refactor

driver/conv.cu

@@ -302,7 +302,7 @@ template <class T>
 void check_error(const Tensor<T>& ref, const Tensor<T>& result)
 {
     float error     = 0;
-    float max_diff  = 0;
+    float max_diff  = -1;
     float ref_value = 0, result_value = 0;
     for(int i = 0; i < ref.mData.size(); ++i)
     {
@@ -338,6 +338,14 @@ int main()
     constexpr unsigned K = 64;
     constexpr unsigned S = 3;
     constexpr unsigned R = 3;
+#elif 0
+    constexpr unsigned N = 72;
+    constexpr unsigned C = 288;
+    constexpr unsigned HI = 38;
+    constexpr unsigned WI = 38;
+    constexpr unsigned K = 72;
+    constexpr unsigned S = 3;
+    constexpr unsigned R = 3;
 #elif 0
     constexpr unsigned N = 1;
     constexpr unsigned C = 1;
@@ -347,13 +355,13 @@ int main()
     constexpr unsigned S = 3;
     constexpr unsigned R = 3;
 #elif 0
-    constexpr unsigned N = 1;
-    constexpr unsigned C = 1;
+    constexpr unsigned N  = 1;
+    constexpr unsigned C  = 1;
     constexpr unsigned HI = 4;
     constexpr unsigned WI = 4;
-    constexpr unsigned K = 1;
-    constexpr unsigned S = 3;
-    constexpr unsigned R = 3;
+    constexpr unsigned K  = 1;
+    constexpr unsigned S  = 3;
+    constexpr unsigned R  = 3;
 #elif 0
     constexpr unsigned N  = 2;
     constexpr unsigned C  = 3;

driver/device_direct_convolution_1.cuh

@@ -26,13 +26,13 @@ void device_direct_convolution_1(
     constexpr auto out_desc         = OutDesc{};
     constexpr unsigned OutTileSizeH = 2;
     constexpr unsigned OutTileSizeW = 2;
-    constexpr unsigned NPerBlock    = 1;
+    constexpr unsigned NPerBlock    = 2;
     constexpr unsigned KPerBlock    = 16;
-    constexpr unsigned CPerBlock    = 4;
-    constexpr unsigned YPerBlock    = 4;
+    constexpr unsigned CPerBlock    = 2;
+    constexpr unsigned YPerBlock    = 2;
     constexpr unsigned XPerBlock    = 16;
 
-    constexpr unsigned NPerThread = 1;
+    constexpr unsigned NPerThread = 2;
     constexpr unsigned KPerThread = 4;
     constexpr unsigned CPerThread = 2;
 

src/include/gridwise_winograd_convolution.cuh

@@ -41,8 +41,8 @@ __global__ void gridwise_winograd_convolution(InGlobalDesc,
     constexpr unsigned HoPerBlock = OutTileSizeH * YPerBlock;
     constexpr unsigned WoPerBlock = OutTileSizeW * XPerBlock;
 
-    constexpr unsigned HiPerBlock = YPerBlock * OutTileSizeH + S - 1;
-    constexpr unsigned WiPerBlock = XPerBlock * OutTileSizeW + R - 1;
+    constexpr unsigned HiPerBlock = HoPerBlock + S - 1;
+    constexpr unsigned WiPerBlock = WoPerBlock + R - 1;
 
     constexpr unsigned InTileSizeH = OutTileSizeH + S - 1;
     constexpr unsigned InTileSizeW = OutTileSizeW + R - 1;
@@ -102,11 +102,8 @@ __global__ void gridwise_winograd_convolution(InGlobalDesc,
     constexpr auto wei_transform_block_desc =
         make_ConstantTensorDescriptor(Sequence<KPerBlock, CPerBlock, InTileSizeH, InTileSizeW>{});
 
-    constexpr unsigned in_transform_block_size  = in_transform_block_desc.GetElementSpace();
-    constexpr unsigned wei_transform_block_size = wei_transform_block_desc.GetElementSpace();
-
-    __shared__ TFloat p_in_transform_block[in_transform_block_size];
-    __shared__ TFloat p_wei_transform_block[wei_transform_block_size];
+    __shared__ TFloat p_in_transform_block[in_transform_block_desc.GetElementSpace()];
+    __shared__ TFloat p_wei_transform_block[wei_transform_block_desc.GetElementSpace()];
 
     // thread data
     constexpr auto in_transform_thread_block_desc =
@@ -126,11 +123,8 @@ __global__ void gridwise_winograd_convolution(InGlobalDesc,
     constexpr auto out_thread_global_desc =
         make_ConstantTensorDescriptor(out_thread_desc.GetLengths(), out_global_desc.GetStrides());
 
-    constexpr unsigned out_transform_thread_size = out_transform_thread_desc.GetElementSpace();
-    constexpr unsigned out_thread_size           = out_thread_desc.GetElementSpace();
-
-    TFloat p_out_transform_thread[out_transform_thread_size];
-    TFloat p_out_thread[out_thread_size];
+    TFloat p_out_transform_thread[out_transform_thread_desc.GetElementSpace()];
+    TFloat p_out_thread[out_thread_desc.GetElementSpace()];
 
 #if 0
     if(blockIdx.x == 0 && threadIdx.x == 0)

src/include/threadwise_tensor_op.cuh

@@ -116,10 +116,13 @@ __device__ void threadwise_4d_tensor_shift_down(Desc, TFloat* __restrict__ p, ID
 
     const unsigned did0_end =
         is_same<decltype(I0), IDim>::value ? desc.GetLength(I0) - shift : desc.GetLength(I0);
+
     const unsigned did1_end =
         is_same<decltype(I1), IDim>::value ? desc.GetLength(I1) - shift : desc.GetLength(I1);
+
     const unsigned did2_end =
         is_same<decltype(I2), IDim>::value ? desc.GetLength(I2) - shift : desc.GetLength(I2);
+
     const unsigned did3_end =
         is_same<decltype(I3), IDim>::value ? desc.GetLength(I3) - shift : desc.GetLength(I3);