Skip to content

GitLab

Commit 316fcc3f authored by Jing Zhang's avatar Jing Zhang
Browse files

rename

parent 59328f3a
Hide whitespace changes
Inline Side-by-side
Showing with 20 additions and 20 deletions
composable_kernel/include/driver/driver_dynamic_convolution_forward_implicit_gemm_v5r1_nchw_kcyx_nkhw.hpp
View file @ 316fcc3f
...@@ -136,13 +136,13 @@ struct DriverDynamicConvolutionForwardImplicitGemm_v5r1_nchw_kcyx_nkhw_pad ...@@ -136,13 +136,13 @@ struct DriverDynamicConvolutionForwardImplicitGemm_v5r1_nchw_kcyx_nkhw_pad
} }
// hack to control index calculation when iterating over a_k_m_global tensor // hack to control index calculation when iterating over a_k_m_global tensor
constexpr auto a_k_m_global_iterator_hacks = constexpr auto a_e_k_global_iterator_hacks =
make_tuple(make_tuple(Sequence<0, 0, 0>{}, Sequence<0, 0, 0>{}), make_tuple(make_tuple(Sequence<0, 0, 0>{}, Sequence<0, 0, 0>{}),
make_tuple(Sequence<0, 0, 0>{}, Sequence<0, 0, 0>{})); make_tuple(Sequence<0, 0, 0>{}, Sequence<0, 0, 0>{}));
constexpr auto a_k_m_global_move_slice_window_iterator_hack = Sequence<0, 0, 0>{}; constexpr auto a_e_k_global_move_slice_window_iterator_hack = Sequence<0, 0, 0>{};
constexpr auto b_k_n_global_iterator_hacks = constexpr auto b_e_n_ho_wo_global_iterator_hacks =
make_tuple(make_tuple(Sequence<0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0>{}, make_tuple(make_tuple(Sequence<0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0>{},
Sequence<0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0>{}, Sequence<0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0>{},
Sequence<0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0>{}, Sequence<0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0>{},
...@@ -152,12 +152,12 @@ struct DriverDynamicConvolutionForwardImplicitGemm_v5r1_nchw_kcyx_nkhw_pad ...@@ -152,12 +152,12 @@ struct DriverDynamicConvolutionForwardImplicitGemm_v5r1_nchw_kcyx_nkhw_pad
Sequence<0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0>{}, Sequence<0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0>{},
Sequence<0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0>{})); Sequence<0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0>{}));
constexpr auto b_k_n_global_move_slice_window_iterator_hack = constexpr auto b_e_n_ho_wo_global_move_slice_window_iterator_hack =
Sequence<0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0>{}; Sequence<0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0>{};
// hack to control index calculation when iterating over c_m0_m1_n0_n1_global tensor // hack to control index calculation when iterating over c_m0_m1_n0_n1_global tensor
// hack for NKHW format // hack for NKHW format
constexpr auto c_k_n_h_w_global_tensor_iterator_hacks = constexpr auto c_k_n_ho_wo_global_tensor_iterator_hacks =
make_tuple(make_tuple(Sequence<0, 1, 0, 0, 0>{}, make_tuple(make_tuple(Sequence<0, 1, 0, 0, 0>{},
Sequence<0, 0, 0, 0, 0>{}, Sequence<0, 0, 0, 0, 0>{},
Sequence<0, 0, 0, 0, 0>{}, Sequence<0, 0, 0, 0, 0>{},
...@@ -202,11 +202,11 @@ struct DriverDynamicConvolutionForwardImplicitGemm_v5r1_nchw_kcyx_nkhw_pad ...@@ -202,11 +202,11 @@ struct DriverDynamicConvolutionForwardImplicitGemm_v5r1_nchw_kcyx_nkhw_pad
Sequence<0, 2, 3, 1>, Sequence<0, 2, 3, 1>,
0, 0,
CThreadTransferDstScalarPerVector_W, CThreadTransferDstScalarPerVector_W,
decltype(a_k_m_global_iterator_hacks), decltype(a_e_k_global_iterator_hacks),
decltype(b_k_n_global_iterator_hacks), decltype(b_e_n_ho_wo_global_iterator_hacks),
decltype(c_k_n_h_w_global_tensor_iterator_hacks), decltype(c_k_n_ho_wo_global_tensor_iterator_hacks),
decltype(a_k_m_global_move_slice_window_iterator_hack), decltype(a_e_k_global_move_slice_window_iterator_hack),
decltype(b_k_n_global_move_slice_window_iterator_hack)>; decltype(b_e_n_ho_wo_global_move_slice_window_iterator_hack)>;
const auto GridSize = (K / KPerBlock) * (Ho / HoPerBlock) * (Wo / WoPerBlock) * N; const auto GridSize = (K / KPerBlock) * (Ho / HoPerBlock) * (Wo / WoPerBlock) * N;
......
composable_kernel/include/driver/driver_dynamic_convolution_forward_implicit_gemm_v5r1_nchw_kcyx_nkhw_outpad.hpp
View file @ 316fcc3f
...@@ -149,13 +149,13 @@ struct DriverDynamicConvolutionForwardImplicitGemm_v5r1_nchw_kcyx_nkhw_outpad ...@@ -149,13 +149,13 @@ struct DriverDynamicConvolutionForwardImplicitGemm_v5r1_nchw_kcyx_nkhw_outpad
} }
// hack to control index calculation when iterating over a_k_m_global tensor // hack to control index calculation when iterating over a_k_m_global tensor
constexpr auto a_k_m_global_iterator_hacks = constexpr auto a_e_k_global_iterator_hacks =
make_tuple(make_tuple(Sequence<0, 0, 0>{}, Sequence<0, 0, 0>{}), make_tuple(make_tuple(Sequence<0, 0, 0>{}, Sequence<0, 0, 0>{}),
make_tuple(Sequence<0, 0, 0>{}, Sequence<0, 0, 0>{})); make_tuple(Sequence<0, 0, 0>{}, Sequence<0, 0, 0>{}));
constexpr auto a_k_m_global_move_slice_window_iterator_hack = Sequence<0, 0, 0>{}; constexpr auto a_e_k_global_move_slice_window_iterator_hack = Sequence<0, 0, 0>{};
constexpr auto b_k_n_global_iterator_hacks = constexpr auto b_e_n_ho_wo_global_iterator_hacks =
make_tuple(make_tuple(Sequence<0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0>{}, make_tuple(make_tuple(Sequence<0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0>{},
Sequence<0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0>{}, Sequence<0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0>{},
Sequence<0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0>{}, Sequence<0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0>{},
...@@ -165,12 +165,12 @@ struct DriverDynamicConvolutionForwardImplicitGemm_v5r1_nchw_kcyx_nkhw_outpad ...@@ -165,12 +165,12 @@ struct DriverDynamicConvolutionForwardImplicitGemm_v5r1_nchw_kcyx_nkhw_outpad
Sequence<0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0>{}, Sequence<0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0>{},
Sequence<0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0>{})); Sequence<0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0>{}));
constexpr auto b_k_n_global_move_slice_window_iterator_hack = constexpr auto b_e_n_ho_wo_global_move_slice_window_iterator_hack =
Sequence<0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0>{}; Sequence<0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0>{};
// hack to control index calculation when iterating over c_m0_m1_n0_n1_global tensor // hack to control index calculation when iterating over c_m0_m1_n0_n1_global tensor
// hack for NKHW format // hack for NKHW format
constexpr auto c_k_n_h_w_global_tensor_iterator_hacks = constexpr auto c_k_n_ho_wo_global_tensor_iterator_hacks =
make_tuple(make_tuple(Sequence<0, 1, 0, 0, 0>{}, make_tuple(make_tuple(Sequence<0, 1, 0, 0, 0>{},
Sequence<0, 0, 0, 0, 0>{}, Sequence<0, 0, 0, 0, 0>{},
Sequence<0, 0, 0, 0, 0>{}, Sequence<0, 0, 0, 0, 0>{},
...@@ -215,11 +215,11 @@ struct DriverDynamicConvolutionForwardImplicitGemm_v5r1_nchw_kcyx_nkhw_outpad ...@@ -215,11 +215,11 @@ struct DriverDynamicConvolutionForwardImplicitGemm_v5r1_nchw_kcyx_nkhw_outpad
Sequence<0, 2, 3, 1>, Sequence<0, 2, 3, 1>,
0, 0,
CThreadTransferDstScalarPerVector_W, CThreadTransferDstScalarPerVector_W,
decltype(a_k_m_global_iterator_hacks), decltype(a_e_k_global_iterator_hacks),
decltype(b_k_n_global_iterator_hacks), decltype(b_e_n_ho_wo_global_iterator_hacks),
decltype(c_k_n_h_w_global_tensor_iterator_hacks), decltype(c_k_n_ho_wo_global_tensor_iterator_hacks),
decltype(a_k_m_global_move_slice_window_iterator_hack), decltype(a_e_k_global_move_slice_window_iterator_hack),
decltype(b_k_n_global_move_slice_window_iterator_hack)>; decltype(b_e_n_ho_wo_global_move_slice_window_iterator_hack)>;
const auto GridSize = (K / KPerBlock) * (Hop / HoPerBlock) * (Wop / WoPerBlock) * N; const auto GridSize = (K / KPerBlock) * (Hop / HoPerBlock) * (Wop / WoPerBlock) * N;
......
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