From 49ceb0fe041cbe62ce6016d917bbc55e9645e3d9 Mon Sep 17 00:00:00 2001
From: Chao Liu <lc.roy86@gmail.com>
Date: Mon, 22 Oct 2018 20:51:26 -0500
Subject: [PATCH] initial cuda run

---
 src/include/device_tensor.cuh      | 26 ++++++++++++--------------
 src/include/direct_convolution.cuh |  6 +++---
 2 files changed, 15 insertions(+), 17 deletions(-)

diff --git a/src/include/device_tensor.cuh b/src/include/device_tensor.cuh
index 87865b53ad..56338452f6 100644
--- a/src/include/device_tensor.cuh
+++ b/src/include/device_tensor.cuh
@@ -14,26 +14,24 @@ struct DeviceTensorDescriptor
         checkCudaErrors(cudaMalloc(&mpLengths, data_sz * mDim));
         checkCudaErrors(cudaMalloc(&mpStrides, data_sz * mDim));
 
-        checkCudaErrors(
-            cudaMemcpy(const_cast<void*>(static_cast<const void*>(host_desc.GetLengths().data())),
-                       mpLengths,
-                       data_sz * mDim,
-                       cudaMemcpyHostToDevice));
-        checkCudaErrors(
-            cudaMemcpy(const_cast<void*>(static_cast<const void*>(host_desc.GetStrides().data())),
-                       mpStrides,
-                       data_sz * mDim,
-                       cudaMemcpyHostToDevice));
+        checkCudaErrors(cudaMemcpy(
+            mpLengths, host_desc.GetLengths().data(), data_sz * mDim, cudaMemcpyHostToDevice));
+        checkCudaErrors(cudaMemcpy(
+            mpStrides, host_desc.GetStrides().data(), data_sz * mDim, cudaMemcpyHostToDevice));
     }
 
     __host__ ~DeviceTensorDescriptor()
     {
-        checkCudaErrors(cudaFree(mpLengths));
-        checkCudaErrors(cudaFree(mpStrides));
+#if 0
+        if(mpLengths != nullptr)
+            checkCudaErrors(cudaFree(mpLengths));
+        if(mpStrides != nullptr)
+            checkCudaErrors(cudaFree(mpStrides));
+#endif
     }
 
     DataType_t mDataType;
     unsigned long mDim;
-    unsigned long* mpLengths;
-    unsigned long* mpStrides;
+    unsigned long* mpLengths = nullptr;
+    unsigned long* mpStrides = nullptr;
 };
diff --git a/src/include/direct_convolution.cuh b/src/include/direct_convolution.cuh
index 2816028ac8..8315dcbb55 100644
--- a/src/include/direct_convolution.cuh
+++ b/src/include/direct_convolution.cuh
@@ -3,10 +3,10 @@
 
 template <class TFloat, int NBlockDim>
 __global__ void direct_convolution(DeviceTensorDescriptor in_desc,
-                                   TFloat* const in,
+                                   TFloat* const p_in,
                                    DeviceTensorDescriptor wei_desc,
-                                   TFloat* const wei,
+                                   TFloat* const p_wei,
                                    DeviceTensorDescriptor out_desc,
-                                   TFloat* out)
+                                   TFloat* p_out)
 {
 }