Skip to content

GitLab

"vscode:/vscode.git/clone" did not exist on "b671cb092082bab9855d028802fccf703db03f16"
Commit b75abe7f authored by rocking's avatar rocking
Browse files

Refactor the base class. implement generic pooling in the future

parent 646f689c
Hide whitespace changes
Inline Side-by-side
Showing with 100 additions and 73 deletions
example/13_pool2d_fwd/pool2d_fwd_common.hpp
View file @ b75abe7f
...@@ -219,14 +219,13 @@ bool pool_test(bool do_verification, ...@@ -219,14 +219,13 @@ bool pool_test(bool do_verification,
static_cast<InDataType*>(in_device_buf.GetDeviceBuffer()), static_cast<InDataType*>(in_device_buf.GetDeviceBuffer()),
static_cast<OutDataType*>(out_device_buf.GetDeviceBuffer()), static_cast<OutDataType*>(out_device_buf.GetDeviceBuffer()),
static_cast<IndexDataType*>(out_indices_device_buf.GetDeviceBuffer()), static_cast<IndexDataType*>(out_indices_device_buf.GetDeviceBuffer()),
N, {N, C, Hi, Wi},
C, {Y, X},
std::array<ck::index_t, 2>{{Hi, Wi}}, {N, C, Ho, Wo},
std::array<ck::index_t, 2>{{Y, X}},
std::array<ck::index_t, 2>{{Ho, Wo}},
window_strides, window_strides,
input_left_pads, input_left_pads,
input_right_pads); input_right_pads,
{2, 3});
if(!pool.IsSupportedArgument(argument_ptr.get())) if(!pool.IsSupportedArgument(argument_ptr.get()))
{ {
......
example/48_pool3d_fwd/pool3d_fwd_common.hpp
View file @ b75abe7f
...@@ -32,7 +32,8 @@ static void pool3d_host_verify(const Tensor<InDataType>& in, ...@@ -32,7 +32,8 @@ static void pool3d_host_verify(const Tensor<InDataType>& in,
const std::array<ck::index_t, 3>& in_left_pads, const std::array<ck::index_t, 3>& in_left_pads,
const std::array<ck::index_t, 3>& /*in_right_pads*/) const std::array<ck::index_t, 3>& /*in_right_pads*/)
{ {
const int32_t reduceLength = window_spatial_lengths[0] * window_spatial_lengths[1]; const int32_t reduceLength =
window_spatial_lengths[0] * window_spatial_lengths[1] * window_spatial_lengths[2];
using ReduceOperation = typename ck::reduce_binary_operator<ReduceOpId>::opType; using ReduceOperation = typename ck::reduce_binary_operator<ReduceOpId>::opType;
...@@ -241,14 +242,13 @@ bool pool3d_test(bool do_verification, ...@@ -241,14 +242,13 @@ bool pool3d_test(bool do_verification,
static_cast<InDataType*>(in_device_buf.GetDeviceBuffer()), static_cast<InDataType*>(in_device_buf.GetDeviceBuffer()),
static_cast<OutDataType*>(out_device_buf.GetDeviceBuffer()), static_cast<OutDataType*>(out_device_buf.GetDeviceBuffer()),
static_cast<IndexDataType*>(out_indices_device_buf.GetDeviceBuffer()), static_cast<IndexDataType*>(out_indices_device_buf.GetDeviceBuffer()),
N, {N, C, Di, Hi, Wi},
C, {Z, Y, X},
std::array<ck::index_t, 3>{{Di, Hi, Wi}}, {N, C, Do, Ho, Wo},
std::array<ck::index_t, 3>{{Z, Y, X}},
std::array<ck::index_t, 3>{{Do, Ho, Wo}},
window_strides, window_strides,
input_left_pads, input_left_pads,
input_right_pads); input_right_pads,
{2, 3, 4});
if(!pool.IsSupportedArgument(argument_ptr.get())) if(!pool.IsSupportedArgument(argument_ptr.get()))
{ {
......
include/ck/tensor_operation/gpu/device/device_pool_fwd.hpp
View file @ b75abe7f
...@@ -13,21 +13,20 @@ namespace ck { ...@@ -13,21 +13,20 @@ namespace ck {
namespace tensor_operation { namespace tensor_operation {
namespace device { namespace device {
template <index_t NDimSpatial, ReduceTensorOp ReduceOpId> template <index_t InOutRank, index_t WindowRank, ReduceTensorOp ReduceOpId>
struct DevicePoolFwd : public BaseOperator struct DevicePoolFwd : public BaseOperator
{ {
virtual std::unique_ptr<BaseArgument> virtual std::unique_ptr<BaseArgument>
MakeArgumentPointer(const void* in_dev, MakeArgumentPointer(const void* p_in_dev,
void* out_dev, void* p_out_dev,
void* out_indices_dev, void* p_out_indices_dev,
ck::index_t N, std::array<ck::index_t, InOutRank> input_lengths,
ck::index_t C, std::array<ck::index_t, WindowRank> window_lengths,
std::array<ck::index_t, NDimSpatial> input_spatial_lengths, std::array<ck::index_t, InOutRank> output_lengths,
std::array<ck::index_t, NDimSpatial> window_spatial_lengths, std::array<ck::index_t, WindowRank> window_strides,
std::array<ck::index_t, NDimSpatial> output_spatial_lengths, std::array<ck::index_t, WindowRank> input_left_pads,
std::array<ck::index_t, NDimSpatial> window_strides, std::array<ck::index_t, WindowRank> input_right_pads,
std::array<ck::index_t, NDimSpatial> input_left_pads, std::array<ck::index_t, WindowRank> pooling_dims) = 0;
std::array<ck::index_t, NDimSpatial> input_right_pads) = 0;
virtual std::unique_ptr<BaseInvoker> MakeInvokerPointer() = 0; virtual std::unique_ptr<BaseInvoker> MakeInvokerPointer() = 0;
}; };
......
include/ck/tensor_operation/gpu/device/impl/device_pool2d_fwd_nhwc_nhwc.hpp
View file @ b75abe7f
...@@ -29,7 +29,7 @@ template <typename InDataType, ...@@ -29,7 +29,7 @@ template <typename InDataType,
ck::index_t ReduceMThreadSliceSize, ck::index_t ReduceMThreadSliceSize,
ck::index_t ReduceKThreadSliceSize, ck::index_t ReduceKThreadSliceSize,
ck::index_t InSrcOutDstVectorSize> ck::index_t InSrcOutDstVectorSize>
struct DevicePool2dFwd_Input_N_Hi_Wi_C_Output_N_Ho_Wo_C : public DevicePoolFwd<2, ReduceOpId> struct DevicePool2dFwd_Input_N_Hi_Wi_C_Output_N_Ho_Wo_C : public DevicePoolFwd<5, 2, ReduceOpId>
{ {
static constexpr auto I0 = Number<0>{}; static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{}; static constexpr auto I1 = Number<1>{};
...@@ -38,6 +38,9 @@ struct DevicePool2dFwd_Input_N_Hi_Wi_C_Output_N_Ho_Wo_C : public DevicePoolFwd<2 ...@@ -38,6 +38,9 @@ struct DevicePool2dFwd_Input_N_Hi_Wi_C_Output_N_Ho_Wo_C : public DevicePoolFwd<2
static constexpr auto I4 = Number<4>{}; static constexpr auto I4 = Number<4>{};
static constexpr auto I5 = Number<5>{}; static constexpr auto I5 = Number<5>{};
static constexpr index_t InOutRank = 5;
static constexpr index_t WindowRank = 2;
using IndexDataType = int32_t; using IndexDataType = int32_t;
using ReduceOperation = typename reduce_binary_operator<ReduceOpId>::opType; using ReduceOperation = typename reduce_binary_operator<ReduceOpId>::opType;
...@@ -57,14 +60,15 @@ struct DevicePool2dFwd_Input_N_Hi_Wi_C_Output_N_Ho_Wo_C : public DevicePoolFwd<2 ...@@ -57,14 +60,15 @@ struct DevicePool2dFwd_Input_N_Hi_Wi_C_Output_N_Ho_Wo_C : public DevicePoolFwd<2
static constexpr ck::index_t ReduceK_BlockTileSize = static constexpr ck::index_t ReduceK_BlockTileSize =
ReduceKThreadClusterSize * ReduceKThreadSliceSize; ReduceKThreadClusterSize * ReduceKThreadSliceSize;
static auto MakeABGridDescriptor_A_M_K_B_M(ck::index_t N, static auto
ck::index_t C, MakeABGridDescriptor_A_M_K_B_M(ck::index_t N,
std::array<ck::index_t, 2> input_spatial_lengths, ck::index_t C,
std::array<ck::index_t, 2> window_spatial_lengths, std::array<ck::index_t, WindowRank> input_spatial_lengths,
std::array<ck::index_t, 2> output_spatial_lengths, std::array<ck::index_t, WindowRank> window_spatial_lengths,
std::array<ck::index_t, 2> window_strides, std::array<ck::index_t, WindowRank> output_spatial_lengths,
std::array<ck::index_t, 2> input_left_pads, std::array<ck::index_t, WindowRank> window_strides,
std::array<ck::index_t, 2> input_right_pads) std::array<ck::index_t, WindowRank> input_left_pads,
std::array<ck::index_t, WindowRank> input_right_pads)
{ {
const index_t Hi = input_spatial_lengths[0]; const index_t Hi = input_spatial_lengths[0];
const index_t Wi = input_spatial_lengths[1]; const index_t Wi = input_spatial_lengths[1];
...@@ -155,12 +159,12 @@ struct DevicePool2dFwd_Input_N_Hi_Wi_C_Output_N_Ho_Wo_C : public DevicePoolFwd<2 ...@@ -155,12 +159,12 @@ struct DevicePool2dFwd_Input_N_Hi_Wi_C_Output_N_Ho_Wo_C : public DevicePoolFwd<2
int* p_out_indices_dev, int* p_out_indices_dev,
ck::index_t N, ck::index_t N,
ck::index_t C, ck::index_t C,
std::array<ck::index_t, 2>& input_spatial_lengths, std::array<ck::index_t, WindowRank>& input_spatial_lengths,
std::array<ck::index_t, 2>& window_spatial_lengths, std::array<ck::index_t, WindowRank>& window_spatial_lengths,
std::array<ck::index_t, 2>& output_spatial_lengths, std::array<ck::index_t, WindowRank>& output_spatial_lengths,
std::array<ck::index_t, 2>& window_strides, std::array<ck::index_t, WindowRank>& window_strides,
std::array<ck::index_t, 2>& input_left_pads, std::array<ck::index_t, WindowRank>& input_left_pads,
std::array<ck::index_t, 2>& input_right_pads) std::array<ck::index_t, WindowRank>& input_right_pads)
: p_in_dev_{p_in_dev}, : p_in_dev_{p_in_dev},
p_out_dev_{p_out_dev}, p_out_dev_{p_out_dev},
p_out_indices_dev_{p_out_indices_dev}, p_out_indices_dev_{p_out_indices_dev},
...@@ -280,22 +284,31 @@ struct DevicePool2dFwd_Input_N_Hi_Wi_C_Output_N_Ho_Wo_C : public DevicePoolFwd<2 ...@@ -280,22 +284,31 @@ struct DevicePool2dFwd_Input_N_Hi_Wi_C_Output_N_Ho_Wo_C : public DevicePoolFwd<2
MakeArgumentPointer(const void* p_in_dev, MakeArgumentPointer(const void* p_in_dev,
void* p_out_dev, void* p_out_dev,
void* p_out_indices_dev, void* p_out_indices_dev,
ck::index_t N, std::array<ck::index_t, InOutRank> input_lengths,
ck::index_t C, std::array<ck::index_t, WindowRank> window_lengths,
std::array<ck::index_t, 2> input_spatial_lengths, std::array<ck::index_t, InOutRank> output_lengths,
std::array<ck::index_t, 2> window_spatial_lengths, std::array<ck::index_t, WindowRank> window_strides,
std::array<ck::index_t, 2> output_spatial_lengths, std::array<ck::index_t, WindowRank> input_left_pads,
std::array<ck::index_t, 2> window_strides, std::array<ck::index_t, WindowRank> input_right_pads,
std::array<ck::index_t, 2> input_left_pads, std::array<ck::index_t, WindowRank>) override
std::array<ck::index_t, 2> input_right_pads) override
{ {
index_t N = input_lengths[0];
index_t C = input_lengths[1];
index_t Hi = input_lengths[2];
index_t Wi = input_lengths[3];
index_t Ho = output_lengths[2];
index_t Wo = output_lengths[3];
std::array<ck::index_t, WindowRank> input_spatial_lengths = {Hi, Wi};
std::array<ck::index_t, WindowRank> output_spatial_lengths = {Ho, Wo};
return std::make_unique<Argument>(static_cast<const InDataType*>(p_in_dev), return std::make_unique<Argument>(static_cast<const InDataType*>(p_in_dev),
static_cast<OutDataType*>(p_out_dev), static_cast<OutDataType*>(p_out_dev),
static_cast<int*>(p_out_indices_dev), static_cast<int*>(p_out_indices_dev),
N, N,
C, C,
input_spatial_lengths, input_spatial_lengths,
window_spatial_lengths, window_lengths,
output_spatial_lengths, output_spatial_lengths,
window_strides, window_strides,
input_left_pads, input_left_pads,
......
include/ck/tensor_operation/gpu/device/impl/device_pool3d_fwd_ndhwc_ndhwc.hpp
View file @ b75abe7f
...@@ -29,7 +29,8 @@ template <typename InDataType, ...@@ -29,7 +29,8 @@ template <typename InDataType,
ck::index_t MThreadSliceSize, ck::index_t MThreadSliceSize,
ck::index_t KThreadSliceSize, ck::index_t KThreadSliceSize,
ck::index_t InSrcOutDstVectorSize> ck::index_t InSrcOutDstVectorSize>
struct DevicePool3dFwd_Input_N_Di_Hi_Wi_C_Output_N_Do_Ho_Wo_C : public DevicePoolFwd<3, ReduceOpId> struct DevicePool3dFwd_Input_N_Di_Hi_Wi_C_Output_N_Do_Ho_Wo_C
: public DevicePoolFwd<5, 3, ReduceOpId>
{ {
static constexpr auto I0 = Number<0>{}; static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{}; static constexpr auto I1 = Number<1>{};
...@@ -38,6 +39,9 @@ struct DevicePool3dFwd_Input_N_Di_Hi_Wi_C_Output_N_Do_Ho_Wo_C : public DevicePoo ...@@ -38,6 +39,9 @@ struct DevicePool3dFwd_Input_N_Di_Hi_Wi_C_Output_N_Do_Ho_Wo_C : public DevicePoo
static constexpr auto I4 = Number<4>{}; static constexpr auto I4 = Number<4>{};
static constexpr auto I5 = Number<5>{}; static constexpr auto I5 = Number<5>{};
static constexpr index_t InOutRank = 5;
static constexpr index_t WindowRank = 3;
using IndexDataType = int32_t; using IndexDataType = int32_t;
using ReduceOperation = typename reduce_binary_operator<ReduceOpId>::opType; using ReduceOperation = typename reduce_binary_operator<ReduceOpId>::opType;
...@@ -55,14 +59,15 @@ struct DevicePool3dFwd_Input_N_Di_Hi_Wi_C_Output_N_Do_Ho_Wo_C : public DevicePoo ...@@ -55,14 +59,15 @@ struct DevicePool3dFwd_Input_N_Di_Hi_Wi_C_Output_N_Do_Ho_Wo_C : public DevicePoo
static constexpr ck::index_t M_BlockTileSize = MThreadClusterSize * MThreadSliceSize; static constexpr ck::index_t M_BlockTileSize = MThreadClusterSize * MThreadSliceSize;
static constexpr ck::index_t K_BlockTileSize = KThreadClusterSize * KThreadSliceSize; static constexpr ck::index_t K_BlockTileSize = KThreadClusterSize * KThreadSliceSize;
static auto MakeABGridDescriptor_A_M_K_B_M(ck::index_t N, static auto
ck::index_t C, MakeABGridDescriptor_A_M_K_B_M(ck::index_t N,
std::array<ck::index_t, 3> input_spatial_lengths, ck::index_t C,
std::array<ck::index_t, 3> window_spatial_lengths, std::array<ck::index_t, WindowRank> input_spatial_lengths,
std::array<ck::index_t, 3> output_spatial_lengths, std::array<ck::index_t, WindowRank> window_spatial_lengths,
std::array<ck::index_t, 3> window_strides, std::array<ck::index_t, WindowRank> output_spatial_lengths,
std::array<ck::index_t, 3> input_left_pads, std::array<ck::index_t, WindowRank> window_strides,
std::array<ck::index_t, 3> input_right_pads) std::array<ck::index_t, WindowRank> input_left_pads,
std::array<ck::index_t, WindowRank> input_right_pads)
{ {
const index_t Di = input_spatial_lengths[0]; const index_t Di = input_spatial_lengths[0];
const index_t Hi = input_spatial_lengths[1]; const index_t Hi = input_spatial_lengths[1];
...@@ -161,12 +166,12 @@ struct DevicePool3dFwd_Input_N_Di_Hi_Wi_C_Output_N_Do_Ho_Wo_C : public DevicePoo ...@@ -161,12 +166,12 @@ struct DevicePool3dFwd_Input_N_Di_Hi_Wi_C_Output_N_Do_Ho_Wo_C : public DevicePoo
int* p_out_indices_dev, int* p_out_indices_dev,
ck::index_t N, ck::index_t N,
ck::index_t C, ck::index_t C,
std::array<ck::index_t, 3>& input_spatial_lengths, std::array<ck::index_t, WindowRank>& input_spatial_lengths,
std::array<ck::index_t, 3>& window_spatial_lengths, std::array<ck::index_t, WindowRank>& window_spatial_lengths,
std::array<ck::index_t, 3>& output_spatial_lengths, std::array<ck::index_t, WindowRank>& output_spatial_lengths,
std::array<ck::index_t, 3>& window_strides, std::array<ck::index_t, WindowRank>& window_strides,
std::array<ck::index_t, 3>& input_left_pads, std::array<ck::index_t, WindowRank>& input_left_pads,
std::array<ck::index_t, 3>& input_right_pads) std::array<ck::index_t, WindowRank>& input_right_pads)
: p_in_dev_{p_in_dev}, : p_in_dev_{p_in_dev},
p_out_dev_{p_out_dev}, p_out_dev_{p_out_dev},
p_out_indices_dev_{p_out_indices_dev}, p_out_indices_dev_{p_out_indices_dev},
...@@ -285,22 +290,33 @@ struct DevicePool3dFwd_Input_N_Di_Hi_Wi_C_Output_N_Do_Ho_Wo_C : public DevicePoo ...@@ -285,22 +290,33 @@ struct DevicePool3dFwd_Input_N_Di_Hi_Wi_C_Output_N_Do_Ho_Wo_C : public DevicePoo
MakeArgumentPointer(const void* p_in_dev, MakeArgumentPointer(const void* p_in_dev,
void* p_out_dev, void* p_out_dev,
void* p_out_indices_dev, void* p_out_indices_dev,
ck::index_t N, std::array<ck::index_t, InOutRank> input_lengths,
ck::index_t C, std::array<ck::index_t, WindowRank> window_lengths,
std::array<ck::index_t, 3> input_spatial_lengths, std::array<ck::index_t, InOutRank> output_lengths,
std::array<ck::index_t, 3> window_spatial_lengths, std::array<ck::index_t, WindowRank> window_strides,
std::array<ck::index_t, 3> output_spatial_lengths, std::array<ck::index_t, WindowRank> input_left_pads,
std::array<ck::index_t, 3> window_strides, std::array<ck::index_t, WindowRank> input_right_pads,
std::array<ck::index_t, 3> input_left_pads, std::array<ck::index_t, WindowRank>) override
std::array<ck::index_t, 3> input_right_pads) override
{ {
index_t N = input_lengths[0];
index_t C = input_lengths[1];
index_t Di = input_lengths[2];
index_t Hi = input_lengths[3];
index_t Wi = input_lengths[4];
index_t Do = output_lengths[2];
index_t Ho = output_lengths[3];
index_t Wo = output_lengths[4];
std::array<ck::index_t, WindowRank> input_spatial_lengths = {Di, Hi, Wi};
std::array<ck::index_t, WindowRank> output_spatial_lengths = {Do, Ho, Wo};
return std::make_unique<Argument>(static_cast<const InDataType*>(p_in_dev), return std::make_unique<Argument>(static_cast<const InDataType*>(p_in_dev),
static_cast<OutDataType*>(p_out_dev), static_cast<OutDataType*>(p_out_dev),
static_cast<int*>(p_out_indices_dev), static_cast<int*>(p_out_indices_dev),
N, N,
C, C,
input_spatial_lengths, input_spatial_lengths,
window_spatial_lengths, window_lengths,
output_spatial_lengths, output_spatial_lengths,
window_strides, window_strides,
input_left_pads, input_left_pads,
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment