bda/gpu-mpi-tests
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

add mem free print, fit in 8GB gpu

Browse Source
This commit is contained in:
Bryce Allen
2020-08-07 13:21:22 -04:00
parent 3ebd09725e
commit 6940ce7ceb
1 changed files with 11 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
mpi_daxpy_nvtx.cc
View File

@@ -35,6 +35,8 @@
static cublasHandle_t handle; static cublasHandle_t handle;
static const int MB = 1024*1024;
void set_rank_device(int n_ranks, int rank) { void set_rank_device(int n_ranks, int rank) {
int n_devices, device, ranks_per_device; int n_devices, device, ranks_per_device;
@@ -66,9 +68,11 @@ void set_rank_device(int n_ranks, int rank) {
int main(int argc, char **argv) { int main(int argc, char **argv) {
int n = 32*1024*1024; int n = 4*MB;
int world_size, world_rank; int world_size, world_rank;
size_t free_mem, total_mem;
double a = 2.0; double a = 2.0;
double sum = 0.0; double sum = 0.0;
@@ -129,11 +133,17 @@ int main(int argc, char **argv) {
CHECK( "d_y", cudaMalloc((void**)&d_y, n*sizeof(*d_y)) ); CHECK( "d_y", cudaMalloc((void**)&d_y, n*sizeof(*d_y)) );
*/ */
nvtxRangePushA("allocateArrays");
CHECK( "m_x", cudaMallocManaged((void**)&m_x, n*sizeof(*m_x)) ); CHECK( "m_x", cudaMallocManaged((void**)&m_x, n*sizeof(*m_x)) );
CHECK( "m_y", cudaMallocManaged((void**)&m_y, n*sizeof(*m_y)) ); CHECK( "m_y", cudaMallocManaged((void**)&m_y, n*sizeof(*m_y)) );
CHECK( "m_allx", cudaMallocManaged((void**)&m_allx, n*sizeof(*m_allx)*world_size) ); CHECK( "m_allx", cudaMallocManaged((void**)&m_allx, n*sizeof(*m_allx)*world_size) );
CHECK( "m_ally", cudaMallocManaged((void**)&m_ally, n*sizeof(*m_ally)*world_size) ); CHECK( "m_ally", cudaMallocManaged((void**)&m_ally, n*sizeof(*m_ally)*world_size) );
nvtxRangePop();
CHECK( "memInfo", cudaMemGetInfo(&free_mem, &total_mem) );
printf("GPU memory %0.3f / %0.3f (%0.3f)\n", free_mem/(double)MB,
(double)total_mem/MB, (double)(total_mem-free_mem)/MB);
nvtxRangePushA("initializeArrays"); nvtxRangePushA("initializeArrays");
for (int i=0; i<n; i++) { for (int i=0; i<n; i++) {