AI/rocm-systems
2
0
Forka 0
Codice Problemi Pull Request Actions Pacchetti Progetti Rilasci Wiki Attività

Calculate and print kernel throughput (#78)

Sfoglia il codice sorgente 
* rccl-prim-test: print GPU info and set iterations

* Calculate and print kernel throughput
This commit is contained in:
Wenkai Du
2019-06-07 10:39:30 -07:00
committato da GitHub
parent 42b488507d
commit ee14676064
1 ha cambiato i file con 10 aggiunte e 8 eliminazioni
Scarica il file Patch Scarica il file Diff Expand all files Collapse all files
tools/rccl-prim-test/rccl_prim_test.cpp
+10 -8
Vedi File
@@ -70,9 +70,8 @@ __global__ void flag_sync_kernel(struct transfer_data_t* transfer_data, struct p
uint64_t curr_time, next_time;
if (idx == 0) {
curr_time = clock();
curr_time = clock64();
}
__syncthreads();
int offset = transfer_data->N * blockIdx.x / gridDim.x;
int n = transfer_data->N / gridDim.x;
@@ -82,10 +81,10 @@ __global__ void flag_sync_kernel(struct transfer_data_t* transfer_data, struct p
if (op == OP_REDUCE) Reduce<UNROLL, THREADS, float>(transfer_data->dest0 + offset, transfer_data->src0 + offset, transfer_data->src1 + offset, n);
if (op == OP_REDUCECOPY) ReduceCopy<UNROLL, THREADS, float>(transfer_data->dest0 + offset, transfer_data->dest1 + offset, transfer_data->src0 + offset, transfer_data->src1 + offset, n);
__syncthreads();
if (idx == 0) {
next_time = clock();
next_time = clock64();
__atomic_fetch_add(&(profiling_data->write_cycles), next_time - curr_time, __ATOMIC_SEQ_CST);
curr_time = next_time;
__atomic_fetch_add(&(profiling_data->bytes_transferred), n * sizeof(float), __ATOMIC_SEQ_CST);
}
}
@@ -353,11 +352,14 @@ int main(int argc,char* argv[])
double speed = (double)(profiling_data_0->bytes_transferred) / (deltaSec*1.0E9);
printf("Transfered %lu bytes in %f s. Throughput %f GB/s\n", profiling_data_0->bytes_transferred, deltaSec, speed);
fprintf(stderr, "GPU 0: write_cycles %ld bytes_transferred %ld\n",
profiling_data_0->write_cycles, profiling_data_0->bytes_transferred);
#define RTC_CLOCK_FREQ 2.7E07
double t0 = (double)profiling_data_0->write_cycles/((double)RTC_CLOCK_FREQ)/(double)workgroups;
fprintf(stderr, "GPU 0: time %.4fs bytes_transferred %lu kernel throughput %.2f GB/s\n",
t0, profiling_data_0->bytes_transferred, (double)profiling_data_0->bytes_transferred/(t0*1.0E9));
fprintf(stderr, "GPU 1: write_cycles %ld bytes_transferred %ld\n",
profiling_data_1->write_cycles, profiling_data_1->bytes_transferred);
double t1 = (double)profiling_data_1->write_cycles/((double)RTC_CLOCK_FREQ)/(double)workgroups;
fprintf(stderr, "GPU 1: time %.4fs bytes_transferred %lu kernel throughput %.2f GB/s\n",
t1, profiling_data_1->bytes_transferred, (double)profiling_data_0->bytes_transferred/(t1*1.0E9));
}
HIPCHECK(hipStreamDestroy(stream_0));