|
|
@@ -33,7 +33,7 @@ static unsigned thread_is_initialized[2];
|
|
|
static pthread_cond_t cond;
|
|
|
static pthread_mutex_t mutex;
|
|
|
|
|
|
-static size_t buffer_size = 1;
|
|
|
+static size_t buffer_size = 4;
|
|
|
static void *cpu_buffer;
|
|
|
static void *gpu_buffer[2];
|
|
|
|
|
|
@@ -41,7 +41,7 @@ static pthread_cond_t cond_go;
|
|
|
static unsigned ready = 0;
|
|
|
static unsigned nready_gpu = 0;
|
|
|
|
|
|
-static unsigned niter = 100000;
|
|
|
+static unsigned niter = 250000;
|
|
|
|
|
|
static pthread_cond_t cond_gpu;
|
|
|
static pthread_mutex_t mutex_gpu;
|
|
|
@@ -50,12 +50,15 @@ static unsigned data_is_available[2];
|
|
|
static cudaStream_t stream[2];
|
|
|
|
|
|
#define ASYNC 1
|
|
|
+#define DO_TRANSFER_GPU_TO_RAM 1
|
|
|
+#define DO_TRANSFER_RAM_TO_GPU 1
|
|
|
|
|
|
void send_data(unsigned src, unsigned dst)
|
|
|
{
|
|
|
cudaError_t cures;
|
|
|
|
|
|
/* Copy data from GPU to RAM */
|
|
|
+#ifdef DO_TRANSFER_GPU_TO_RAM
|
|
|
#ifdef ASYNC
|
|
|
cures = cudaMemcpyAsync(cpu_buffer, gpu_buffer[src], buffer_size, cudaMemcpyDeviceToHost, stream[src]);
|
|
|
assert(!cures);
|
|
|
@@ -69,6 +72,7 @@ void send_data(unsigned src, unsigned dst)
|
|
|
cures = cudaThreadSynchronize();
|
|
|
assert(!cures);
|
|
|
#endif
|
|
|
+#endif
|
|
|
|
|
|
/* Tell the other GPU that data is in RAM */
|
|
|
pthread_mutex_lock(&mutex_gpu);
|
|
|
@@ -93,8 +97,9 @@ void recv_data(unsigned src, unsigned dst)
|
|
|
//fprintf(stderr, "RECV on %d\n", dst);
|
|
|
|
|
|
/* Upload data */
|
|
|
+#ifdef DO_TRANSFER_RAM_TO_GPU
|
|
|
#ifdef ASYNC
|
|
|
- cures = cudaMemcpyAsync(gpu_buffer[dst], cpu_buffer, buffer_size, cudaMemcpyDeviceToHost, stream[dst]);
|
|
|
+ cures = cudaMemcpyAsync(gpu_buffer[dst], cpu_buffer, buffer_size, cudaMemcpyHostToDevice, stream[dst]);
|
|
|
assert(!cures);
|
|
|
cures = cudaThreadSynchronize();
|
|
|
assert(!cures);
|
|
|
@@ -105,6 +110,7 @@ void recv_data(unsigned src, unsigned dst)
|
|
|
cures = cudaThreadSynchronize();
|
|
|
assert(!cures);
|
|
|
#endif
|
|
|
+#endif
|
|
|
}
|
|
|
|
|
|
void *launch_gpu_thread(void *arg)
|
|
|
@@ -112,7 +118,6 @@ void *launch_gpu_thread(void *arg)
|
|
|
unsigned *idptr = arg;
|
|
|
unsigned id = *idptr;
|
|
|
|
|
|
- fprintf(stderr, "Initialize device %d\n", id);
|
|
|
cudaSetDevice(id);
|
|
|
cudaFree(0);
|
|
|
|
|
|
@@ -130,8 +135,6 @@ void *launch_gpu_thread(void *arg)
|
|
|
|
|
|
pthread_mutex_unlock(&mutex);
|
|
|
|
|
|
- fprintf(stderr, "Device %d GOGO\n", id);
|
|
|
-
|
|
|
unsigned iter;
|
|
|
for (iter = 0; iter < niter; iter++)
|
|
|
{
|
Cédric Augonnet