[ROCm/composable_kernel commit: a21b0d27a5]
This commit is contained in:
Chao Liu
2018-11-20 10:51:28 -06:00
parent cb8367c0eb
commit 69a23647d8
5 changed files with 50 additions and 50 deletions

View File

@@ -20,12 +20,12 @@ template <class TFloat,
unsigned NBlockOpLen3,
unsigned BlockSize,
unsigned GridSize>
__global__ void gridwise_convolution(InGlobalDesc,
TFloat* const __restrict__ p_in_global,
WeiGlobalDesc,
TFloat* const __restrict__ p_wei_global,
OutGlobalDesc,
TFloat* __restrict__ p_out_global)
__global__ void gridwise_direct_convolution_1(InGlobalDesc,
TFloat* const __restrict__ p_in_global,
WeiGlobalDesc,
TFloat* const __restrict__ p_wei_global,
OutGlobalDesc,
TFloat* __restrict__ p_out_global)
{
constexpr auto I0 = Index<0>{};
constexpr auto I1 = Index<1>{};

View File

@@ -25,12 +25,12 @@ template <class TFloat,
unsigned NBlockOpLen3,
unsigned BlockSize,
unsigned GridSize>
__global__ void gridwise_convolution(InGlobalDesc,
TFloat* const __restrict__ p_in_global,
WeiGlobalDesc,
TFloat* const __restrict__ p_wei_global,
OutGlobalDesc,
TFloat* __restrict__ p_out_global)
__global__ void gridwise_direct_convolution_2(InGlobalDesc,
TFloat* const __restrict__ p_in_global,
WeiGlobalDesc,
TFloat* const __restrict__ p_wei_global,
OutGlobalDesc,
TFloat* __restrict__ p_out_global)
{
constexpr auto I0 = Index<0>{};
constexpr auto I1 = Index<1>{};