From 3ccde397f9f5550fdd7134a09bfa1b387d475b28 Mon Sep 17 00:00:00 2001 From: Hieu Hoang Date: Thu, 14 Dec 2017 22:11:11 +0000 Subject: use mblas::copy --- src/amun/gpu/mblas/nth_element.cu | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-) (limited to 'src/amun/gpu') diff --git a/src/amun/gpu/mblas/nth_element.cu b/src/amun/gpu/mblas/nth_element.cu index c4a426a1..31261f5c 100644 --- a/src/amun/gpu/mblas/nth_element.cu +++ b/src/amun/gpu/mblas/nth_element.cu @@ -156,9 +156,8 @@ void NthElement::getValueByKey(std::vector& out, const mblas::Matrix &d_i // need a model with multiple scorers to test this method out.resize(d_breakdown.size()); - mblas::VectorWrapper breakdownWrap(d_breakdown); - const mblas::MatrixWrapper inWrap(d_in); - + //mblas::VectorWrapper breakdownWrap(d_breakdown); + //const mblas::MatrixWrapper inWrap(d_in); //gGetValueByKey<<<1, lastN_, 0, stream_>>> // (breakdownWrap, inWrap, h_res_idx, lastN_); /* @@ -168,8 +167,7 @@ void NthElement::getValueByKey(std::vector& out, const mblas::Matrix &d_i << h_res.size() << endl; */ - HANDLE_ERROR( cudaMemcpyAsync(out.data(), d_breakdown.data(), d_breakdown.size() * sizeof(float), - cudaMemcpyDeviceToHost, mblas::CudaStreamHandler::GetStream()) ); + mblas::copy(d_breakdown.data(), d_breakdown.size(), out.data(), cudaMemcpyDeviceToHost); HANDLE_ERROR( cudaStreamSynchronize(mblas::CudaStreamHandler::GetStream())); } -- cgit v1.2.3