Remove various synchronisations from cuda API calls, instrument monitor (#4205)

* Remove various synchronisations from cuda API calls, instrument monitor with nvtx profiler ranges.
2019-03-10 15:01:23 +13:00
parent f83e62dca5
commit 4eeeded7d1
9 changed files with 116 additions and 104 deletions
--- a/src/common/transform.h
+++ b/src/common/transform.h
@@ -145,8 +145,6 @@ class Transform {
            static_cast<int>(dh::DivRoundUp(*(range_.end()), kBlockThreads));
        detail::LaunchCUDAKernel<<<GRID_SIZE, kBlockThreads>>>(
            _func, shard_range, UnpackHDV(_vectors, device)...);
-        dh::safe_cuda(cudaGetLastError());
-        dh::safe_cuda(cudaDeviceSynchronize());
      }
    }
 #else