This is the sequential piece of code I am trying to parallelize in CUDA
/*
Sequential (Single Thread) APSP on CPU.
*/
void floyd_sequential(int *mat, const size_t N)
{
for(int k = 0; k < N; k ++)
for(int i = 0; i < N; i ++)
for(int j = 0; j < N; j ++)
{
int i0 = i*N + j;
int i1 = i*N + k;
int i2 = k*N + j;
if(mat[i1] != -1 && mat[i2] != -1)
mat[i0] = (mat[i0] != -1 && mat[i0] < mat[i1] + mat[i2]) ?
mat[i0] : (mat[i1] + mat[i2]);
}
}
This is my CUDA implementation
// ParallelComputing.cpp : Defines the entry point for the console application.
//
#include <stdio.h>
#include <cuda.h>
#include <stdlib.h>
#define DIMENSION 10;
__global__ void gpu_Floyd(int *result, int N)
{
int j,k;
int Row = blockIdx.y * blockDim.y + threadIdx.y;
for(k = 0; k < N; k++)
{
for(j = 0; j < N; j++)
{
int i0 = Row * N + j;
int i1 = Row * N + k;
int i2 = k * N + j;
if(result[i0] != -1 && result[i2] != -1)
result[i0] = (result[i0] != -1 && result[i0] < result[i1] + result[i2]) ?
result[i0] : (result[i1] + result[i2]);
__syncthreads();
}
}
}
void GenMatrix(int *mat, const size_t N)
{
for(int i = 0; i < N*N; i ++)
mat[i] = rand()%32 - 1;
}
bool CmpArray(const int *l, const int *r, const size_t eleNum)
{
for(int i = 0; i < eleNum; i ++)
if(l[i] != r[i])
{
printf("ERROR: l[%d] = %d, r[%d] = %d\n", i, l[i], i, r[i]);
return false;
}
return true;
}
int main(int argc, char **argv)
{
// generate a random matrix.
size_t N = 10;
int *mat = (int*)malloc(sizeof(int)*N*N);
GenMatrix(mat, N);
// compute the reference result.
int *ref = (int*)malloc(sizeof(int)*N*N);
memcpy(ref, mat, sizeof(int)*N*N);
Floyd_sequential(ref, N);
//CUDA Portion
int Grid_Dim_x = 1, Grid_Dim_y = 1;
int noThreads_x, noThreads_y;
int *result = (int*)malloc(sizeof(int)*N*N);
memcpy(result, mat, sizeof(int)*N*N);
int *d_result;
// compute your results
cudaMalloc((void **)&d_result, N*N);
cudaMemcpy(result, N * N, cudaMemcpyHostToDevice);
gpu_Floyd<<<1024, 256>>>(d_result, N);
cudaMemcpy(result, d_result, cudaMemcpyDeviceToHost);
// compare your result with reference result
if(CmpArray(result, ref, N*N))
printf("The matrix matches.\n");
else
printf("The matrix do not match.\n");
free(ref);
free(result);
cudaFree(d_result);
}
However, my output always shows the matrices do not match.
I understand that in CUDA we try to map each element in the matrix to each row. However, I am trying to explore possibilities by mapping each row of the matrix to a thread instead.
As has already been mentioned, your provided GPU code does not compile, so I'm curious how you got to the observation that your output matrices do not match.
Here are some of the problems with your code:
cudaMalloc, just like malloc allocates bytes, so this is not correct:
cudaMalloc((void **)&d_result, N*N);
instead you want this:
cudaMalloc((void **)&d_result, N*N*sizeof(int));
likewise cudaMemcpy, just like memcpy, operates on bytes, and furthermore cudaMemcpy requires 4 parameters so this is not correct:
cudaMemcpy(result, N * N, cudaMemcpyHostToDevice);
instead you probably want this:
cudaMemcpy(d_result, result, N * N*sizeof(int), cudaMemcpyHostToDevice);
and your other cudaMemcpy line needs to be fixed similarly.
I'd also advise doing proper cuda error checking
Your kernel is written as if it's expecting a 2 dimensional thread array, or at least one dimensional in y, whereas you are launching a one dimensional grid in x:
gpu_Floyd<<<1024, 256>>>(d_result, N);
therefore all your kernel built-in variables in y will be 1 or 0 always, and this line of code:
int Row = blockIdx.y * blockDim.y + threadIdx.y;
will evaluate to zero for all threads in your 1-D grid in x.
Your gpu kernel is putting the results in the same matrix as the input data. For sequential code this may or may not matter, but for code that is intended to run in parallel, it can often lead to race conditions, because the order of operations (i.e. order of thread execution) is largely undefined.
Below you will find a canonical, simple implementation of the Floyd-Warshall algorithm in CUDA.
The CUDA code is accompanied with a sequential implementation and both are based on the simplifying assumption that the edges are non-negative. The full, minimum distance paths are also reconstructed in both the cases. Despite the simplifying assumption, it should be possible to grasp the relevant parallelization idea, namely that a two-dimensional thread grid is exploited and that each thread along x is assigned to a matrix column, while each block along y is assigned to a matrix row. In this way, all the columns are loaded by the threadIdx.x == 0 threads of each block in shared memory.
// --- Assumption: graph with positive edges
#include <stdio.h>
#include <string>
#include <map>
#include <iostream>
#include <fstream>
#include "Utilities.cuh"
#define BLOCKSIZE 256
using namespace std;
map<string, int> nameToNum; // --- names of vertices
map<string, map<string, int>> weightMap; // --- weights of edges
/************************/
/* READ GRAPH FROM FILE */
/************************/
int *readGraphFromFile(int &N, char *fileName) {
string vertex1, vertex2;
ifstream graphFile;
int currentWeight;
N = 0; // --- Init the number of found vertices
graphFile.open(fileName); // --- Open the graph file
graphFile >> vertex1; // --- Read first vertex
while(vertex1 != "--END--") { // --- Loop untile end of file has not been found
graphFile >> vertex2; // --- Read second vertex
graphFile >> currentWeight; // --- Read weight between first and second vertex
if (nameToNum.count(vertex1) == 0) { // --- If vertex has not yet been added ...
nameToNum[vertex1] = N; // assign a progressive number to the vertex
weightMap[vertex1][vertex1] = 0; // assign a zero weight to the "self-edge"
N++; // --- Update the found number of vertices
}
if (nameToNum.count(vertex2) == 0) {
nameToNum[vertex2] = N;
weightMap[vertex2][vertex2] = 0;
N++;
}
weightMap[vertex1][vertex2] = currentWeight; // --- Update weight between vertices 1 and 2
graphFile >> vertex1;
}
graphFile.close(); // --- Close the graph file
// --- Construct the array
int *weightMatrix = (int*) malloc(N * N * sizeof(int));
// --- Loop over all the vertex couples in the wights matrix
for (int ii = 0; ii < N; ii++)
for (int jj = 0; jj < N; jj++)
weightMatrix[ii * N + jj] = INT_MAX / 2; // --- Init the weights matrix elements to infinity
map<string, int>::iterator i, j;
// --- Loop over all the vertex couples in the map
// (*i).first and (*j).first are the weight entries of the map, while (*i).second and (*j).second are their corresponding indices
for (i = nameToNum.begin(); i != nameToNum.end(); ++i)
for (j = nameToNum.begin(); j != nameToNum.end(); ++j) {
// --- If there is connection between vertices (*i).first and (*j).first, the update the weight matrix
if (weightMap[(*i).first].count((*j).first) != 0)
weightMatrix[N * (*i).second + (*j).second] = weightMap[(*i).first][(*j).first];
}
return weightMatrix;
}
/************************************/
/* PRINT MINIMUM DISTANCES FUNCTION */
/************************************/
void printMinimumDistances(int N, int *a) {
map<string, int>::iterator i;
// --- Prints all the node labels at the first row
for (i = nameToNum.begin(); i != nameToNum.end(); ++i) printf("\t%s", i->first.c_str());
printf("\n");
i = nameToNum.begin();
// --- Loop over the rows
for (int p = 0; p < N; p++) {
printf("%s\t", i -> first.c_str());
// --- Loop over the columns
for (int q = 0; q < N; q++) {
int dd = a[p * N + q];
if (dd != INT_MAX / 2) printf("%d\t", dd);
else printf("--\t");
}
printf("\n");
i++;
}
}
void printPathRecursive(int row, int col, int *minimumDistances, int *path, int N) {
map<string, int>::iterator i = nameToNum.begin();
map<string, int>::iterator j = nameToNum.begin();
if (row == col) {advance(i, row); printf("%s\t", i -> first.c_str()); }
else {
if (path[row * N + col] == INT_MAX / 2) printf("%row %row %row No path exists\t\n", minimumDistances[row * N + col], row, col);
else {
printPathRecursive(row, path[row * N + col], minimumDistances, path, N);
advance(j, col);
printf("%s\t", j -> first.c_str());
}
}
}
void printPath(int N, int *minimumDistances, int *path) {
map<string, int>::iterator i;
map<string, int>::iterator j;
// --- Loop over the rows
i = nameToNum.begin();
for (int p = 0; p < N; p++) {
// --- Loop over the columns
j = nameToNum.begin();
for (int q = 0; q < N; q++) {
printf("From %s to %s\t", i -> first.c_str(), j -> first.c_str());
printPathRecursive(p, q, minimumDistances, path, N);
printf("\n");
j++;
}
i++;
}
}
/**********************/
/* FLOYD-WARSHALL CPU */
/**********************/
void h_FloydWarshall(int *h_graphMinimumDistances, int *h_graphPath, const int N) {
for (int k = 0; k < N; k++)
for (int row = 0; row < N; row++)
for (int col = 0; col < N; col++) {
if (h_graphMinimumDistances[row * N + col] > (h_graphMinimumDistances[row * N + k] + h_graphMinimumDistances[k * N + col])) {
h_graphMinimumDistances[row * N + col] = (h_graphMinimumDistances[row * N + k] + h_graphMinimumDistances[k * N + col]);
h_graphPath[row * N + col] = h_graphPath[k * N + col];
}
}
}
/*************************/
/* FLOYD-WARSHALL KERNEL */
/*************************/
__global__ void d_FloydWarshall(int k, int *d_graphMinimumDistances, int *d_graphPath, int N) {
int col = blockIdx.x * blockDim.x + threadIdx.x; // --- Each thread along x is assigned to a matrix column
int row = blockIdx.y; // --- Each block along y is assigned to a matrix row
if (col >= N) return;
int arrayIndex = N * row + col;
// --- All the blocks load the entire k-th column into shared memory
__shared__ int d_graphMinimumDistances_row_k;
if(threadIdx.x == 0) d_graphMinimumDistances_row_k = d_graphMinimumDistances[N * row + k];
__syncthreads();
if (d_graphMinimumDistances_row_k == INT_MAX / 2) // --- If element (row, k) = infinity, no update is needed
return;
int d_graphMinimumDistances_k_col = d_graphMinimumDistances[k * N + col];
if(d_graphMinimumDistances_k_col == INT_MAX / 2) // --- If element (k, col) = infinity, no update is needed
return;
int candidateBetterDistance = d_graphMinimumDistances_row_k + d_graphMinimumDistances_k_col;
if (candidateBetterDistance < d_graphMinimumDistances[arrayIndex]) {
d_graphMinimumDistances[arrayIndex] = candidateBetterDistance;
d_graphPath[arrayIndex] = d_graphPath[k * N + col];
}
}
/********/
/* MAIN */
/********/
int main() {
int N = 0; // --- Number of vertices
// --- Read graph array from file
int *h_graphArray = readGraphFromFile(N, "graph2.txt");
printf("\n******************\n");
printf("* Original graph *\n");
printf("******************\n");
printMinimumDistances(N, h_graphArray);
// --- Floyd-Warshall on CPU
int *h_graphMinimumDistances = (int *) malloc(N * N * sizeof(int));
int *h_graphPath = (int *) malloc(N * N * sizeof(int));
memcpy(h_graphMinimumDistances, h_graphArray, N * N * sizeof(int));
for (int k = 0; k < N; k++)
for (int l = 0; l < N; l++)
if (h_graphArray[k * N + l] == INT_MAX / 2) h_graphPath[k * N + l] = INT_MAX / 2;
else h_graphPath[k * N + l] = k;
h_FloydWarshall(h_graphMinimumDistances, h_graphPath, N);
printf("\n*************************\n");
printf("* CPU result: distances *\n");
printf("*************************\n");
printMinimumDistances(N, h_graphMinimumDistances);
printf("\n********************\n");
printf("* CPU result: path *\n");
printf("********************\n");
printPath(N, h_graphMinimumDistances, h_graphPath);
// --- Graph array device allocation and host-device memory transfer
int *d_graphMinimumDistances; gpuErrchk(cudaMalloc(&d_graphMinimumDistances, N * N * sizeof(int)));
gpuErrchk(cudaMemcpy(d_graphMinimumDistances, h_graphArray, N * N * sizeof(int), cudaMemcpyHostToDevice));
int *d_graphPath; gpuErrchk(cudaMalloc(&d_graphPath, N * N * sizeof(int)));
for (int k = 0; k < N; k++)
for (int l = 0; l < N; l++)
if (h_graphArray[k * N + l] == INT_MAX / 2) h_graphPath[k * N + l] = INT_MAX / 2;
else h_graphPath[k * N + l] = k;
gpuErrchk(cudaMemcpy(d_graphPath, h_graphPath, N * N * sizeof(int), cudaMemcpyHostToDevice));
// --- Iterations
for (int k = 0; k < N; k++) {
d_FloydWarshall <<<dim3(iDivUp(N, BLOCKSIZE), N), BLOCKSIZE>>>(k, d_graphMinimumDistances, d_graphPath, N);
#ifdef DEBUG
gpuErrchk(cudaPeekAtLastError());
gpuErrchk(cudaDeviceSynchronize());
#endif
}
// --- Copy results back to the host
gpuErrchk(cudaMemcpy(h_graphMinimumDistances, d_graphMinimumDistances, N * N * sizeof(int), cudaMemcpyDeviceToHost));
gpuErrchk(cudaMemcpy(h_graphPath, d_graphPath, N * N * sizeof(int), cudaMemcpyDeviceToHost));
printf("\n**************\n");
printf("* GPU result *\n");
printf("**************\n");
printMinimumDistances(N, h_graphMinimumDistances);
printf("\n********************\n");
printf("* GPU result: path *\n");
printf("********************\n");
printPath(N, h_graphMinimumDistances, h_graphPath);
}
Related
I kinda understand how to deal with 2D cuda. But batched cholesky has a 4D towards the end of the algorithm. I attached cholesky and my cuda code if anyone could give me a hint.
int i, k, m, n;
// Batched Cholesky factorization.
for (i = 0; i < batch; i++) {
float *pA = &dA[i*N*N];
// Single Cholesky factorization.
for (k = 0; k < N; k++) {
// Panel factorization.
pA[k*N+k] = sqrtf(pA[k*N+k]);
for (m = k+1; m < N; m++)
pA[k*N+m] /= pA[k*N+k];
// Update of the trailing submatrix.
for (n = k+1; n < N; n++)
for (m = n; m < N; m++)
pA[n*N+m] -= (pA[k*N+n]*pA[k*N+m]);
}
}
Cuda:
int i = blockIdx.x * blockDim.x + threadIdx.x;
int k = blockIdx.y * blockDim.y + threadIdx.y;
int m = blockIdx.z * blockDim.z + threadIdx.z;
int n = blockIdx.z * blockDim.z + threadIdx.z;
if( k >= N || m >= N || n >= N || i >= batch ) return;
float *pA = &dA[i*N*N];
pA[k*N+k] = sqrtf(pA[k*N+k]);
pA[k*N+m] /= pA[k*N+k];
pA[n*N+m] -= (pA[k*N+n]*pA[k*N+m]);
starter:
dim3 dimBlock( (batch+31)/32, (n+31)/32, (n+31)/32 );
dim3 dimGrid( 32, 32, 32);
spotrf_batched_kernel<<< dimBlock, dimGrid, 0, stream>>>(n, batch, dA);
I am going to leave this here without much comment. The code is relatively self-explanatory. This implementation is completely faithful to your serial version, with the following features:
Each block performs exactly one factorization in the batch. Run as many blocks as there are batched matrices to factorize.
Because the factorization is all done at block scope, synchronization between parallel operations is possible, so the order of operations of the factorization is respected
The only parallelism the algorithm exposes is within the row operations of the factorization and update operations
Blocks should be sized according to the number of rows in the batch matrix size in round multiples of the warp size (32 on all CUDA capable devices to date)
The code below has been extremely lightly tested and is not guaranteed to work or be correct. Use at your own peril:
#include <iostream>
#include <algorithm>
__global__
void batchkernel(float** batches, int nbatches, int N, int LDA)
{
if (blockIdx.x < nbatches) {
float* pA = batches[blockIdx.x];
for (int k = 0; k < N; k++) {
// Panel factorization.
if (threadIdx.x == 0) {
pA[k*LDA+k] = sqrtf(pA[k*LDA+k]);
}
__syncthreads();
for (int m = threadIdx.x; ((m < N) && (threadIdx.x > k)); m+=blockDim.x) {
pA[k*LDA+m] /= pA[k*LDA+k];
}
__syncthreads();
// Update of the trailing submatrix.
for (int n = k+1; (n < N); n++) {
for (int m = threadIdx.x; ((m < N) && (threadIdx.x >= n)); m+=blockDim.x) {
pA[n*LDA+m] -= pA[k*LDA+n] * pA[k*LDA+m];
}
}
__syncthreads();
}
}
}
void refCholeskey(float* pA, int N)
{
int k, m, n;
// Single Cholesky factorization.
for (k = 0; k < N; k++) {
// Panel factorization.
pA[k*N+k] = sqrtf(pA[k*N+k]);
for (m = k+1; m < N; m++)
pA[k*N+m] /= pA[k*N+k];
// Update of the trailing submatrix.
for (n = k+1; n < N; n++)
for (m = n; m < N; m++)
pA[n*N+m] -= (pA[k*N+n]*pA[k*N+m]);
}
}
int main()
{
// B = np.random.random((10,10))
// SPDmatrix = (0.5*(B+B.T)) + B.shape[0]*np.eye(B.shape[0])
const int N = 10;
const int LDA = 10;
float SPDmatrix[LDA*N] = {
10.22856331, 0.17380577, 0.61779525, 0.66592082, 0.46915566,
0.09946502, 0.69386511, 0.35224291, 0.53155506, 0.51441469,
0.17380577, 10.67971161, 0.34481401, 0.64766522, 0.22372943,
0.55896022, 0.59083588, 0.48872497, 0.54049871, 0.74764959,
0.61779525, 0.34481401, 10.229388, 0.40904432, 0.5015491,
0.52152334, 0.19684814, 0.28262256, 0.04384535, 0.61919751,
0.66592082, 0.64766522, 0.40904432, 10.78410647, 0.12708693,
0.3241063, 0.6984497, 0.65074097, 0.08027563, 0.56332844,
0.46915566, 0.22372943, 0.5015491, 0.12708693, 10.52234091,
0.76346103, 0.80932473, 0.8234331, 0.52737611, 0.65777357,
0.09946502, 0.55896022, 0.52152334, 0.3241063, 0.76346103,
10.54906761, 0.32865411, 0.32467483, 0.80720007, 0.36287463,
0.69386511, 0.59083588, 0.19684814, 0.6984497, 0.80932473,
0.32865411, 10.29729551, 0.34707933, 0.69379356, 0.87612982,
0.35224291, 0.48872497, 0.28262256, 0.65074097, 0.8234331,
0.32467483, 0.34707933, 10.42929929, 0.78849458, 0.159371,
0.53155506, 0.54049871, 0.04384535, 0.08027563, 0.52737611,
0.80720007, 0.69379356, 0.78849458, 10.49604818, 0.43871288,
0.51441469, 0.74764959, 0.61919751, 0.56332844, 0.65777357,
0.36287463, 0.87612982, 0.159371, 0.43871288, 10.94535485 };
const int nbatches = 8;
float** batches;
cudaMallocManaged((void **)&batches, nbatches * sizeof(float*));
for(int i=0; i<nbatches; i++) {
cudaMallocManaged((void **)&batches[i], N * LDA * sizeof(float));
cudaMemcpy(batches[i], SPDmatrix, N * LDA * sizeof(float), cudaMemcpyDefault);
}
int blocksz = 32;
int nblocks = nbatches;
batchkernel<<<nblocks, blocksz>>>(batches, nbatches, N, LDA);
refCholeskey(SPDmatrix, N);
cudaDeviceSynchronize();
float maxabsrelerror = 0.0f;
for(int i = 0; i < N*N; i++) {
float absrelerror = std::fabs(SPDmatrix[i] - batches[0][i]) / std::fabs(SPDmatrix[i]);
maxabsrelerror = std::max(absrelerror, maxabsrelerror);
}
std::cout << "Maximum absolute relative error = " << maxabsrelerror << std::endl;
cudaDeviceReset();
return 0;
}
I want to write a prefix scan for large arrays using the instruction in GPUgem, It's a homework for my parallel class. I did follow all the steps in the book but still my code's not working. I got it to work for array size 4096 but it's not working for larger arrays. Here is my code :
#include <stdio.h>
#include <sys/time.h>
#define THREADS 1024
typedef int mytype;
__global__ void phaseI(mytype *g_odata, mytype *g_idata, int n, mytype *aux)
{
__shared__ mytype temp[THREADS];
const int tid1 = threadIdx.x;
int offset = 1;
temp[2*tid1] = g_idata[2*tid1]; // load input into shared memory
temp[2*tid1+1] = g_idata[2*tid1+1];
for (int d = THREADS>>1; d > 0; d >>= 1) // build sum in place up the tree
{
__syncthreads();
if (tid1 < d)
{
int ai = offset*(2*tid1+1)-1;
int bi = offset*(2*tid1+2)-1;
temp[bi] += temp[ai];
}
offset *= 2;
}
__syncthreads();
if (tid1 == 0) {
aux[blockIdx.x] = temp[THREADS - 1];
temp[THREADS - 1] = 0;
}
for (int d = 1; d < THREADS; d *= 2) // traverse down tree & build scan
{
offset >>= 1;
__syncthreads();
if (tid1 < d)
{
int ai = offset*(2*tid1+1)-1;
int bi = offset*(2*tid1+2)-1;
mytype t = temp[ai];
temp[ai] = temp[bi];
temp[bi] += t;
}
}
__syncthreads();
g_odata[2*thid] = temp[2*thid]; // write results to device memory
g_odata[2*thid+1] = temp[2*thid+1];
}
__global__ void phaseII(mytype *g_odata, mytype *aux, int n)
{
const int tid1 = threadIdx.x;
const int B = (n / THREADS);
int offset = 1;
for (int d = B>>1; d > 0; d >>= 1) // build sum in place up the tree
{
__syncthreads();
if (tid1 < d)
{
int ai = offset*(2*tid1+1)-1;
int bi = offset*(2*tid1+2)-1;
temp[bi] += temp[ai];
}
offset *= 2;
}
__syncthreads();
if (tid1 == 0 && blockIdx.x == 0) {
aux[B - 1] = 0;
}
for (int d = 1; d < B; d *= 2) // traverse down tree & build scan
{
offset >>= 1;
__syncthreads();
if (tid1 < d)
{
int ai = offset*(2*tid1+1)-1;
int bi = offset*(2*tid1+2)-1;
mytype t = temp[ai];
temp[ai] = temp[bi];
temp[bi] += t;
}
}
__syncthreads();
g_odata[2*thid] += aux[blockIdx.x];
g_odata[2*thid+1] += aux[blockIdx.x];
}
int main(int argc, char *argv[])
{
if (argc != 2) {
printf("usage: %s n\n", argv[0]);
return -1;
}
const int n = atoi(argv[1]);
mytype *h_i, *d_i, *h_o, *d_o, *d_temp;
const int size = n * sizeof(mytype);
h_i = (mytype *)malloc(size);
h_o = (mytype *)malloc(size);
if ((h_i == NULL) || (h_o == NULL)) {
printf("malloc failed\n");
return -1;
}
for (int i = 0; i < n; i++) {
h_i[i] = i;
h_o[i] = 0;
}
cudaMalloc(&d_i, size);
cudaMalloc(&d_temp, (n / THREADS) );
cudaMalloc(&d_o, size);
cudaMemset(d_o, 0, size);
cudaMemset(d_temp, 0, (n / THREADS));
cudaMemcpy(d_i, h_i, size, cudaMemcpyHostToDevice);
int blocks = n / THREADS;
phaseI<<<blocks, THREADS / 2 >>>(d_o, d_i, n, d_temp);
phaseII<<<blocks, THREADS / 2>>>(d_o, d_temp, n);
cudaThreadSynchronize();
cudaMemcpy(h_o, d_o, size, cudaMemcpyDeviceToHost);
printf("\n");
for (int i = 0; i < n ; i++) {
printf(" %d", h_o[i]);
}
printf("\n\n");
return 0;
}
Does anyone have any idea what I'm doing wrong?
One possible error I see in your code is here:
aux[thid] = temp[THREADS];
If your temp array is temp[1024], as you say, and each block has 1024 threads, as you say, then if THREADS is 1024, temp[THREADS] will access your shared memory array out-of-bounds (one past the end.) An array of 1024 elements only has valid indices from 0 to 1023.
Beyond that, it seems like you're asking how to take the last element out of a shared memory array (temp) and place it in a position in a (presumably global) aux array, which has one element for each block.
Here's a fully worked example:
$ cat t831.cu
#include <stdio.h>
#define THREADS 1024
#define BLOCKS 20
__global__ void kernel(int *aux){
__shared__ int temp[THREADS];
temp[threadIdx.x] = threadIdx.x + blockIdx.x;
__syncthreads();
if (threadIdx.x == 0)
aux[blockIdx.x] = temp[THREADS-1];
}
int main(){
int *h_data, *d_data;
const int dsize = BLOCKS*sizeof(int);
h_data=(int *)malloc(dsize);
cudaMalloc(&d_data, dsize);
memset(h_data, 0, dsize);
cudaMemset(d_data, 0, dsize);
kernel<<<BLOCKS, THREADS>>>(d_data);
cudaMemcpy(h_data, d_data, dsize, cudaMemcpyDeviceToHost);
for (int i = 0; i < BLOCKS; i++) printf("%d, ", h_data[i]);
printf("\n");
return 0;
}
$ nvcc -o t831 t831.cu
$ cuda-memcheck ./t831
========= CUDA-MEMCHECK
1023, 1024, 1025, 1026, 1027, 1028, 1029, 1030, 1031, 1032, 1033, 1034, 1035, 1036, 1037, 1038, 1039, 1040, 1041, 1042,
========= ERROR SUMMARY: 0 errors
$
Closed. This question needs debugging details. It is not currently accepting answers.
Edit the question to include desired behavior, a specific problem or error, and the shortest code necessary to reproduce the problem. This will help others answer the question.
Closed 7 years ago.
Improve this question
I've written the CUDA code below. It's supposed to transpose a matrix using tiling blocks, and the code works when using small values, but when using, for example:
TILE = 32, matrix 128 x 128, it doesn't complete the transpose, it stops after 96. In host this is my dimension thread/block
dim3 dimGrid((nEven + TILE_DIM - 1) / TILE_DIM, (nEven + TILE_DIM - 1) / TILE_DIM);
dim3 dimBlock(TILE_DIM, TILE_DIM);
where I let the threads number == to tile block number,
the global code is simple and it should theoretically work:
__global__ void transposeMain( int *idata)
{
__shared__ int tile2[TILE_DIM][TILE_DIM];
int yyy = blockIdx.y * TILE_DIM ; // col values (0,32,64,96)
int xxx = blockIdx.x * TILE_DIM ; // row values (0,32,64,96)
if (xxx < nEven && yyy < nEven)
{
tile2[threadIdx.x][threadIdx.y] = idata[(threadIdx.x + xxx)*nEven + (threadIdx.y + yyy)];
__syncthreads();
idata[(threadIdx.y + yyy)*nEven + (threadIdx.x + xxx)] = tile2[threadIdx.x][threadIdx.y];
}
}
Any idea what might be the problem?
The problem is you are trying to do an in-place transpose.
CUDA device code execution is broken up into threadblocks. Threadblocks (groups of threads) can execute in any order, and do not all (typically) execute at the same time. So when you read a tile in here:
tile2[threadIdx.x][threadIdx.y] = idata[(threadIdx.x + xxx)*nEven + (threadIdx.y + yyy)];
That is OK. But when you write the tile:
idata[(threadIdx.y + yyy)*nEven + (threadIdx.x + xxx)] = tile2[threadIdx.x][threadIdx.y];
You are frequently over-writing data (in some other tile in the original matrix) which you haven't read yet (because the threadblock responsible for reading that tile hasn't even begun to execute yet). Once you overwrite it like this, it's lost.
The solution (for square matrix transpose) has several aspects to it:
Each threadblock must first read 2 tiles. These 2 tiles from the input data will be swapped.
Then each threadblock can write those two tiles.
The tiles along the main diagonal need special casing.
since most threadblocks are handling 2 tiles, only threadblocks on or on one side of the main diagonal need do any work.
You haven't shown a complete MCVE (which is expected when you have questions like this), and your code has other issues such as the potential for uncoalesced access (lower performance) so I'm not going to try to "fix" your code.
Instead, here's a fully worked example, lifted from here:
$ cat t469.cu
#include <stdio.h>
#include <cublas_v2.h>
#include <time.h>
#include <sys/time.h>
#define uS_PER_SEC 1000000
#define uS_PER_mS 1000
#define N 4096
#define M 4096
#define TILE_DIM 32
#define BLOCK_ROWS 8
__global__ void transposeCoalesced(float *odata, const float *idata)
{
__shared__ float tile[TILE_DIM][TILE_DIM+1];
int x = blockIdx.x * TILE_DIM + threadIdx.x;
int y = blockIdx.y * TILE_DIM + threadIdx.y;
int width = gridDim.x * TILE_DIM;
for (int j = 0; j < TILE_DIM; j += BLOCK_ROWS)
tile[threadIdx.y+j][threadIdx.x] = idata[(y+j)*width + x];
__syncthreads();
x = blockIdx.y * TILE_DIM + threadIdx.x; // transpose block offset
y = blockIdx.x * TILE_DIM + threadIdx.y;
for (int j = 0; j < TILE_DIM; j += BLOCK_ROWS)
odata[(y+j)*width + x] = tile[threadIdx.x][threadIdx.y + j];
}
__global__ void iptransposeCoalesced(float *data)
{
__shared__ float tile_s[TILE_DIM][TILE_DIM+1];
__shared__ float tile_d[TILE_DIM][TILE_DIM+1];
int x = blockIdx.x * TILE_DIM + threadIdx.x;
int y = blockIdx.y * TILE_DIM + threadIdx.y;
int width = gridDim.x * TILE_DIM;
if (blockIdx.y>blockIdx.x) { // handle off-diagonal case
int dx = blockIdx.y * TILE_DIM + threadIdx.x;
int dy = blockIdx.x * TILE_DIM + threadIdx.y;
for (int j = 0; j < TILE_DIM; j += BLOCK_ROWS)
tile_s[threadIdx.y+j][threadIdx.x] = data[(y+j)*width + x];
for (int j = 0; j < TILE_DIM; j += BLOCK_ROWS)
tile_d[threadIdx.y+j][threadIdx.x] = data[(dy+j)*width + dx];
__syncthreads();
for (int j = 0; j < TILE_DIM; j += BLOCK_ROWS)
data[(dy+j)*width + dx] = tile_s[threadIdx.x][threadIdx.y + j];
for (int j = 0; j < TILE_DIM; j += BLOCK_ROWS)
data[(y+j)*width + x] = tile_d[threadIdx.x][threadIdx.y + j];
}
else if (blockIdx.y==blockIdx.x){ // handle on-diagonal case
for (int j = 0; j < TILE_DIM; j += BLOCK_ROWS)
tile_s[threadIdx.y+j][threadIdx.x] = data[(y+j)*width + x];
__syncthreads();
for (int j = 0; j < TILE_DIM; j += BLOCK_ROWS)
data[(y+j)*width + x] = tile_s[threadIdx.x][threadIdx.y + j];
}
}
int validate(const float *mat, const float *mat_t, int n, int m){
int result = 1;
for (int i = 0; i < n; i++)
for (int j = 0; j < m; j++)
if (mat[(i*m)+j] != mat_t[(j*n)+i]) result = 0;
return result;
}
int main(){
timeval t1, t2;
float *matrix = (float *) malloc (N * M * sizeof(float));
for (int i = 0; i < N; i ++)
for (int j = 0; j < M; j++)
matrix[(i*M) + j] = i;
// Starting the timer
gettimeofday(&t1, NULL);
float *matrixT = (float *) malloc (N * M * sizeof(float));
for (int i = 0; i < N; i++)
for (int j = 0; j < M; j++)
matrixT[(j*N)+i] = matrix[(i*M)+j]; // matrix is obviously filled
//Ending the timer
gettimeofday(&t2, NULL);
if (!validate(matrix, matrixT, N, M)) {printf("fail!\n"); return 1;}
float et1 = (((t2.tv_sec*uS_PER_SEC)+t2.tv_usec) - ((t1.tv_sec*uS_PER_SEC)+t1.tv_usec))/(float)uS_PER_mS;
printf("CPU time = %fms\n", et1);
float *h_matrixT , *d_matrixT , *d_matrix;
h_matrixT = (float *) (malloc (N * M * sizeof(float)));
cudaMalloc((void **)&d_matrixT , N * M * sizeof(float));
cudaMalloc((void**)&d_matrix , N * M * sizeof(float));
cudaMemcpy(d_matrix , matrix , N * M * sizeof(float) , cudaMemcpyHostToDevice);
//Starting the timer
gettimeofday(&t1, NULL);
const float alpha = 1.0;
const float beta = 0.0;
cublasHandle_t handle;
//gettimeofday(&t1, NULL);
cublasCreate(&handle);
gettimeofday(&t1, NULL);
cublasSgeam(handle, CUBLAS_OP_T, CUBLAS_OP_N, N, M, &alpha, d_matrix, M, &beta, d_matrix, N, d_matrixT, N);
cudaDeviceSynchronize();
gettimeofday(&t2, NULL);
cublasDestroy(handle);
//Ending the timer
float et2 = (((t2.tv_sec*uS_PER_SEC)+t2.tv_usec) - ((t1.tv_sec*uS_PER_SEC)+t1.tv_usec))/(float)uS_PER_mS;
printf("GPU Sgeam time = %fms\n", et2);
cudaMemcpy(h_matrixT , d_matrixT , N * M * sizeof(float) , cudaMemcpyDeviceToHost);
if (!validate(matrix, h_matrixT, N, M)) {printf("fail!\n"); return 1;}
cudaMemset(d_matrixT,0, N*M*sizeof(float));
memset(h_matrixT, 0, N*M*sizeof(float));
dim3 threads(TILE_DIM, BLOCK_ROWS);
dim3 blocks(N/TILE_DIM, M/TILE_DIM);
gettimeofday(&t1, NULL);
transposeCoalesced<<<blocks, threads >>>(d_matrixT, d_matrix);
cudaDeviceSynchronize();
gettimeofday(&t2, NULL);
cudaMemcpy(h_matrixT , d_matrixT , N * M * sizeof(float) , cudaMemcpyDeviceToHost);
if (!validate(matrix, h_matrixT, N, M)) {printf("fail!\n"); return 1;}
float et3 = (((t2.tv_sec*uS_PER_SEC)+t2.tv_usec) - ((t1.tv_sec*uS_PER_SEC)+t1.tv_usec))/(float)uS_PER_mS;
printf("GPU kernel time = %fms\n", et3);
memset(h_matrixT, 0, N*M*sizeof(float));
gettimeofday(&t1, NULL);
iptransposeCoalesced<<<blocks, threads >>>(d_matrix);
cudaDeviceSynchronize();
gettimeofday(&t2, NULL);
cudaMemcpy(h_matrixT , d_matrix , N * M * sizeof(float) , cudaMemcpyDeviceToHost);
if (!validate(matrix, h_matrixT, N, M)) {printf("fail!\n"); return 1;}
float et4 = (((t2.tv_sec*uS_PER_SEC)+t2.tv_usec) - ((t1.tv_sec*uS_PER_SEC)+t1.tv_usec))/(float)uS_PER_mS;
printf("GPU in-place kernel time = %fms\n", et4);
cudaFree(d_matrix);
cudaFree(d_matrixT);
return 0;
}
$ nvcc -arch=sm_20 -o t469 t469.cu -lcublas
$ ./t469
CPU time = 450.095001ms
GPU Sgeam time = 1.937000ms
GPU kernel time = 1.694000ms
GPU in-place kernel time = 1.839000ms
$
Note that this compares several different approaches to matrix transpose.
If you study the iptransposeCoalesced you will see that it is adhering to the 4 specific aspects I outlined above.
It is fishy to use __syncthreads(); in the if statement in CUDA. Try to move it outside this block by simple:
if (xxx < nEven && yyy < nEven)
{
tile2[threadIdx.x][threadIdx.y] = idata[(threadIdx.x + xxx)*nEven + (threadIdx.y + yyy)];
}
__syncthreads();
if (xxx < nEven && yyy < nEven)
{
idata[(threadIdx.y + yyy)*nEven + (threadIdx.x + xxx)] = tile2[threadIdx.x][threadIdx.y];
}
I am trying to use cudaMemcpy3D to transfer dynamically allocated 3d matrix (tensor). Tensor is allocated as contiguous block of memory (see code below). I tried various combinations of cudaExtent and cudaMemcpy3DParms, however the order of elements gets mixed up. I created the following example to demonstrate the issue:
#include <stdio.h>
int ***alloc_tensor(int Nx, int Ny, int Nz) {
int i, j;
int ***tensor;
tensor = (int ***) malloc((size_t) (Nx * sizeof(int **)));
tensor[0] = (int **) malloc((size_t) (Nx * Ny * sizeof(int *)));
tensor[0][0] = (int *) malloc((size_t) (Nx * Ny * Nz * sizeof(int)));
for(j = 1; j < Ny; j++)
tensor[0][j] = tensor[0][j-1] + Nz;
for(i = 1; i < Nx; i++) {
tensor[i] = tensor[i - 1] + Ny;
tensor[i][0] = tensor[i - 1][0] + Ny * Nz;
for(j = 1; j < Ny; j++)
tensor[i][j] = tensor[i][j - 1] + Nz;
}
return tensor;
}
__global__ void kernel(cudaPitchedPtr tensor, int Nx, int Ny, int Nz) {
int i, j, k;
char *tensorslice;
int *tensorrow;
for (i = 0; i < Nx; i++) {
for (j = 0; j < Ny; j++) {
for (k = 0; k < Nz; k++) {
tensorslice = ((char *)tensor.ptr) + k * tensor.pitch * Nx;
tensorrow = (int *)(tensorslice + i * tensor.pitch);
printf("d_tensor[%d][%d][%d] = %d\n", i, j, k, tensorrow[j]);
}
}
}
}
int main() {
int i, j, k, value = 0;
int Nx = 2, Ny = 6, Nz = 4;
int ***h_tensor;
struct cudaPitchedPtr d_tensor;
h_tensor = alloc_tensor(Nx, Ny, Nz);
cudaMalloc3D(&d_tensor, make_cudaExtent(Nx * sizeof(int), Ny, Nz));
for(i = 0; i < Nx; i++) {
for(j = 0; j < Ny; j++) {
for(k = 0; k < Nz; k++) {
h_tensor[i][j][k] = value++;
printf("h_tensor[%d][%d][%d] = %d\n", i, j, k, h_tensor[i][j][k]);
}
}
}
cudaMemcpy3DParms cpy = { 0 };
cpy.srcPtr = make_cudaPitchedPtr(h_tensor[0][0], Nx * sizeof(int), Ny, Nz);
cpy.dstPtr = d_tensor;
cpy.extent = make_cudaExtent(Nx * sizeof(int), Ny, Nz);
cpy.kind = cudaMemcpyHostToDevice;
cudaMemcpy3D(&cpy);
kernel<<<1, 1>>>(d_tensor, Nx, Ny, Nz);
// ... clean-up
}
Output for host variable (h_tensor) and device (d_tensor) differ, looking like
h_tensor[0][0][0] = 0
h_tensor[0][0][1] = 1
h_tensor[0][0][2] = 2
h_tensor[0][0][3] = 3
h_tensor[0][1][0] = 4
h_tensor[0][1][1] = 5
h_tensor[0][1][2] = 6
...
d_tensor[0][0][0] = 0
d_tensor[0][0][1] = 12
d_tensor[0][0][2] = 24
d_tensor[0][0][3] = 36
d_tensor[0][1][0] = 1
d_tensor[0][1][1] = 13
d_tensor[0][1][2] = 25
...
What am I doing wrong? What would be the correct way to use cudaMemcpy3D?
Any time you are having trouble with a cuda code, it's a good idea to do proper cuda error checking. The code you have posted here, at least, does not run correctly for me - the cudaMemcpy3D line throws an error. This is due to item 2 below. (I suspect the code you used to generate the output was not identical to the code you have shown here, but that's just a guess.)
Your usage of make_cudaPitchedPtr is not correct:
cpy.srcPtr = make_cudaPitchedPtr(h_tensor[0][0], Nx * sizeof(int), Ny, Nz);
review the API documentation. Making a CUDA pitched pointer this way is no different between 2D and 3D. So it makes no sense to pass 3 different dimensions as you are doing. Instead do this:
cpy.srcPtr = make_cudaPitchedPtr(h_tensor[0][0], Nx * sizeof(int), Nx, Ny);
The remaining issues I found I attribute to incorrect understanding of 3 dimensions in C. The last subscript on a multiply-subscripted array is the rapidly varying dimension, i.e. it is the one where adjacent values in memory occupy adjacent index values. Your usage of Z in the 3rd dimension is confusing to me due to this. Your host allocation was using Nx in the first subscript place, but your device indexing didn't match. There are obviously multiple ways to handle this. If you don't like my arrangement, you can change it, but the host and device indexing must match.
Anyway, the following code modifications worked for me:
#include <stdio.h>
int ***alloc_tensor(int Nx, int Ny, int Nz) {
int i, j;
int ***tensor;
tensor = (int ***) malloc((size_t) (Nx * sizeof(int **)));
tensor[0] = (int **) malloc((size_t) (Nx * Ny * sizeof(int *)));
tensor[0][0] = (int *) malloc((size_t) (Nx * Ny * Nz * sizeof(int)));
for(j = 1; j < Ny; j++)
tensor[0][j] = tensor[0][j-1] + Nz;
for(i = 1; i < Nx; i++) {
tensor[i] = tensor[i - 1] + Ny;
tensor[i][0] = tensor[i - 1][0] + Ny * Nz;
for(j = 1; j < Ny; j++)
tensor[i][j] = tensor[i][j - 1] + Nz;
}
return tensor;
}
__global__ void kernel(cudaPitchedPtr tensor, int Nx, int Ny, int Nz) {
int i, j, k;
char *tensorslice;
int *tensorrow;
for (i = 0; i < Nx; i++) {
for (j = 0; j < Ny; j++) {
for (k = 0; k < Nz; k++) {
tensorslice = ((char *)tensor.ptr) + k * tensor.pitch * Ny;
tensorrow = (int *)(tensorslice + j * tensor.pitch);
printf("d_tensor[%d][%d][%d] = %d\n", i, j, k, tensorrow[i]);
}
}
}
}
int main() {
int i, j, k, value = 0;
int Nx = 2, Ny = 6, Nz = 4;
int ***h_tensor;
struct cudaPitchedPtr d_tensor;
h_tensor = alloc_tensor(Nz, Ny, Nx);
cudaMalloc3D(&d_tensor, make_cudaExtent(Nx * sizeof(int), Ny, Nz));
for(i = 0; i < Nx; i++) {
for(j = 0; j < Ny; j++) {
for(k = 0; k < Nz; k++) {
h_tensor[k][j][i] = value++;
//printf("h_tensor[%d][%d][%d] = %d\n", i, j, k, h_tensor[i][j][k]);
}
}
}
for(i = 0; i < Nx; i++) {
for(j = 0; j < Ny; j++) {
for(k = 0; k < Nz; k++) {
//h_tensor[i][j][k] = value++;
printf("h_tensor[%d][%d][%d] = %d\n", i, j, k, h_tensor[k][j][i]);
}
}
}
cudaMemcpy3DParms cpy = { 0 };
cpy.srcPtr = make_cudaPitchedPtr(h_tensor[0][0], Nx * sizeof(int), Nx, Ny);
cpy.dstPtr = d_tensor;
cpy.extent = make_cudaExtent(Nx * sizeof(int), Ny, Nz);
cpy.kind = cudaMemcpyHostToDevice;
cudaMemcpy3D(&cpy);
kernel<<<1, 1>>>(d_tensor, Nx, Ny, Nz);
cudaDeviceSynchronize();
// ... clean-up
}
I wrote a code that uses GPU to do the forward solve (get U matrix from A = LU where here the diagonal entries of U are not set to unity). My code works fine for matrices of dimension less than 128x128 which are a factor of the block size which is 16. I am puzzled why larger matrices give me the wrong result. Something goes wrong in the last row of blocks.
------------ gaussian.h --------------------
// Thread block size
#define BLOCK_SIZE 16 //for forward solve A = LU
#define BLOCK_SIZE2 32 //for finding the pivot factor
// A[i,j] = A[i,j] - A[k,j] / A[k,k] * A[i,k]
// pivot factor = A[k,j] / A[k,k]
////////////////////////////////////////////////////////////////////////////////////////
------------ gaussian.cu --------------------
// includes, system
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <math.h>
#include <sys/time.h>
#include <cuda.h>
#include "gaussian_kernel.cu"
#define OUTPUT
void runTest(int argc, char** argv);
double gettime() {
struct timeval t;
gettimeofday(&t,NULL);
return t.tv_sec+t.tv_usec*1e-6;
}
int main(int argc, char** argv)
{
runTest(argc, argv);
}
void runTest(int argc, char** argv)
{
int dim;
if (argc == 2)
{
dim = atoi(argv[1]);
}
else{
printf("Wrong Usage\n");
exit(1);
}
// Note: A is a square matrix of size NxN where N % BLOCK_SIZE = 0
// Every row of A has a pivot column
// It is known that all the arithmetic operations involved for
// Gaussian Elimination are of type int (not float)
// allocate host memory for matrix A augmented with vector b (Ax=b)
unsigned int size_A = dim * (dim + BLOCK_SIZE);
unsigned int mem_size_A = sizeof(int) * size_A;
int* h_A = (int*) malloc(mem_size_A);
// initialize host memory, generate a test case such as below
// augmented matrix A with padding to make the size evenly divisible by BLOCK_SIZE
// ----A----- | b --padding--
// 1 1 1 1 .. | 0 * * * ..
// 1 2 2 2 .. | 1 * * * ..
// 1 2 3 3 .. | 2 * * * ..
// 1 2 3 4 .. | 3 * * * ..
// .......... | ...
// * means uninitialized entry
// A is size NxN
// Augmented matrix with padding is size Nx(N+BLOCK_SIZE)
int dimRow = dim;
int dimCol = dim + BLOCK_SIZE;
for(int i = 0; i < dim; i++)
{
h_A[(i + 1) * dimCol - BLOCK_SIZE] = i; // b vector stored in (N+1)th column
// of augmented matrix
for (int j = 0; j < dimRow - i; j++)
{
h_A[j + i + i * dimCol] = i + 1; // A[i][j] entry
h_A[j * dimCol + i + i * dimCol] = i + 1; // A[j][i] entry
}
}
//display the test case [ A | b ]
for ( int m = 0 ; m < dimRow; m++)
{
for ( int n = 0 ; n <= dimRow; n++)
//for ( int n = 0 ; n < dimCol; n++)
{
printf("%d ", h_A[m * dimCol + n]);
}
printf("\n");
}
// allocate device memory for the augmented matrix A
int* d_A;
cudaMalloc(&d_A, mem_size_A);
// start timer
double timer1 = gettime();
// copy host memory to device
cudaMemcpy(d_A, h_A, mem_size_A, cudaMemcpyHostToDevice);
// setup execution parameters for obtaining the pivot factor
int gridP = (dimRow / BLOCK_SIZE2) + ( (dimRow % BLOCK_SIZE2) == 0 ? 0 : 1 );
// add 1 if dimRow/BLOCK_SIZE2 is not evenly divisible
// setup execution parameters for the forward solve (Gaussian Elimination)
dim3 threads(BLOCK_SIZE, BLOCK_SIZE);
dim3 grid(dimCol / threads.x, dimRow / threads.y);
// execute the kernel
for ( int i = 0 ; i < dim; i++)
{
Gaussian_Pivot_CUDA<<< gridP, BLOCK_SIZE2 >>>(d_A, dimCol, i);
Gaussian_Elim_CUDA<<< grid, threads >>>(d_A, dimCol, i);
}
// copy result from device to host
cudaMemcpy(h_A, d_A, mem_size_A, cudaMemcpyDeviceToHost);
// stop timer
double timer2 = gettime();
printf("GPU time = %lf\n",(timer2-timer1)*1000);
// Back substitution
int X[dimRow];
for (int row = dimRow - 1; row >= 0; row--)
{
X[row] = h_A[(row + 1) * dimCol - BLOCK_SIZE]; // b[row] entry
for (int col = dimRow - 1; col > row; col--)
{
X[row] -= h_A[row * dimCol + col] * X[col];
}
X[row] /= h_A[row * dimCol + row];
printf("X[%d] = %d\t",row,X[row]);
}
printf("\n");
#ifdef OUTPUT
// result of Gaussian Elimination
// ----A----- | b --padding--
// 1 1 1 1 .. | 0 * * * ..
// 0 1 1 1 .. | 1 * * * ..
// 0 0 1 1 .. | 1 * * * ..
// 0 0 0 1 .. | 1 * * * ..
// .......... | ...
// * means garbage entry
for ( int m = 0 ; m < dimRow; m++)
{
for ( int n = 0 ; n <= dimRow; n++)
{
printf("%d ", h_A[m * dimCol + n]);
}
printf("\n");
}
#endif
free(h_A);
cudaFree(d_A);
}
///////////////////////////////////////////////////////////////////////////////
------------ gaussian_kernel.cu --------------------
#include "gaussian.h"
__global__
void Gaussian_Pivot_CUDA(int* A, int widthA, int Pcol)
{
// find the pivot factor
// A[i,j] = A[i,j] - A[k,j] / A[k,k] * A[i,k]
// pivot factor = A[k,j] / A[k,k]
int bx = blockIdx.x;
int tx = threadIdx.x;
int index_row = BLOCK_SIZE2 * bx + tx;
int index_rowA = index_row * widthA;
__shared__ int A_RowRow;
if (tx == 0)
{
A_RowRow = A[Pcol * widthA + Pcol]; // get A[k,k] where k is the pivot column
// for this problem pivot column = pivot row
}
__syncthreads();
A[index_rowA - 1] = A[index_rowA + Pcol] / A_RowRow;
// INVALID WRITE HERE!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
// store pivot factor at A[dimCol - 1][row]
}
__global__
void Gaussian_Elim_CUDA(int* A, int widthA, int Pcol)
{
// Block index
int bx = blockIdx.x;
int by = blockIdx.y;
// Thread index
int tx = threadIdx.x;
int ty = threadIdx.y;
int index_col = BLOCK_SIZE * bx + tx;
int index_row = BLOCK_SIZE * by + ty;
int index = widthA * index_row + index_col;
// d_A[index] "=" d_A[index_row][index_col]
// store pivot factor for rows that we are working on in this block
__shared__ int pivotFactor[BLOCK_SIZE];
if ( ty == 0 ) // use ty instead of tx for coalesced accesses
{
pivotFactor[tx] = A[(index_row + tx) * widthA - 1];
// INVALID WRITE HERE!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
}
__syncthreads();
// implement the gaussian elimination for the current row
if ( index_row > Pcol )
{
A[index] -= A[Pcol * widthA + index_col] * pivotFactor[ty];
// A[i,j] = A[i,j] - A[k,j] / A[k,k] * A[i,k]
// pivot factor = A[k,j] / A[k,k]
}
}
cuda-memcheck told me that I was reading and writing stuff out of bounds.
My indexing was off. In Gaussian_Pivot_CUDA, the last line should be A[widthA*(index_row+1)-1)]=...
In Gaussian_Elim_CUDA, when writing pivotFactor I have to change ty to tx and read A[widthA*(index_row+1)-1)].
Thank you #Eugene !!!
If sizeof(int)==2, then This array index will overflow when dim is 128:
h_A[j * dimCol + i + i * dimCol] = i + 1; // A[j][i] entry
(127*144)+127+(127*144) = 36703 > 32767.