));
float milliSeconds2=0;
cudaEventElapsedTime(&milliSeconds2,start,stop);
long result_device1=0,result_host1=0;
long result_device2=0,result_host2=0;
for(int i=0;i<size1;i++) {
result_device1 += h_sum1[i];
result_device2 += h_sum2[i];
}
for(int thread_id=0;thread_id<GRID_SIZE*BLOCK_SIZE;thread_id++)
for(int i=0;i<10;i++) {
result_host1 += h_array[i*10+thread_id];
result_host2 += h_array[10*thread_id+i];
}
cout << "Device result1 = " << result_device1 << endl;
cout << "Host result1 = " << result_host1 << endl;
cout << "Time1 (ms) = " << milliSeconds1 << endl;
cout << "Device result2 = " << result_device2 << endl;
cout << "Host result2 = " << result_host2 << endl;
cout << "Time2 (ms) = " << milliSeconds2 << endl;
gpuErrchk(cudaFree(d_array));
gpuErrchk(cudaFree(d_sum1));
gpuErrchk(cudaFree(d_sum2));
return 0;
}
The Cuda Event timer output is as under:
Device result1 = 57659226
Host result1 = 57659226
Time1 (ms) = 5.21952
Device result2 = 57674257
Host result2 = 57674257
Time2 (ms) = 11.8356

Error when debugging sample in OPENCV Camera Calibration using OPEN CV. Assertion failed ( count >=0) cornersubpix. Opencv 3.2 with VS15

I am getting an error when using Camera_Calibration.cpp on OpenCv Tutorial (photo). i running it in release mode.
.
I got assertion fail. I debugged this project by command window with in_VID5.xml
in_VID5.xml:
<?xml version="1.0"?>
<opencv_storage>
<Settings>
<!-- Number of inner corners per a item row and column. (square, circle) -->
<BoardSize_Width>7</BoardSize_Width>
<BoardSize_Height>5</BoardSize_Height>
<!-- The size of a square in some user defined metric system (pixel, millimeter)-->
<Square_Size>30</Square_Size>
<!-- The type of input used for camera calibration. One of: CHESSBOARD CIRCLES_GRID ASYMMETRIC_CIRCLES_GRID -->
<Calibrate_Pattern>"CHESSBOARD"</Calibrate_Pattern>
<!-- The input to use for calibration.
To use an input camera -> give the ID of the camera, like "1"
To use an input video -> give the path of the input video, like "/tmp/x.avi"
To use an image list -> give the path to the XML or YAML file containing the list of the images, like "/tmp/circles_list.xml"
-->
<Input>"1"</Input>
<!-- If true (non-zero) we flip the input images around the horizontal axis.-->
<Input_FlipAroundHorizontalAxis>0</Input_FlipAroundHorizontalAxis>
<!-- Time delay between frames in case of camera. -->
<Input_Delay>100</Input_Delay>
<!-- How many frames to use, for calibration. -->
<Calibrate_NrOfFrameToUse>10</Calibrate_NrOfFrameToUse>
<!-- Consider only fy as a free parameter, the ratio fx/fy stays the same as in the input cameraMatrix.
Use or not setting. 0 - False Non-Zero - True-->
<Calibrate_FixAspectRatio> 1 </Calibrate_FixAspectRatio>
<!-- If true (non-zero) tangential distortion coefficients are set to zeros and stay zero.-->
<Calibrate_AssumeZeroTangentialDistortion>1</Calibrate_AssumeZeroTangentialDistortion>
<!-- If true (non-zero) the principal point is not changed during the global optimization.-->
<Calibrate_FixPrincipalPointAtTheCenter> 1 </Calibrate_FixPrincipalPointAtTheCenter>
<!-- The name of the output log file. -->
<Write_outputFileName>"out_camera_data.xml"</Write_outputFileName>
<!-- If true (non-zero) we write to the output file the feature points.-->
<Write_DetectedFeaturePoints>1</Write_DetectedFeaturePoints>
<!-- If true (non-zero) we write to the output file the extrinsic camera parameters.-->
<Write_extrinsicParameters>1</Write_extrinsicParameters>
<!-- If true (non-zero) we show after calibration the undistorted images.-->
<Show_UndistortedImage>1</Show_UndistortedImage>
</Settings>
</opencv_storage>
My code:
#include
#include <sstream>
#include <time.h>
#include <opencv2/core/core.hpp>
#include <opencv2/imgproc/imgproc.hpp>
#include <opencv2/calib3d/calib3d.hpp>
#include <opencv2/highgui/highgui.hpp>
using namespace cv;
using namespace std;
void help()
{
cout << "This is a camera calibration sample." << endl
<< "Usage: calibration configurationFile" << endl
<< "Near the sample file you'll find the configuration file, which has detailed help of "
"how to edit it. It may be any OpenCV supported file format XML/YAML." << endl;
}
class Settings
{
public:
Settings() : goodInput(false) {}
enum Pattern { NOT_EXISTING, CHESSBOARD, CIRCLES_GRID, ASYMMETRIC_CIRCLES_GRID };
enum InputType { INVALID, CAMERA, VIDEO_FILE, IMAGE_LIST };
void write(FileStorage& fs) const //Write serialization for this class
{
fs << "{" << "BoardSize_Width" << boardSize.width
<< "BoardSize_Height" << boardSize.height
<< "Square_Size" << squareSize
<< "Calibrate_Pattern" << patternToUse
<< "Calibrate_NrOfFrameToUse" << nrFrames
<< "Calibrate_FixAspectRatio" << aspectRatio
<< "Calibrate_AssumeZeroTangentialDistortion" << calibZeroTangentDist
<< "Calibrate_FixPrincipalPointAtTheCenter" << calibFixPrincipalPoint
<< "Write_DetectedFeaturePoints" << bwritePoints
<< "Write_extrinsicParameters" << bwriteExtrinsics
<< "Write_outputFileName" << outputFileName
<< "Show_UndistortedImage" << showUndistorsed
<< "Input_FlipAroundHorizontalAxis" << flipVertical
<< "Input_Delay" << delay
<< "Input" << input
<< "}";
}
void read(const FileNode& node) //Read serialization for this class
{
node["BoardSize_Width"] >> boardSize.width;
node["BoardSize_Height"] >> boardSize.height;
node["Calibrate_Pattern"] >> patternToUse;
node["Square_Size"] >> squareSize;
node["Calibrate_NrOfFrameToUse"] >> nrFrames;
node["Calibrate_FixAspectRatio"] >> aspectRatio;
node["Write_DetectedFeaturePoints"] >> bwritePoints;
node["Write_extrinsicParameters"] >> bwriteExtrinsics;
node["Write_outputFileName"] >> outputFileName;
node["Calibrate_AssumeZeroTangentialDistortion"] >> calibZeroTangentDist;
node["Calibrate_FixPrincipalPointAtTheCenter"] >> calibFixPrincipalPoint;
node["Input_FlipAroundHorizontalAxis"] >> flipVertical;
node["Show_UndistortedImage"] >> showUndistorsed;
node["Input"] >> input;
node["Input_Delay"] >> delay;
interprate();
}
void interprate()
{
goodInput = true;
if (boardSize.width <= 0 || boardSize.height <= 0)
{
cerr << "Invalid Board size: " << boardSize.width << " " << boardSize.height << endl;
goodInput = false;
}
if (squareSize <= 10e-6)
{
cerr << "Invalid square size " << squareSize << endl;
goodInput = false;
}
if (nrFrames <= 0)
{
cerr << "Invalid number of frames " << nrFrames << endl;
goodInput = false;
}
if (input.empty()) // Check for valid input
inputType = INVALID;
else
{
if (input[0] >= '0' && input[0] <= '9')
{
stringstream ss(input);
ss >> cameraID;
inputType = CAMERA;
}
else
{
if (readStringList(input, imageList))
{
inputType = IMAGE_LIST;
nrFrames = (nrFrames < imageList.size()) ? nrFrames : imageList.size();
}
else
inputType = VIDEO_FILE;
}
if (inputType == CAMERA)
inputCapture.open(cameraID);
if (inputType == VIDEO_FILE)
inputCapture.open(input);
if (inputType != IMAGE_LIST && !inputCapture.isOpened())
inputType = INVALID;
}
if (inputType == INVALID)
{
cerr << " Inexistent input: " << input;
goodInput = false;
}
flag = 0;
if (calibFixPrincipalPoint) flag |= CV_CALIB_FIX_PRINCIPAL_POINT;
if (calibZeroTangentDist) flag |= CV_CALIB_ZERO_TANGENT_DIST;
if (aspectRatio) flag |= CV_CALIB_FIX_ASPECT_RATIO;
calibrationPattern = NOT_EXISTING;
if (!patternToUse.compare("CHESSBOARD")) calibrationPattern = CHESSBOARD;
if (!patternToUse.compare("CIRCLES_GRID")) calibrationPattern = CIRCLES_GRID;
if (!patternToUse.compare("ASYMMETRIC_CIRCLES_GRID")) calibrationPattern = ASYMMETRIC_CIRCLES_GRID;
if (calibrationPattern == NOT_EXISTING)
{
cerr << " Inexistent camera calibration mode: " << patternToUse << endl;
goodInput = false;
}
atImageList = 0;
}
Mat nextImage()
{
Mat result;
if (inputCapture.isOpened())
{
Mat view0;
inputCapture >> view0;
view0.copyTo(result);
}
else if (atImageList < (int)imageList.size())
result = imread(imageList[atImageList++], CV_LOAD_IMAGE_COLOR);
return result;
}
static bool readStringList(const string& filename, vector<string>& l)
{
l.clear();
FileStorage fs(filename, FileStorage::READ);
if (!fs.isOpened())
return false;
FileNode n = fs.getFirstTopLevelNode();
if (n.type() != FileNode::SEQ)
return false;
FileNodeIterator it = n.begin(), it_end = n.end();
for (; it != it_end; ++it)
l.push_back((string)*it);
return true;
}
public:
Size boardSize; // The size of the board -> Number of items by width and height
Pattern calibrationPattern;// One of the Chessboard, circles, or asymmetric circle pattern
float squareSize; // The size of a square in your defined unit (point, millimeter,etc).
int nrFrames; // The number of frames to use from the input for calibration
float aspectRatio; // The aspect ratio
int delay; // In case of a video input
bool bwritePoints; // Write detected feature points
bool bwriteExtrinsics; // Write extrinsic parameters
bool calibZeroTangentDist; // Assume zero tangential distortion
bool calibFixPrincipalPoint;// Fix the principal point at the center
bool flipVertical; // Flip the captured images around the horizontal axis
string outputFileName; // The name of the file where to write
bool showUndistorsed; // Show undistorted images after calibration
string input; // The input ->
int cameraID;
vector<string> imageList;
int atImageList;
VideoCapture inputCapture;
InputType inputType;
bool goodInput;
int flag;
private:
string patternToUse;
};
void write(FileStorage& fs, const std::string&, const Settings& x)
{
x.write(fs);
}
void read(const FileNode& node, Settings& x, const Settings& default_value = Settings())
{
if (node.empty())
x = default_value;
else
x.read(node);
}
enum { DETECTION = 0, CAPTURING = 1, CALIBRATED = 2 };
bool runCalibrationAndSave(Settings& s, Size imageSize, Mat& cameraMatrix, Mat& distCoeffs,
vector<vector<Point2f> > imagePoints);
int main(int argc, char* argv[])
{
help();
Settings s;
const string inputSettingsFile = argc > 1 ? argv[1] : "in_VID5.xml";
FileStorage fs(inputSettingsFile, FileStorage::READ); // Read the settings
if (!fs.isOpened())
{
cout << "Could not open the configuration file: \"" << inputSettingsFile << "\"" << endl;
return -1;
}
fs["Settings"] >> s;
fs.release(); // close Settings file
if (!s.goodInput)
{
cout << "Invalid input detected. Application stopping. " << endl;
return -1;
}
vector<vector<Point2f> > imagePoints;
Mat cameraMatrix, distCoeffs;
Size imageSize;
int mode = s.inputType == Settings::IMAGE_LIST ? CAPTURING : DETECTION;
clock_t prevTimestamp = 0;
const Scalar RED(0, 0, 255), GREEN(0, 255, 0);
const char ESC_KEY = 27;
for (int i = 0;; ++i)
{
Mat view;
bool blinkOutput = false;
view = s.nextImage();
//----- If no more image, or got enough, then stop calibration and show result -------------
if (mode == CAPTURING && imagePoints.size() >= (unsigned)s.nrFrames)
{
if (runCalibrationAndSave(s, imageSize, cameraMatrix, distCoeffs, imagePoints))
mode = CALIBRATED;
else
mode = DETECTION;
}
if (view.empty()) // If no more images then run calibration, save and stop loop.
{
if (imagePoints.size() > 0)
runCalibrationAndSave(s, imageSize, cameraMatrix, distCoeffs, imagePoints);
break;
}
imageSize = view.size(); // Format input image.
if (s.flipVertical) flip(view, view, 0);
vector<Point2f> pointBuf;
bool found;
switch (s.calibrationPattern) // Find feature points on the input format
{
case Settings::CHESSBOARD:
found = findChessboardCorners(view, s.boardSize, pointBuf,
CV_CALIB_CB_ADAPTIVE_THRESH | CV_CALIB_CB_FAST_CHECK | CV_CALIB_CB_NORMALIZE_IMAGE);
break;
case Settings::CIRCLES_GRID:
found = findCirclesGrid(view, s.boardSize, pointBuf);
break;
case Settings::ASYMMETRIC_CIRCLES_GRID:
found = findCirclesGrid(view, s.boardSize, pointBuf, CALIB_CB_ASYMMETRIC_GRID);
break;
}
if (found) // If done with success,
{
// improve the found corners' coordinate accuracy for chessboard
if (s.calibrationPattern == Settings::CHESSBOARD)
{
Mat viewGray;
cvtColor(view, viewGray, CV_BGR2GRAY);
cornerSubPix(viewGray, pointBuf, Size(11, 11),
Size(-1, -1), TermCriteria(CV_TERMCRIT_EPS + CV_TERMCRIT_ITER, 30, 0.1));
}
if (mode == CAPTURING && // For camera only take new samples after delay time
(!s.inputCapture.isOpened() || clock() - prevTimestamp > s.delay*1e-3*CLOCKS_PER_SEC))
{
imagePoints.push_back(pointBuf);
prevTimestamp = clock();
blinkOutput = s.inputCapture.isOpened();
}
// Draw the corners.
drawChessboardCorners(view, s.boardSize, Mat(pointBuf), found);
}
//----------------------------- Output Text ------------------------------------------------
string msg = (mode == CAPTURING) ? "100/100" :
mode == CALIBRATED ? "Calibrated" : "Press 'g' to start";
int baseLine = 0;
Size textSize = getTextSize(msg, 1, 1, 1, &baseLine);
Point textOrigin(view.cols - 2 * textSize.width - 10, view.rows - 2 * baseLine - 10);
if (mode == CAPTURING)
{
if (s.showUndistorsed)
msg = format("%d/%d Undist", (int)imagePoints.size(), s.nrFrames);
else
msg = format("%d/%d", (int)imagePoints.size(), s.nrFrames);
}
putText(view, msg, textOrigin, 1, 1, mode == CALIBRATED ? GREEN : RED);
if (blinkOutput)
bitwise_not(view, view);
//------------------------- Video capture output undistorted ------------------------------
if (mode == CALIBRATED && s.showUndistorsed)
{
Mat temp = view.clone();
undistort(temp, view, cameraMatrix, distCoeffs);
}
//------------------------------ Show image and check for input commands -------------------
imshow("Image View", view);
char key = waitKey(s.inputCapture.isOpened() ? 50 : s.delay);
if (key == ESC_KEY)
break;
if (key == 'u' && mode == CALIBRATED)
s.showUndistorsed = !s.showUndistorsed;
if (s.inputCapture.isOpened() && key == 'g')
{
mode = CAPTURING;
imagePoints.clear();
}
}
// -----------------------Show the undistorted image for the image list ------------------------
if (s.inputType == Settings::IMAGE_LIST && s.showUndistorsed)
{
Mat view, rview, map1, map2;
initUndistortRectifyMap(cameraMatrix, distCoeffs, Mat(),
getOptimalNewCameraMatrix(cameraMatrix, distCoeffs, imageSize, 1, imageSize, 0),
imageSize, CV_16SC2, map1, map2);
for (int i = 0; i < (int)s.imageList.size(); i++)
{
view = imread(s.imageList[i], 1);
if (view.empty())
continue;
remap(view, rview, map1, map2, INTER_LINEAR);
imshow("Image View", rview);
char c = waitKey();
if (c == ESC_KEY || c == 'q' || c == 'Q')
break;
}
}
return 0;
}
double computeReprojectionErrors(const vector<vector<Point3f> >& objectPoints,
const vector<vector<Point2f> >& imagePoints,
const vector<Mat>& rvecs, const vector<Mat>& tvecs,
const Mat& cameraMatrix, const Mat& distCoeffs,
vector<float>& perViewErrors)
{
vector<Point2f> imagePoints2;
int i, totalPoints = 0;
double totalErr = 0, err;
perViewErrors.resize(objectPoints.size());
for (i = 0; i < (int)objectPoints.size(); ++i)
{
projectPoints(Mat(objectPoints[i]), rvecs[i], tvecs[i], cameraMatrix,
distCoeffs, imagePoints2);
err = norm(Mat(imagePoints[i]), Mat(imagePoints2), CV_L2);
int n = (int)objectPoints[i].size();
perViewErrors[i] = (float)std::sqrt(err*err / n);
totalErr += err*err;
totalPoints += n;
}
return std::sqrt(totalErr / totalPoints);
}
void calcBoardCornerPositions(Size boardSize, float squareSize, vector<Point3f>& corners,
Settings::Pattern patternType /*= Settings::CHESSBOARD*/)
{
corners.clear();
switch (patternType)
{
case Settings::CHESSBOARD:
case Settings::CIRCLES_GRID:
for (int i = 0; i < boardSize.height; ++i)
for (int j = 0; j < boardSize.width; ++j)
corners.push_back(Point3f(float(j*squareSize), float(i*squareSize), 0));
break;
case Settings::ASYMMETRIC_CIRCLES_GRID:
for (int i = 0; i < boardSize.height; i++)
for (int j = 0; j < boardSize.width; j++)
corners.push_back(Point3f(float((2 * j + i % 2)*squareSize), float(i*squareSize), 0));
break;
}
}
bool runCalibration(Settings& s, Size& imageSize, Mat& cameraMatrix, Mat& distCoeffs,
vector<vector<Point2f> > imagePoints, vector<Mat>& rvecs, vector<Mat>& tvecs,
vector<float>& reprojErrs, double& totalAvgErr)
{
cameraMatrix = Mat::eye(3, 3, CV_64F);
if (s.flag & CV_CALIB_FIX_ASPECT_RATIO)
cameraMatrix.at<double>(0, 0) = 1.0;
distCoeffs = Mat::zeros(8, 1, CV_64F);
vector<vector<Point3f> > objectPoints(1);
calcBoardCornerPositions(s.boardSize, s.squareSize, objectPoints[0], s.calibrationPattern);
objectPoints.resize(imagePoints.size(), objectPoints[0]);
//Find intrinsic and extrinsic camera parameters
double rms = calibrateCamera(objectPoints, imagePoints, imageSize, cameraMatrix,
distCoeffs, rvecs, tvecs, s.flag | CV_CALIB_FIX_K4 | CV_CALIB_FIX_K5);
cout << "Re-projection error reported by calibrateCamera: " << rms << endl;
bool ok = checkRange(cameraMatrix) && checkRange(distCoeffs);
totalAvgErr = computeReprojectionErrors(objectPoints, imagePoints,
rvecs, tvecs, cameraMatrix, distCoeffs, reprojErrs);
return ok;
}
// Print camera parameters to the output file
void saveCameraParams(Settings& s, Size& imageSize, Mat& cameraMatrix, Mat& distCoeffs,
const vector<Mat>& rvecs, const vector<Mat>& tvecs,
const vector<float>& reprojErrs, const vector<vector<Point2f> >& imagePoints,
double totalAvgErr)
{
FileStorage fs(s.outputFileName, FileStorage::WRITE);
time_t t;
time(&t);
struct tm *t2 = localtime(&t);
char buf[1024];
strftime(buf, sizeof(buf) - 1, "%c", t2);
fs << "calibration_Time" << buf;
if (!rvecs.empty() || !reprojErrs.empty())
fs << "nrOfFrames" << (int)std::max(rvecs.size(), reprojErrs.size());
fs << "image_Width" << imageSize.width;
fs << "image_Height" << imageSize.height;
fs << "board_Width" << s.boardSize.width;
fs << "board_Height" << s.boardSize.height;
fs << "square_Size" << s.squareSize;
if (s.flag & CV_CALIB_FIX_ASPECT_RATIO)
fs << "FixAspectRatio" << s.aspectRatio;
if (s.flag)
{
sprintf(buf, "flags: %s%s%s%s",
s.flag & CV_CALIB_USE_INTRINSIC_GUESS ? " +use_intrinsic_guess" : "",
s.flag & CV_CALIB_FIX_ASPECT_RATIO ? " +fix_aspectRatio" : "",
s.flag & CV_CALIB_FIX_PRINCIPAL_POINT ? " +fix_principal_point" : "",
s.flag & CV_CALIB_ZERO_TANGENT_DIST ? " +zero_tangent_dist" : "");
cvWriteComment(*fs, buf, 0);
}
fs << "flagValue" << s.flag;
fs << "Camera_Matrix" << cameraMatrix;
fs << "Distortion_Coefficients" << distCoeffs;
fs << "Avg_Reprojection_Error" << totalAvgErr;
if (!reprojErrs.empty())
fs << "Per_View_Reprojection_Errors" << Mat(reprojErrs);
if (!rvecs.empty() && !tvecs.empty())
{
CV_Assert(rvecs[0].type() == tvecs[0].type());
Mat bigmat((int)rvecs.size(), 6, rvecs[0].type());
for (int i = 0; i < (int)rvecs.size(); i++)
{
Mat r = bigmat(Range(i, i + 1), Range(0, 3));
Mat t = bigmat(Range(i, i + 1), Range(3, 6));
CV_Assert(rvecs[i].rows == 3 && rvecs[i].cols == 1);
CV_Assert(tvecs[i].rows == 3 && tvecs[i].cols == 1);
//*.t() is MatExpr (not Mat) so we can use assignment operator
r = rvecs[i].t();
t = tvecs[i].t();
}
cvWriteComment(*fs, "a set of 6-tuples (rotation vector + translation vector) for each view", 0);
fs << "Extrinsic_Parameters" << bigmat;
}
if (!imagePoints.empty())
{
Mat imagePtMat((int)imagePoints.size(), imagePoints[0].size(), CV_32FC2);
for (int i = 0; i < (int)imagePoints.size(); i++)
{
Mat r = imagePtMat.row(i).reshape(2, imagePtMat.cols);
Mat imgpti(imagePoints[i]);
imgpti.copyTo(r);
}
fs << "Image_points" << imagePtMat;
}
}
bool runCalibrationAndSave(Settings& s, Size imageSize, Mat& cameraMatrix, Mat& distCoeffs, vector<vector<Point2f> > imagePoints)
{
vector<Mat> rvecs, tvecs;
vector<float> reprojErrs;
double totalAvgErr = 0;
bool ok = runCalibration(s, imageSize, cameraMatrix, distCoeffs, imagePoints, rvecs, tvecs,
reprojErrs, totalAvgErr);
cout << (ok ? "Calibration succeeded" : "Calibration failed")
<< ". avg re projection error = " << totalAvgErr;
if (ok)
saveCameraParams(s, imageSize, cameraMatrix, distCoeffs, rvecs, tvecs, reprojErrs,
imagePoints, totalAvgErr);
return ok;
}

cublas cublasDgetrfBatched() batched LU factorization doesn't work with matrices bigger than 32x32

I wrote a cuda function for Matlab to perform a LU factorization of a batch of matrices using cublasDgetrfBatched(). The toolkit documentation of this function is here.
It works fine for matrices up to size 32x32. But it fails with status code CUBLAS_STATUS_INVALID_VALUE for bigger matrices. Below is my source code (gpuBatchedLU.cu):
#include "mex.h"
#include "gpu/mxGPUArray.h"
/* Includes, cuda */
#include <cuda_runtime.h>
#include <cublas_v2.h>
#include <string>
#include <sstream>
static std::string cublasGetErrorString(cublasStatus_t error) {
switch (error) {
case CUBLAS_STATUS_SUCCESS:
return "CUBLAS_STATUS_SUCCESS";
case CUBLAS_STATUS_NOT_INITIALIZED:
return "CUBLAS_STATUS_NOT_INITIALIZED";
case CUBLAS_STATUS_ALLOC_FAILED:
return "CUBLAS_STATUS_ALLOC_FAILED";
case CUBLAS_STATUS_INVALID_VALUE:
return "CUBLAS_STATUS_INVALID_VALUE";
case CUBLAS_STATUS_ARCH_MISMATCH:
return "CUBLAS_STATUS_ARCH_MISMATCH";
case CUBLAS_STATUS_MAPPING_ERROR:
return "CUBLAS_STATUS_MAPPING_ERROR";
case CUBLAS_STATUS_EXECUTION_FAILED:
return "CUBLAS_STATUS_EXECUTION_FAILED";
case CUBLAS_STATUS_INTERNAL_ERROR:
return "CUBLAS_STATUS_INTERNAL_ERROR";
}
return "<unknown>";
}
inline bool cublasAssert(cublasStatus_t code, const char* file, int line) {
if (code != CUBLAS_STATUS_SUCCESS) {
std::stringstream ss;
ss << "cublasAssert: " << cublasGetErrorString(code) << " in "
<< std::string(file) << ", line " << line << ".";
mexErrMsgTxt(ss.str().c_str());
}
return code == CUBLAS_STATUS_SUCCESS;
}
inline bool cudaAssert(cudaError_t code, const char* file, int line) {
if (code != cudaSuccess) {
std::stringstream ss;
ss << "cudaAssert: " << cudaGetErrorString(code) << " in "
<< std::string(file) << ", line " << line << ".";
mexErrMsgTxt(ss.str().c_str());
}
return code == cudaSuccess;
}
inline bool mexGPUAssert(int code, const char* file, int line) {
if (code != MX_GPU_SUCCESS) {
std::stringstream ss;
ss << "mexGPUAssert: could not initialize the Mathworks GPU API in "
<< std::string(file) << ", line " << line << ".";
mexErrMsgTxt(ss.str().c_str());
}
return code == MX_GPU_SUCCESS;
}
#define cublasErrchk(ans) { cublasAssert((ans), __FILE__, __LINE__); }
#define cudaErrchk(ans) { cudaAssert((ans), __FILE__, __LINE__); }
#define mxGPUErrchk(ans) { mexGPUAssert((ans), __FILE__, __LINE__); }
void mexFunction(int nlhs, mxArray *plhs[], /* Output variables */int nrhs,
const mxArray *prhs[]) /* Input variables */{
if (nrhs != 1) { /* end if not one function arguments */
mexErrMsgTxt("This function requires one input argument.");
return;
}
if (nlhs > 3) { /* take three outputs */
mexErrMsgTxt("This function takes a maximum of three output variables.");
return;
}
mxGPUErrchk(mxInitGPU());
const mxGPUArray* in1_gpu = mxGPUCreateFromMxArray(prhs[0]);
size_t ndims = mxGPUGetNumberOfDimensions(in1_gpu);
const size_t* dim = (const size_t*) mxGPUGetDimensions(in1_gpu);
if (ndims != 3) { /* end if input arguments are of different dimensions */
mexErrMsgTxt("The input argument must be a 3-dimensional array.");
return;
}
cublasHandle_t handle;
cublasErrchk(cublasCreate(&handle));
int no_matrices = dim[2];
int nrow = dim[0];
int ncol = dim[1];
int matrix_size = nrow * ncol;
size_t i;
std::stringstream ss;
ss << "dim[2] = " << dim[2] << "\nno_matrices = " << no_matrices << "\nnrow = " << nrow << "\nmatrix_size = " << nrow << " x " << ncol << " = " << matrix_size << std::endl;
mexPrintf(ss.str().c_str());
mxGPUArray* gpu_array_inout = mxGPUCopyFromMxArray(prhs[0]);
double* inout_storage = (double*) mxGPUGetData(gpu_array_inout);
size_t info_dimensions[1] = { no_matrices };
mxGPUArray* gpu_array_info = mxGPUCreateGPUArray(1, (mwSize*) info_dimensions, mxINT32_CLASS, mxREAL,
MX_GPU_INITIALIZE_VALUES);
int* out_info = (int*) mxGPUGetData(gpu_array_info);
mexPrintf("after defining gpu_array_info\n");
size_t pivot_dimensions[2] = { nrow, no_matrices };
mxGPUArray* gpu_array_pivot = mxGPUCreateGPUArray(2, (mwSize*) pivot_dimensions, mxINT32_CLASS, mxREAL,
MX_GPU_DO_NOT_INITIALIZE);
int* out_pivot = (int*) mxGPUGetData(gpu_array_pivot);
mexPrintf("after defining gpu_array_pivot\n");
double** inout_pointers_CPU = (double**) malloc(no_matrices * sizeof(double*));
for (i = 0; i < no_matrices; i++) {
inout_pointers_CPU[i] = (double*) ((char*) inout_storage + i * ((size_t) matrix_size) * sizeof(double));
}
double** inout_pointers_GPU;
cudaErrchk(cudaMalloc((void** )&inout_pointers_GPU, no_matrices * sizeof(double*)));
cudaErrchk(
cudaMemcpy(inout_pointers_GPU, inout_pointers_CPU, no_matrices * sizeof(double*), cudaMemcpyHostToDevice));
free(inout_pointers_CPU);
ss.clear();
ss << "check again before calling cublasDgetrfBatched:\nnrow = " << nrow << "\nno_matrices = " << no_matrices << std::endl;
mexPrintf(ss.str().c_str());
cublasErrchk(cublasDgetrfBatched(handle, nrow, inout_pointers_GPU, nrow, out_pivot, out_info, no_matrices));
cublasErrchk(cublasDestroy(handle));
cudaErrchk(cudaFree(inout_pointers_GPU));
if (mxIsGPUArray(prhs[0])) {
plhs[0] = mxGPUCreateMxArrayOnGPU(gpu_array_inout);
if (nlhs > 1) {
plhs[1] = mxGPUCreateMxArrayOnGPU(gpu_array_pivot);
if (nlhs > 2) {
plhs[2] = mxGPUCreateMxArrayOnGPU(gpu_array_info);
}
}
} else {
plhs[0] = mxGPUCreateMxArrayOnCPU(gpu_array_inout);
if (nlhs > 1) {
plhs[1] = mxGPUCreateMxArrayOnCPU(gpu_array_pivot);
if (nlhs > 2) {
plhs[2] = mxGPUCreateMxArrayOnCPU(gpu_array_info);
}
}
}
mxGPUDestroyGPUArray(gpu_array_inout);
mxGPUDestroyGPUArray(gpu_array_pivot);
mxGPUDestroyGPUArray(gpu_array_info);
mxFree((void*) dim);
return;
}
I compile as follows:
mex -L/usr/local/cuda/lib64 -lcudart -lcublas gpuBatchedLU.cu
And I call from MATLAB:
[a1,b1,c1]=gpuBatchedLU(randn(32,32,5)); %no problem
[a2,b2,c2]=gpuBatchedLU(randn(33,33,5)); %produces CUBLAS_STATUS_INVALID_VALUE
I use Matlab R2013b with the parallel toolbox, Cuda 5.5, and a NVS 5200M graphics chip.
Can anyone replicate this problem? I would appreciate any suggestions on how to solve this problem.
The problem seems to be with Matlab R2013b using libcublas.so in version 5.0. The file link is in /MATLAB/R2013b/bin/glnxa64/. Once I changed the link to the libcublas.so of my Cuda 5.5 installation it worked fine.

Debug Assertion Failed - MSVCP110D.dll

#include "stdafx.h"
#include <iostream>
#include <string>
using namespace std;
int main()
{
string token = "000000:::AAAA:::000011:::Hello 8:::::::D Jay!";
string * stringArray = new string[token.size()];
string interim;
int r = 0;
int arrayCounter = 0;
for(int x = 0; x < token.length(); x++)
{
if(token[x] != ':')
{
interim[r] = token[x];
r++;
}
}
for (int x = 0; x < r; x++)
{
cout << interim[x] << endl;
}
system("pause");
return 0;
}
I am new and learning, and have narrowed it down to the line:
interim[r] = token[x];
..But i don't know why it crashes. Advice? I am coding in Visual C++ VSE2012
The string interim has a size of zero. Setting interim[r] = token[x] modifies the string at location r without changing its size. With a size of zero this is undefined behavior.
interim += token[x] is probably what you want.
Example:
#include <iostream>
#include <string>
using namespace std;
int main()
{
string token = "000000:::AAAA:::000011:::Hello 8:::::::D Jay!";
string interim;
for(int x = 0; x < token.length(); x++)
{
if(token[x] != ':')
{
interim += token[x];
}
}
cout << interim << endl;
system("pause");
return 0;
}
Output:
000000AAAA000011Hello 8D Jay!
Press any key to continue . . .

Resources