Skip to content

GitLab

Commit 666bdad1 authored by root's avatar root
Browse files

clean

parent f744524e
Hide whitespace changes
Inline Side-by-side
Showing with 240 additions and 242 deletions
composable_kernel/include/driver/driver_dynamic_convolution_forward_implicit_gemm_v5r1_nchw_kcyx_nkhw.hpp
View file @ 666bdad1
......@@ -13,13 +13,13 @@ template <index_t BlockSize,
typename Float,
typename AccFloat,
index_t KPerBlock,
index_t HPerBlock,
index_t WPerBlock,
index_t CYXPerBlock,
index_t HoPerBlock,
index_t WoPerBlock,
index_t EPerBlock,
index_t KPerThread,
index_t HPerThread,
index_t WPerThread,
index_t CYXPerThread,
index_t EPerThread,
typename GemmABlockTransferThreadSliceLengths_GemmK_GemmM,
typename GemmABlockTransferThreadClusterLengths_GemmK_GemmM,
index_t GemmABlockTransferSrcScalarPerVector_GemmK,
......@@ -123,10 +123,10 @@ struct DriverDynamicConvolutionForwardImplicitGemm_v5r1_nchw_kcyx_nkhw_pad
make_tuple(Sequence<1>{}, Sequence<0>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}));
const auto CYX = C * Y * X;
const auto E = C * Y * X;
if(!(K % KPerBlock == 0 && Ho % HPerBlock == 0 && Wo % WPerBlock == 0 &&
CYX % CYXPerBlock == 0))
if(!(K % KPerBlock == 0 && Ho % HoPerBlock == 0 && Wo % WoPerBlock == 0 &&
E % EPerBlock == 0))
{
throw std::runtime_error("wrong! GEMM size no divisible");
}
......@@ -165,7 +165,7 @@ struct DriverDynamicConvolutionForwardImplicitGemm_v5r1_nchw_kcyx_nkhw_pad
#if 1
// GEMM
using gridwise_gemm = GridwiseDynamicGemm_km_kn_mn_v3<
using gridwise_gemm = GridwiseDynamicGemm_km_kn_mn_v2<
BlockSize,
Float,
AccFloat,
......@@ -174,13 +174,13 @@ struct DriverDynamicConvolutionForwardImplicitGemm_v5r1_nchw_kcyx_nkhw_pad
decltype(in_gemmk_n_ho_wo_global_desc),
decltype(out_gemmm_n_ho_wo_global_desc),
KPerBlock,
HPerBlock,
WPerBlock,
CYXPerBlock,
HoPerBlock,
WoPerBlock,
EPerBlock,
KPerThread,
HPerThread,
WPerThread,
CYXPerThread,
EPerThread,
GemmABlockTransferThreadSliceLengths_GemmK_GemmM,
GemmABlockTransferThreadClusterLengths_GemmK_GemmM,
Sequence<1, 0>,
......@@ -205,11 +205,11 @@ struct DriverDynamicConvolutionForwardImplicitGemm_v5r1_nchw_kcyx_nkhw_pad
decltype(a_k_m_global_move_slice_window_iterator_hack),
decltype(b_k_n_global_move_slice_window_iterator_hack)>;
const auto GridSize = (K / KPerBlock) * (Ho / HPerBlock) * (Wo / WPerBlock) * N;
const auto GridSize = (K / KPerBlock) * (Ho / HoPerBlock) * (Wo / WoPerBlock) * N;
const bool has_main_k_block_loop = (CYX + CYXPerBlock) / (2 * CYXPerBlock) > 1;
const bool has_main_k_block_loop = (E + EPerBlock) / (2 * EPerBlock) > 1;
const bool has_double_tail_k_block_loop = (CYX / CYXPerBlock) % 2 == 0;
const bool has_double_tail_k_block_loop = (E / EPerBlock) % 2 == 0;
index_t nrepeat = 100;
......@@ -225,6 +225,7 @@ struct DriverDynamicConvolutionForwardImplicitGemm_v5r1_nchw_kcyx_nkhw_pad
for(index_t j = 0; j < nrepeat; ++j)
{
#if 0
{
const auto kernel =
run_gridwise_operation<gridwise_gemm,
......@@ -251,7 +252,7 @@ struct DriverDynamicConvolutionForwardImplicitGemm_v5r1_nchw_kcyx_nkhw_pad
integral_constant<bool, true>{},
integral_constant<bool, false>{});
}
#if 0
#else
if(has_main_k_block_loop && has_double_tail_k_block_loop)
{
const auto kernel =
......
composable_kernel/include/tensor_operation/gridwise_dynamic_gemm_v2.hpp
View file @ 666bdad1
......@@ -21,11 +21,11 @@ template <index_t BlockSize,
index_t KPerBlock,
index_t HPerBlock,
index_t WPerBlock,
index_t CYXPerBlock,
index_t EPerBlock,
index_t KPerThread,
index_t HPerThread,
index_t WPerThread,
index_t CYXPerThread,
index_t EPerThread,
typename ABlockTransferThreadSliceLengths_K_M,
typename ABlockTransferThreadClusterLengths_K_M,
typename ABlockTransferThreadClusterArrangeOrder,
......@@ -57,20 +57,20 @@ struct GridwiseDynamicGemm_km_kn_mn_v2
// A matrix in LDS memory, dst of blockwise copy
// be careful of LDS alignment
constexpr auto a_cyx_k_block_desc = make_dynamic_naive_tensor_descriptor_aligned_v2(
make_tuple(Number<CYXPerBlock>{}, Number<KPerBlock>{}), max_lds_align);
constexpr auto a_e_k_block_desc = make_dynamic_naive_tensor_descriptor_aligned_v2(
make_tuple(Number<EPerBlock>{}, Number<KPerBlock>{}), max_lds_align);
// LDS allocation for A and B: be careful of alignment
constexpr auto a_block_space_size =
math::integer_least_multiple(a_cyx_k_block_desc.GetElementSpaceSize(), max_lds_align);
math::integer_least_multiple(a_e_k_block_desc.GetElementSpaceSize(), max_lds_align);
return 2 * (a_block_space_size) * sizeof(Float);
}
template <bool HasMainKBlockLoop, bool HasDoubleTailKBlockLoop>
__device__ void Run(const AGlobalDesc& a_cyx_k_global_desc,
__device__ void Run(const AGlobalDesc& a_e_k_global_desc,
const Float* __restrict__ p_a_global,
const BGlobalDesc& b_cyx_n_h_w_global_desc,
const BGlobalDesc& b_e_n_h_w_global_desc,
const Float* __restrict__ p_b_global,
const CGlobalDesc& c_k_n_h_w_global_desc,
Float* __restrict__ p_c_global,
......@@ -83,15 +83,15 @@ struct GridwiseDynamicGemm_km_kn_mn_v2
constexpr auto I2 = Number<2>{};
constexpr auto I3 = Number<3>{};
const auto CYX = a_cyx_k_global_desc.GetLength(I0);
const auto K = a_cyx_k_global_desc.GetLength(I1);
const auto E = a_e_k_global_desc.GetLength(I0);
const auto K = a_e_k_global_desc.GetLength(I1);
const auto N = b_cyx_n_h_w_global_desc.GetLength(I1);
const auto H = b_cyx_n_h_w_global_desc.GetLength(I2);
const auto W = b_cyx_n_h_w_global_desc.GetLength(I3);
const auto N = b_e_n_h_w_global_desc.GetLength(I1);
const auto H = b_e_n_h_w_global_desc.GetLength(I2);
const auto W = b_e_n_h_w_global_desc.GetLength(I3);
// divide block work by [M, N]
#if 1
#if 0
const auto k_block_work_num = K / Number<KPerBlock>{};
const auto h_block_work_num = H / Number<HPerBlock>{};
const auto w_block_work_num = W / Number<WPerBlock>{};
......@@ -99,6 +99,8 @@ struct GridwiseDynamicGemm_km_kn_mn_v2
const index_t k_block_work_id = get_block_1d_id() / hw_block_work_num;
const index_t hw_block_work_id = get_block_1d_id() - k_block_work_id * hw_block_work_num;
const index_t h_block_work_id = hw_block_work_id / w_block_work_num;
const index_t w_block_work_id = hw_block_work_id - h_block_work_id * w_block_work_num;
#else
// Hack: this force result into SGPR
......@@ -110,10 +112,10 @@ struct GridwiseDynamicGemm_km_kn_mn_v2
const index_t k_block_work_id =
__builtin_amdgcn_readfirstlane(get_block_1d_id() / hw_block_work_num);
const index_t hw_block_work_id = get_block_1d_id() - k_block_work_id * hw_block_work_num;
#endif
const index_t h_block_work_id = hw_block_work_id / w_block_work_num;
const index_t h_block_work_id =
__builtin_amdgcn_readfirstlane(hw_block_work_id / w_block_work_num);
const index_t w_block_work_id = hw_block_work_id - h_block_work_id * w_block_work_num;
#endif
// lds max alignment
constexpr auto max_lds_align =
......@@ -121,14 +123,13 @@ struct GridwiseDynamicGemm_km_kn_mn_v2
// A matrix in LDS memory, dst of blockwise copy
// be careful of LDS alignment
constexpr auto a_cyx_k_block_desc = make_dynamic_naive_tensor_descriptor_aligned_v2(
make_tuple(Number<CYXPerBlock>{}, Number<KPerBlock>{}), max_lds_align);
constexpr auto a_e_k_block_desc = make_dynamic_naive_tensor_descriptor_aligned_v2(
make_tuple(Number<EPerBlock>{}, Number<KPerBlock>{}), max_lds_align);
// B matrix in LDS memory, dst of blockwise copy
// be careful of LDS alignment
constexpr auto b_cyx_n_h_w_block_desc =
make_dynamic_naive_tensor_descriptor_packed_v2(make_tuple(
Number<CYXPerBlock>{}, Number<1>{}, Number<HPerBlock>{}, Number<WPerBlock>{}));
constexpr auto b_e_n_h_w_block_desc = make_dynamic_naive_tensor_descriptor_packed_v2(
make_tuple(Number<EPerBlock>{}, Number<1>{}, Number<HPerBlock>{}, Number<WPerBlock>{}));
// c_thread_mtx definition: this is a mess
// TODO:: more elegent way of defining c_thread_mtx
......@@ -136,18 +137,17 @@ struct GridwiseDynamicGemm_km_kn_mn_v2
make_dynamic_naive_tensor_descriptor_packed_v2(make_tuple(
Number<KPerThread>{}, Number<1>{}, Number<HPerThread>{}, Number<WPerThread>{}));
const auto blockwise_gemm =
BlockwiseGemm_km_kn_m0m1n0n1_v3<BlockSize,
decltype(a_cyx_k_block_desc),
decltype(b_cyx_n_h_w_block_desc),
decltype(c_k_n_h_w_thread_desc),
KPerThread, // KPerThreadSubC
HPerThread, // HPerThreadSubC
WPerThread, // WPerThreadSubC
CYXPerThread, // CYXPerThreadLoop
1, // ThreadGemmADataPerRead_K
1 // ThreadGemmBDataPerRead_W
>{};
const auto blockwise_gemm = BlockwiseGemm_km_kn_m0m1n0n1_v3<BlockSize,
decltype(a_e_k_block_desc),
decltype(b_e_n_h_w_block_desc),
decltype(c_k_n_h_w_thread_desc),
KPerThread, // KPerThreadSubC
HPerThread, // HPerThreadSubC
WPerThread, // WPerThreadSubC
EPerThread, // EPerThreadLoop
1, // ThreadGemmADataPerRead_K
1 // ThreadGemmBDataPerRead_W
>{};
auto c_thread_mtx_index = blockwise_gemm.GetBeginOfThreadMatrixC(get_thread_local_1d_id());
......@@ -166,14 +166,14 @@ struct GridwiseDynamicGemm_km_kn_mn_v2
auto a_blockwise_copy =
BlockwiseDynamicTensorSliceTransfer_v4<BlockSize,
InMemoryDataOperation::Set,
Sequence<CYXPerBlock, KPerBlock>,
Sequence<EPerBlock, KPerBlock>,
ABlockTransferThreadSliceLengths_K_M,
ABlockTransferThreadClusterLengths_K_M,
ABlockTransferThreadClusterArrangeOrder,
Float,
Float,
decltype(a_cyx_k_global_desc),
decltype(a_cyx_k_block_desc),
decltype(a_e_k_global_desc),
decltype(a_e_k_block_desc),
ABlockTransferSrcAccessOrder,
Sequence<0, 1>,
ABlockTransferSrcVectorDim,
......@@ -186,21 +186,21 @@ struct GridwiseDynamicGemm_km_kn_mn_v2
1,
AThreadTransferSrcResetCoordinateAfterRun,
true>(
a_cyx_k_global_desc,
a_e_k_global_desc,
make_multi_index(0, k_block_data_on_global),
a_cyx_k_block_desc,
a_e_k_block_desc,
make_multi_index(0, 0));
constexpr auto b_cyx_n_h_w_thread_desc =
constexpr auto b_e_n_h_w_thread_desc =
make_dynamic_naive_tensor_descriptor_packed_v2(make_tuple(
Number<CYXPerBlock>{}, Number<1>{}, Number<HPerThread>{}, Number<WPerThread>{}));
Number<EPerBlock>{}, Number<1>{}, Number<HPerThread>{}, Number<WPerThread>{}));
auto b_threadwise_transfer = ThreadwiseDynamicTensorSliceTransfer_v2<
Float,
Float,
decltype(b_cyx_n_h_w_global_desc),
decltype(b_cyx_n_h_w_thread_desc),
Sequence<CYXPerBlock, 1, HPerThread, WPerThread>,
decltype(b_e_n_h_w_global_desc),
decltype(b_e_n_h_w_thread_desc),
Sequence<EPerBlock, 1, HPerThread, WPerThread>,
Sequence<3, 2, 0, 1>, // BBlockTransferSrcAccessOrder,
3, // BBlockTransferSrcVectorDim,
1, // BBlockTransferSrcScalarPerVector,
......@@ -208,12 +208,12 @@ struct GridwiseDynamicGemm_km_kn_mn_v2
AddressSpace::Vgpr,
InMemoryDataOperation::Set,
1,
true>(b_cyx_n_h_w_global_desc,
true>(b_e_n_h_w_global_desc,
make_multi_index(0, 0, h_thread_data_on_global, w_thread_data_on_global));
// LDS allocation for A and B: be careful of alignment
constexpr auto a_block_space_size =
math::integer_least_multiple(a_cyx_k_block_desc.GetElementSpaceSize(), max_lds_align);
math::integer_least_multiple(a_e_k_block_desc.GetElementSpaceSize(), max_lds_align);
Float* p_a_block_double = p_shared_block;
......@@ -223,37 +223,37 @@ struct GridwiseDynamicGemm_km_kn_mn_v2
// zero out threadwise output
threadwise_matrix_set_zero_v3(c_k_n_h_w_thread_desc, p_c_thread);
constexpr auto a_block_slice_copy_step = make_multi_index(CYXPerBlock, 0);
constexpr auto b_thread_slice_copy_step = make_multi_index(CYXPerBlock, 0, 0, 0);
constexpr auto a_block_slice_copy_step = make_multi_index(EPerBlock, 0);
constexpr auto b_thread_slice_copy_step = make_multi_index(EPerBlock, 0, 0, 0);
// hack to control index calculation when iterating over A and B matrix for threadwise copy
constexpr auto a_k_m_global_iterator_hacks = AGlobalIteratorHacks{};
constexpr auto b_cyx_n_h_w_global_iterator_hacks = BGlobalIteratorHacks{};
constexpr auto a_k_m_global_iterator_hacks = AGlobalIteratorHacks{};
constexpr auto b_e_n_h_w_global_iterator_hacks = BGlobalIteratorHacks{};
// hack to control index calculation when move slice window for A and B matrix for
// threadwise copy
constexpr auto a_k_m_global_move_slice_window_iterator_hack =
AGlobalMoveSliceWindowIteratorHacks{};
constexpr auto b_cyx_n_h_w_global_move_slice_window_iterator_hack =
constexpr auto b_e_n_h_w_global_move_slice_window_iterator_hack =
BGlobalMoveSliceWindowIteratorHacks{};
constexpr auto b_thread_space_size = b_cyx_n_h_w_thread_desc.GetElementSpaceSize();
constexpr auto b_thread_space_size = b_e_n_h_w_thread_desc.GetElementSpaceSize();
Float p_b_thread[b_thread_space_size * 2];
Float* p_b_thread_double = p_b_thread;
// LDS double buffer: preload data into LDS
{
a_blockwise_copy.RunRead(a_cyx_k_global_desc, p_a_global, a_k_m_global_iterator_hacks);
a_blockwise_copy.RunRead(a_e_k_global_desc, p_a_global, a_k_m_global_iterator_hacks);
b_threadwise_transfer.Run(b_cyx_n_h_w_global_desc,
b_threadwise_transfer.Run(b_e_n_h_w_global_desc,
p_b_global,
b_cyx_n_h_w_thread_desc,
b_e_n_h_w_thread_desc,
make_tuple(I0, I0, I0, I0),
p_b_thread_double,
b_cyx_n_h_w_global_iterator_hacks);
b_e_n_h_w_global_iterator_hacks);
a_blockwise_copy.RunWrite(a_cyx_k_block_desc, p_a_block_double);
a_blockwise_copy.RunWrite(a_e_k_block_desc, p_a_block_double);
}
#if 1
......@@ -272,89 +272,89 @@ struct GridwiseDynamicGemm_km_kn_mn_v2
do
{
// even iteration
a_blockwise_copy.MoveSrcSliceWindow(a_cyx_k_global_desc,
a_blockwise_copy.MoveSrcSliceWindow(a_e_k_global_desc,
a_block_slice_copy_step,
a_k_m_global_move_slice_window_iterator_hack);
b_threadwise_transfer.MoveSrcSliceWindow(b_cyx_n_h_w_global_desc,
b_threadwise_transfer.MoveSrcSliceWindow(b_e_n_h_w_global_desc,
b_thread_slice_copy_step);
__syncthreads();
// LDS doubel buffer: load next data from device mem
a_blockwise_copy.RunRead(
a_cyx_k_global_desc, p_a_global, a_k_m_global_iterator_hacks);
a_e_k_global_desc, p_a_global, a_k_m_global_iterator_hacks);
b_threadwise_transfer.Run(b_cyx_n_h_w_global_desc,
b_threadwise_transfer.Run(b_e_n_h_w_global_desc,
p_b_global,
b_cyx_n_h_w_thread_desc,
b_e_n_h_w_thread_desc,
make_tuple(I0, I0, I0, I0),
p_b_thread_odd,
b_cyx_n_h_w_global_iterator_hacks);
b_e_n_h_w_global_iterator_hacks);
// LDS double buffer: GEMM on current data
blockwise_gemm.Run(p_a_block_even, p_b_thread_even, p_c_thread);
// LDS double buffer: store next data to LDS
a_blockwise_copy.RunWrite(a_cyx_k_block_desc, p_a_block_odd);
a_blockwise_copy.RunWrite(a_e_k_block_desc, p_a_block_odd);
// odd iteration
a_blockwise_copy.MoveSrcSliceWindow(a_cyx_k_global_desc,
a_blockwise_copy.MoveSrcSliceWindow(a_e_k_global_desc,
a_block_slice_copy_step,
a_k_m_global_move_slice_window_iterator_hack);
b_threadwise_transfer.MoveSrcSliceWindow(b_cyx_n_h_w_global_desc,
b_threadwise_transfer.MoveSrcSliceWindow(b_e_n_h_w_global_desc,
b_thread_slice_copy_step);
__syncthreads();
// LDS doubel buffer: load next data from device mem
a_blockwise_copy.RunRead(
a_cyx_k_global_desc, p_a_global, a_k_m_global_iterator_hacks);
a_e_k_global_desc, p_a_global, a_k_m_global_iterator_hacks);
b_threadwise_transfer.Run(b_cyx_n_h_w_global_desc,
b_threadwise_transfer.Run(b_e_n_h_w_global_desc,
p_b_global,
b_cyx_n_h_w_thread_desc,
b_e_n_h_w_thread_desc,
make_tuple(I0, I0, I0, I0),
p_b_thread_even,
b_cyx_n_h_w_global_iterator_hacks);
b_e_n_h_w_global_iterator_hacks);
// LDS double buffer: GEMM on current data
blockwise_gemm.Run(p_a_block_odd, p_b_thread_odd, p_c_thread);
// LDS double buffer: store next data to LDS
a_blockwise_copy.RunWrite(a_cyx_k_block_desc, p_a_block_even);
a_blockwise_copy.RunWrite(a_e_k_block_desc, p_a_block_even);
b_block_data_begin += 2 * CYXPerBlock;
} while(b_block_data_begin < CYX - 2 * CYXPerBlock);
b_block_data_begin += 2 * EPerBlock;
} while(b_block_data_begin < E - 2 * EPerBlock);
}
// LDS double buffer: tail
if constexpr(HasDoubleTailKBlockLoop) // if has 2 iteration left
{
a_blockwise_copy.MoveSrcSliceWindow(a_cyx_k_global_desc,
a_blockwise_copy.MoveSrcSliceWindow(a_e_k_global_desc,
a_block_slice_copy_step,
a_k_m_global_move_slice_window_iterator_hack);
b_threadwise_transfer.MoveSrcSliceWindow(b_cyx_n_h_w_global_desc,
b_threadwise_transfer.MoveSrcSliceWindow(b_e_n_h_w_global_desc,
b_thread_slice_copy_step);
__syncthreads();
// LDS double buffer: load last data from device mem
a_blockwise_copy.RunRead(a_cyx_k_global_desc, p_a_global, a_k_m_global_iterator_hacks);
a_blockwise_copy.RunRead(a_e_k_global_desc, p_a_global, a_k_m_global_iterator_hacks);
b_threadwise_transfer.Run(b_cyx_n_h_w_global_desc,
b_threadwise_transfer.Run(b_e_n_h_w_global_desc,
p_b_global,
b_cyx_n_h_w_thread_desc,
b_e_n_h_w_thread_desc,
make_tuple(I0, I0, I0, I0),
p_b_thread_double + b_thread_space_size,
b_cyx_n_h_w_global_iterator_hacks);
b_e_n_h_w_global_iterator_hacks);
// LDS double buffer: GEMM on 2nd-last data
blockwise_gemm.Run(p_a_block_double, p_b_thread_double, p_c_thread);
// LDS double buffer: store last data to LDS
a_blockwise_copy.RunWrite(a_cyx_k_block_desc, p_a_block_double + a_block_space_size);
a_blockwise_copy.RunWrite(a_e_k_block_desc, p_a_block_double + a_block_space_size);
__syncthreads();
......@@ -410,9 +410,9 @@ struct GridwiseDynamicGemm_km_kn_mn_v2
// pass tensor descriptor by reference
template <bool HasMainKBlockLoop, bool HasDoubleTailKBlockLoop>
__device__ void Run(const AGlobalDesc& a_cyx_k_global_desc,
__device__ void Run(const AGlobalDesc& a_e_k_global_desc,
const Float* __restrict__ p_a_global,
const BGlobalDesc& b_cyx_n_h_w_global_desc,
const BGlobalDesc& b_e_n_h_w_global_desc,
const Float* __restrict__ p_b_global,
const CGlobalDesc& c_k_n_h_w_global_desc,
Float* __restrict__ p_c_global,
......@@ -423,9 +423,9 @@ struct GridwiseDynamicGemm_km_kn_mn_v2
__shared__ Float p_shared_block[shared_block_size];
Run(a_cyx_k_global_desc,
Run(a_e_k_global_desc,
p_a_global,
b_cyx_n_h_w_global_desc,
b_e_n_h_w_global_desc,
p_b_global,
c_k_n_h_w_global_desc,
p_c_global,
......@@ -436,22 +436,22 @@ struct GridwiseDynamicGemm_km_kn_mn_v2
// pass tensor descriptors by their pointers
template <bool HasMainKBlockLoop, bool HasDoubleTailKBlockLoop>
__device__ void Run(const AGlobalDesc* p_a_cyx_k_global_desc,
__device__ void Run(const AGlobalDesc* p_a_e_k_global_desc,
const Float* __restrict__ p_a_global,
const BGlobalDesc* p_b_cyx_n_h_w_global_desc,
const BGlobalDesc* p_b_e_n_h_w_global_desc,
const Float* __restrict__ p_b_global,
const CGlobalDesc* p_c_k_n_h_w_global_desc,
Float* __restrict__ p_c_global,
integral_constant<bool, HasMainKBlockLoop>,
integral_constant<bool, HasDoubleTailKBlockLoop>) const
{
const auto a_cyx_k_global_desc = *p_a_cyx_k_global_desc;
const auto b_cyx_n_h_w_global_desc = *p_b_cyx_n_h_w_global_desc;
const auto c_k_n_h_w_global_desc = *p_c_k_n_h_w_global_desc;
const auto a_e_k_global_desc = *p_a_e_k_global_desc;
const auto b_e_n_h_w_global_desc = *p_b_e_n_h_w_global_desc;
const auto c_k_n_h_w_global_desc = *p_c_k_n_h_w_global_desc;
Run(a_cyx_k_global_desc,
Run(a_e_k_global_desc,
p_a_global,
b_cyx_n_h_w_global_desc,
b_e_n_h_w_global_desc,
p_b_global,
c_k_n_h_w_global_desc,
p_c_global,
......@@ -461,25 +461,24 @@ struct GridwiseDynamicGemm_km_kn_mn_v2
// pass tensor descriptors by void*
template <bool HasMainKBlockLoop, bool HasDoubleTailKBlockLoop>
__device__ void Run(const void* p_a_cyx_k_global_desc,
__device__ void Run(const void* p_a_e_k_global_desc,
const Float* __restrict__ p_a_global,
const void* p_b_cyx_n_h_w_global_desc,
const void* p_b_e_n_h_w_global_desc,
const Float* __restrict__ p_b_global,
const void* p_c_k_n_h_w_global_desc,
Float* __restrict__ p_c_global,
integral_constant<bool, HasMainKBlockLoop>,
integral_constant<bool, HasDoubleTailKBlockLoop>) const
{
const auto a_cyx_k_global_desc =
*reinterpret_cast<const AGlobalDesc*>(p_a_cyx_k_global_desc);
const auto b_cyx_n_h_w_global_desc =
*reinterpret_cast<const BGlobalDesc*>(p_b_cyx_n_h_w_global_desc);
const auto a_e_k_global_desc = *reinterpret_cast<const AGlobalDesc*>(p_a_e_k_global_desc);
const auto b_e_n_h_w_global_desc =
*reinterpret_cast<const BGlobalDesc*>(p_b_e_n_h_w_global_desc);
const auto c_k_n_h_w_global_desc =
*reinterpret_cast<const CGlobalDesc*>(p_c_k_n_h_w_global_desc);
Run(a_cyx_k_global_desc,
Run(a_e_k_global_desc,
p_a_global,
b_cyx_n_h_w_global_desc,
b_e_n_h_w_global_desc,
p_b_global,
c_k_n_h_w_global_desc,
p_c_global,
......@@ -498,11 +497,11 @@ template <index_t BlockSize,
index_t KPerBlock,
index_t HPerBlock,
index_t WPerBlock,
index_t CYXPerBlock,
index_t EPerBlock,
index_t KPerThread,
index_t HPerThread,
index_t WPerThread,
index_t CYXPerThread,
index_t EPerThread,
typename ABlockTransferThreadSliceLengths_K_M,
typename ABlockTransferThreadClusterLengths_K_M,
typename ABlockTransferThreadClusterArrangeOrder,
......@@ -529,28 +528,28 @@ struct GridwiseDynamicGemm_km_kn_mn_v3
{
__host__ __device__ static constexpr index_t GetSharedMemoryNumberOfByte()
{
const auto CYX = 4 * 3 * 3;
const auto K = 16;
const auto E = 4 * 3 * 3;
const auto K = 16;
constexpr auto max_lds_align =
math::lcm(Number<ABlockTransferDstScalarPerVector_M>{}, Number<K>{});
// A matrix in LDS memory, dst of blockwise copy
// be careful of LDS alignment
constexpr auto a_cyx_k_desc = make_dynamic_naive_tensor_descriptor_aligned_v2(
make_tuple(Number<CYX>{}, Number<K>{}), max_lds_align);
constexpr auto a_e_k_desc = make_dynamic_naive_tensor_descriptor_aligned_v2(
make_tuple(Number<E>{}, Number<K>{}), max_lds_align);
// LDS allocation for A and B: be careful of alignment
constexpr auto a_block_space_size =
math::integer_least_multiple(a_cyx_k_desc.GetElementSpaceSize(), max_lds_align);
math::integer_least_multiple(a_e_k_desc.GetElementSpaceSize(), max_lds_align);
return a_block_space_size * sizeof(Float);
}
template <bool HasMainKBlockLoop, bool HasDoubleTailKBlockLoop>
__device__ void Run(const AGlobalDesc& a_cyx_k_global_desc,
__device__ void Run(const AGlobalDesc& a_e_k_global_desc,
const Float* __restrict__ p_a_global,
const BGlobalDesc& b_cyx_n_h_w_global_desc,
const BGlobalDesc& b_e_n_h_w_global_desc,
const Float* __restrict__ p_b_global,
const CGlobalDesc& c_k_n_h_w_global_desc,
Float* __restrict__ p_c_global,
......@@ -563,12 +562,12 @@ struct GridwiseDynamicGemm_km_kn_mn_v3
constexpr auto I2 = Number<2>{};
constexpr auto I3 = Number<3>{};
const auto CYX = a_cyx_k_global_desc.GetLength(I0);
const auto K = a_cyx_k_global_desc.GetLength(I1);
const auto E = a_e_k_global_desc.GetLength(I0);
const auto K = a_e_k_global_desc.GetLength(I1);
const auto N = b_cyx_n_h_w_global_desc.GetLength(I1);
const auto H = b_cyx_n_h_w_global_desc.GetLength(I2);
const auto W = b_cyx_n_h_w_global_desc.GetLength(I3);
const auto N = b_e_n_h_w_global_desc.GetLength(I1);
const auto H = b_e_n_h_w_global_desc.GetLength(I2);
const auto W = b_e_n_h_w_global_desc.GetLength(I3);
// divide block work by [M, N]
#if 1
......@@ -601,17 +600,16 @@ struct GridwiseDynamicGemm_km_kn_mn_v3
// A matrix in LDS memory, dst of blockwise copy
// be careful of LDS alignment
constexpr auto a_cyx_k_block_desc = make_dynamic_naive_tensor_descriptor_aligned_v2(
make_tuple(Number<CYXPerBlock>{}, Number<KPerBlock>{}), max_lds_align);
constexpr auto a_e_k_block_desc = make_dynamic_naive_tensor_descriptor_aligned_v2(
make_tuple(Number<EPerBlock>{}, Number<KPerBlock>{}), max_lds_align);
constexpr auto a_cyx_k_desc = make_dynamic_naive_tensor_descriptor_aligned_v2(
make_tuple(Number<CYX>{}, Number<K>{}), max_lds_align);
constexpr auto a_e_k_desc = make_dynamic_naive_tensor_descriptor_aligned_v2(
make_tuple(Number<E>{}, Number<K>{}), max_lds_align);
// B matrix in LDS memory, dst of blockwise copy
// be careful of LDS alignment
constexpr auto b_cyx_n_h_w_block_desc =
make_dynamic_naive_tensor_descriptor_packed_v2(make_tuple(
Number<CYXPerBlock>{}, Number<1>{}, Number<HPerBlock>{}, Number<WPerBlock>{}));
constexpr auto b_e_n_h_w_block_desc = make_dynamic_naive_tensor_descriptor_packed_v2(
make_tuple(Number<EPerBlock>{}, Number<1>{}, Number<HPerBlock>{}, Number<WPerBlock>{}));
// c_thread_mtx definition: this is a mess
// TODO:: more elegent way of defining c_thread_mtx
......@@ -619,18 +617,17 @@ struct GridwiseDynamicGemm_km_kn_mn_v3
make_dynamic_naive_tensor_descriptor_packed_v2(make_tuple(
Number<KPerThread>{}, Number<1>{}, Number<HPerThread>{}, Number<WPerThread>{}));
const auto blockwise_gemm =
BlockwiseGemm_km_kn_m0m1n0n1_v3<BlockSize,
decltype(a_cyx_k_block_desc),
decltype(b_cyx_n_h_w_block_desc),
decltype(c_k_n_h_w_thread_desc),
KPerThread, // KPerThreadSubC
HPerThread, // HPerThreadSubC
WPerThread, // WPerThreadSubC
CYXPerThread, // CYXPerThreadLoop
1, // ThreadGemmADataPerRead_K
1 // ThreadGemmBDataPerRead_W
>{};
const auto blockwise_gemm = BlockwiseGemm_km_kn_m0m1n0n1_v3<BlockSize,
decltype(a_e_k_block_desc),
decltype(b_e_n_h_w_block_desc),
decltype(c_k_n_h_w_thread_desc),
KPerThread, // KPerThreadSubC
HPerThread, // HPerThreadSubC
WPerThread, // WPerThreadSubC
EPerThread, // EPerThreadLoop
1, // ThreadGemmADataPerRead_K
1 // ThreadGemmBDataPerRead_W
>{};
auto c_thread_mtx_index = blockwise_gemm.GetBeginOfThreadMatrixC(get_thread_local_1d_id());
......@@ -646,43 +643,44 @@ struct GridwiseDynamicGemm_km_kn_mn_v3
const index_t w_thread_data_on_global = w_block_data_on_global + w_thread_id * WPerThread;
// A matrix blockwise copy
auto a_blockwise_copy = BlockwiseDynamicTensorSliceTransfer_v4<
BlockSize,
InMemoryDataOperation::Set,
Sequence<CYX, K>,
ABlockTransferThreadSliceLengths_K_M,
ABlockTransferThreadClusterLengths_K_M,
ABlockTransferThreadClusterArrangeOrder,
Float,
Float,
decltype(a_cyx_k_global_desc),
decltype(a_cyx_k_desc),
ABlockTransferSrcAccessOrder,
Sequence<0, 1>,
ABlockTransferSrcVectorDim,
1,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_M,
AddressSpace::Global,
AddressSpace::Lds,
1,
1,
AThreadTransferSrcResetCoordinateAfterRun,
true>(a_cyx_k_global_desc,
make_multi_index(0, k_block_data_on_global),
a_cyx_k_desc,
make_multi_index(0, 0));
auto a_blockwise_copy =
BlockwiseDynamicTensorSliceTransfer_v4<BlockSize,
InMemoryDataOperation::Set,
Sequence<E, K>,
ABlockTransferThreadSliceLengths_K_M,
ABlockTransferThreadClusterLengths_K_M,
ABlockTransferThreadClusterArrangeOrder,
Float,
Float,
decltype(a_e_k_global_desc),
decltype(a_e_k_desc),
ABlockTransferSrcAccessOrder,
Sequence<0, 1>,
ABlockTransferSrcVectorDim,
1,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_M,
AddressSpace::Global,
AddressSpace::Lds,
1,
1,
AThreadTransferSrcResetCoordinateAfterRun,
true>(
a_e_k_global_desc,
make_multi_index(0, k_block_data_on_global),
a_e_k_desc,
make_multi_index(0, 0));
constexpr auto b_cyx_n_h_w_thread_desc =
constexpr auto b_e_n_h_w_thread_desc =
make_dynamic_naive_tensor_descriptor_packed_v2(make_tuple(
Number<CYXPerBlock>{}, Number<1>{}, Number<HPerThread>{}, Number<WPerThread>{}));
Number<EPerBlock>{}, Number<1>{}, Number<HPerThread>{}, Number<WPerThread>{}));
auto b_threadwise_transfer = ThreadwiseDynamicTensorSliceTransfer_v2<
Float,
Float,
decltype(b_cyx_n_h_w_global_desc),
decltype(b_cyx_n_h_w_thread_desc),
Sequence<CYXPerBlock, 1, HPerThread, WPerThread>,
decltype(b_e_n_h_w_global_desc),
decltype(b_e_n_h_w_thread_desc),
Sequence<EPerBlock, 1, HPerThread, WPerThread>,
Sequence<3, 2, 0, 1>, // BBlockTransferSrcAccessOrder,
3, // BBlockTransferSrcVectorDim,
1, // BBlockTransferSrcScalarPerVector,
......@@ -690,7 +688,7 @@ struct GridwiseDynamicGemm_km_kn_mn_v3
AddressSpace::Vgpr,
InMemoryDataOperation::Set,
1,
true>(b_cyx_n_h_w_global_desc,
true>(b_e_n_h_w_global_desc,
make_multi_index(0, 0, h_thread_data_on_global, w_thread_data_on_global));
Float* p_a_block = p_shared_block;
......@@ -701,36 +699,36 @@ struct GridwiseDynamicGemm_km_kn_mn_v3
// zero out threadwise output
threadwise_matrix_set_zero_v3(c_k_n_h_w_thread_desc, p_c_thread);
constexpr auto b_thread_slice_copy_step = make_multi_index(CYXPerBlock, 0, 0, 0);
constexpr auto b_thread_slice_copy_step = make_multi_index(EPerBlock, 0, 0, 0);
// hack to control index calculation when iterating over A and B matrix for threadwise copy
constexpr auto a_k_m_global_iterator_hacks = AGlobalIteratorHacks{};
constexpr auto b_cyx_n_h_w_global_iterator_hacks = BGlobalIteratorHacks{};
constexpr auto a_k_m_global_iterator_hacks = AGlobalIteratorHacks{};
constexpr auto b_e_n_h_w_global_iterator_hacks = BGlobalIteratorHacks{};
// hack to control index calculation when move slice window for A and B matrix for
// threadwise copy
constexpr auto a_k_m_global_move_slice_window_iterator_hack =
AGlobalMoveSliceWindowIteratorHacks{};
constexpr auto b_cyx_n_h_w_global_move_slice_window_iterator_hack =
constexpr auto b_e_n_h_w_global_move_slice_window_iterator_hack =
BGlobalMoveSliceWindowIteratorHacks{};
constexpr auto b_thread_space_size = b_cyx_n_h_w_thread_desc.GetElementSpaceSize();
constexpr auto b_thread_space_size = b_e_n_h_w_thread_desc.GetElementSpaceSize();
Float p_b_thread[b_thread_space_size * 2];
Float* p_b_thread_double = p_b_thread;
// LDS double buffer: preload data into LDS
{
a_blockwise_copy.RunRead(a_cyx_k_global_desc, p_a_global, a_k_m_global_iterator_hacks);
a_blockwise_copy.RunRead(a_e_k_global_desc, p_a_global, a_k_m_global_iterator_hacks);
b_threadwise_transfer.Run(b_cyx_n_h_w_global_desc,
b_threadwise_transfer.Run(b_e_n_h_w_global_desc,
p_b_global,
b_cyx_n_h_w_thread_desc,
b_e_n_h_w_thread_desc,
make_tuple(I0, I0, I0, I0),
p_b_thread_double,
b_cyx_n_h_w_global_iterator_hacks);
b_e_n_h_w_global_iterator_hacks);
a_blockwise_copy.RunWrite(a_cyx_k_desc, p_a_block);
a_blockwise_copy.RunWrite(a_e_k_desc, p_a_block);
}
__syncthreads();
......@@ -748,69 +746,69 @@ struct GridwiseDynamicGemm_km_kn_mn_v3
do
{
// even iteration
b_threadwise_transfer.MoveSrcSliceWindow(b_cyx_n_h_w_global_desc,
b_threadwise_transfer.MoveSrcSliceWindow(b_e_n_h_w_global_desc,
b_thread_slice_copy_step);
b_threadwise_transfer.Run(b_cyx_n_h_w_global_desc,
b_threadwise_transfer.Run(b_e_n_h_w_global_desc,
p_b_global,
b_cyx_n_h_w_thread_desc,
b_e_n_h_w_thread_desc,
make_tuple(I0, I0, I0, I0),
p_b_thread_odd,
b_cyx_n_h_w_global_iterator_hacks);
b_e_n_h_w_global_iterator_hacks);
// LDS double buffer: GEMM on current data
blockwise_gemm.Run(p_a_block + a_cyx_k_block_desc.CalculateOffset(
make_tuple(b_block_data_begin, 0)),
p_b_thread_even,
p_c_thread);
blockwise_gemm.Run(
p_a_block + a_e_k_block_desc.CalculateOffset(make_tuple(b_block_data_begin, 0)),
p_b_thread_even,
p_c_thread);
b_block_data_begin += CYXPerBlock;
b_block_data_begin += EPerBlock;
b_threadwise_transfer.MoveSrcSliceWindow(b_cyx_n_h_w_global_desc,
b_threadwise_transfer.MoveSrcSliceWindow(b_e_n_h_w_global_desc,
b_thread_slice_copy_step);
b_threadwise_transfer.Run(b_cyx_n_h_w_global_desc,
b_threadwise_transfer.Run(b_e_n_h_w_global_desc,
p_b_global,
b_cyx_n_h_w_thread_desc,
b_e_n_h_w_thread_desc,
make_tuple(I0, I0, I0, I0),
p_b_thread_even,
b_cyx_n_h_w_global_iterator_hacks);
b_e_n_h_w_global_iterator_hacks);
// LDS double buffer: GEMM on current data
blockwise_gemm.Run(p_a_block + a_cyx_k_block_desc.CalculateOffset(
make_tuple(b_block_data_begin, 0)),
p_b_thread_odd,
p_c_thread);
blockwise_gemm.Run(
p_a_block + a_e_k_block_desc.CalculateOffset(make_tuple(b_block_data_begin, 0)),
p_b_thread_odd,
p_c_thread);
b_block_data_begin += CYXPerBlock;
b_block_data_begin += EPerBlock;
} while(b_block_data_begin < CYX - 2 * CYXPerBlock);
} while(b_block_data_begin < E - 2 * EPerBlock);
}
// LDS double buffer: tail
if constexpr(HasDoubleTailKBlockLoop) // if has 2 iteration left
{
b_threadwise_transfer.MoveSrcSliceWindow(b_cyx_n_h_w_global_desc,
b_threadwise_transfer.MoveSrcSliceWindow(b_e_n_h_w_global_desc,
b_thread_slice_copy_step);
b_threadwise_transfer.Run(b_cyx_n_h_w_global_desc,
b_threadwise_transfer.Run(b_e_n_h_w_global_desc,
p_b_global,
b_cyx_n_h_w_thread_desc,
b_e_n_h_w_thread_desc,
make_tuple(I0, I0, I0, I0),
p_b_thread_double + b_thread_space_size,
b_cyx_n_h_w_global_iterator_hacks);
b_e_n_h_w_global_iterator_hacks);
// LDS double buffer: GEMM on 2nd-last data
blockwise_gemm.Run(
p_a_block + a_cyx_k_block_desc.CalculateOffset(make_tuple(b_block_data_begin, 0)),
p_a_block + a_e_k_block_desc.CalculateOffset(make_tuple(b_block_data_begin, 0)),
p_b_thread_double,
p_c_thread);
b_block_data_begin += CYXPerBlock;
b_block_data_begin += EPerBlock;
// LDS double buffer: GEMM on last data
blockwise_gemm.Run(
p_a_block + a_cyx_k_block_desc.CalculateOffset(make_tuple(b_block_data_begin, 0)),
p_a_block + a_e_k_block_desc.CalculateOffset(make_tuple(b_block_data_begin, 0)),
p_b_thread_double + b_thread_space_size,
p_c_thread);
}
......@@ -818,7 +816,7 @@ struct GridwiseDynamicGemm_km_kn_mn_v3
{
// LDS double buffer: GEMM on last data
blockwise_gemm.Run(
p_a_block + a_cyx_k_block_desc.CalculateOffset(make_tuple(b_block_data_begin, 0)),
p_a_block + a_e_k_block_desc.CalculateOffset(make_tuple(b_block_data_begin, 0)),
p_b_thread_double,
p_c_thread);
}
......@@ -862,9 +860,9 @@ struct GridwiseDynamicGemm_km_kn_mn_v3
// pass tensor descriptor by reference
template <bool HasMainKBlockLoop, bool HasDoubleTailKBlockLoop>
__device__ void Run(const AGlobalDesc& a_cyx_k_global_desc,
__device__ void Run(const AGlobalDesc& a_e_k_global_desc,
const Float* __restrict__ p_a_global,
const BGlobalDesc& b_cyx_n_h_w_global_desc,
const BGlobalDesc& b_e_n_h_w_global_desc,
const Float* __restrict__ p_b_global,
const CGlobalDesc& c_k_n_h_w_global_desc,
Float* __restrict__ p_c_global,
......@@ -875,9 +873,9 @@ struct GridwiseDynamicGemm_km_kn_mn_v3
__shared__ Float p_shared_block[shared_block_size];
Run(a_cyx_k_global_desc,
Run(a_e_k_global_desc,
p_a_global,
b_cyx_n_h_w_global_desc,
b_e_n_h_w_global_desc,
p_b_global,
c_k_n_h_w_global_desc,
p_c_global,
......@@ -888,22 +886,22 @@ struct GridwiseDynamicGemm_km_kn_mn_v3
// pass tensor descriptors by their pointers
template <bool HasMainKBlockLoop, bool HasDoubleTailKBlockLoop>
__device__ void Run(const AGlobalDesc* p_a_cyx_k_global_desc,
__device__ void Run(const AGlobalDesc* p_a_e_k_global_desc,
const Float* __restrict__ p_a_global,
const BGlobalDesc* p_b_cyx_n_h_w_global_desc,
const BGlobalDesc* p_b_e_n_h_w_global_desc,
const Float* __restrict__ p_b_global,
const CGlobalDesc* p_c_k_n_h_w_global_desc,
Float* __restrict__ p_c_global,
integral_constant<bool, HasMainKBlockLoop>,
integral_constant<bool, HasDoubleTailKBlockLoop>) const
{
const auto a_cyx_k_global_desc = *p_a_cyx_k_global_desc;
const auto b_cyx_n_h_w_global_desc = *p_b_cyx_n_h_w_global_desc;
const auto c_k_n_h_w_global_desc = *p_c_k_n_h_w_global_desc;
const auto a_e_k_global_desc = *p_a_e_k_global_desc;
const auto b_e_n_h_w_global_desc = *p_b_e_n_h_w_global_desc;
const auto c_k_n_h_w_global_desc = *p_c_k_n_h_w_global_desc;
Run(a_cyx_k_global_desc,
Run(a_e_k_global_desc,
p_a_global,
b_cyx_n_h_w_global_desc,
b_e_n_h_w_global_desc,
p_b_global,
c_k_n_h_w_global_desc,
p_c_global,
......@@ -913,25 +911,24 @@ struct GridwiseDynamicGemm_km_kn_mn_v3
// pass tensor descriptors by void*
template <bool HasMainKBlockLoop, bool HasDoubleTailKBlockLoop>
__device__ void Run(const void* p_a_cyx_k_global_desc,
__device__ void Run(const void* p_a_e_k_global_desc,
const Float* __restrict__ p_a_global,
const void* p_b_cyx_n_h_w_global_desc,
const void* p_b_e_n_h_w_global_desc,
const Float* __restrict__ p_b_global,
const void* p_c_k_n_h_w_global_desc,
Float* __restrict__ p_c_global,
integral_constant<bool, HasMainKBlockLoop>,
integral_constant<bool, HasDoubleTailKBlockLoop>) const
{
const auto a_cyx_k_global_desc =
*reinterpret_cast<const AGlobalDesc*>(p_a_cyx_k_global_desc);
const auto b_cyx_n_h_w_global_desc =
*reinterpret_cast<const BGlobalDesc*>(p_b_cyx_n_h_w_global_desc);
const auto a_e_k_global_desc = *reinterpret_cast<const AGlobalDesc*>(p_a_e_k_global_desc);
const auto b_e_n_h_w_global_desc =
*reinterpret_cast<const BGlobalDesc*>(p_b_e_n_h_w_global_desc);
const auto c_k_n_h_w_global_desc =
*reinterpret_cast<const CGlobalDesc*>(p_c_k_n_h_w_global_desc);
Run(a_cyx_k_global_desc,
Run(a_e_k_global_desc,
p_a_global,
b_cyx_n_h_w_global_desc,
b_e_n_h_w_global_desc,
p_b_global,
c_k_n_h_w_global_desc,
p_c_global,
......
composable_kernel/include/tensor_operation/threadwise_gemm_v3.hpp
View file @ 666bdad1
......@@ -57,10 +57,10 @@ struct ThreadwiseGemm_km_kn_mn_v3
constexpr auto H = BDesc{}.GetLength(I2);
constexpr auto W = BDesc{}.GetLength(I3);
constexpr auto CYX = ADesc{}.GetLength(I0);
constexpr auto K = ADesc{}.GetLength(I1);
constexpr auto E = ADesc{}.GetLength(I0);
constexpr auto K = ADesc{}.GetLength(I1);
static_for<0, CYX, 1>{}([&](auto e) {
static_for<0, E, 1>{}([&](auto e) {
static_for<0, K, 1>{}([&](auto k) {
static_for<0, H, 1>{}([&](auto h) {
static_for<0, W, 1>{}([&](auto w) {
......
driver/include/device_dynamic_convolution_forward_implicit_gemm_v5r1_nchw_kcyx_nkhw.hpp
View file @ 666bdad1
......@@ -73,15 +73,15 @@ void device_dynamic_convolution_forward_implicit_gemm_v5r1_nchw_kcyx_nkhw(InDesc
constexpr index_t KPerBlock = 16;
constexpr index_t HPerBlock = 16;
constexpr index_t WPerBlock = 16;
constexpr index_t CYXPerBlock = 2 * 3 * 3;
constexpr index_t CYXPerBlock = 4;
constexpr index_t KPerThread = 4;
constexpr index_t HPerThread = 2;
constexpr index_t WPerThread = 2;
constexpr index_t CYXPerThread = 2;
constexpr index_t CYXPerThread = 4;
using GemmABlockTransferThreadSliceLengths_GemmK_GemmM = Sequence<4, 1>;
using GemmABlockTransferThreadClusterLengths_GemmK_GemmM = Sequence<9, 16>;
using GemmABlockTransferThreadSliceLengths_GemmK_GemmM = Sequence<1, 1>;
using GemmABlockTransferThreadClusterLengths_GemmK_GemmM = Sequence<4, 16>;
constexpr index_t GemmABlockTransferSrcScalarPerVector_GemmK = 1;
constexpr index_t GemmABlockTransferDstScalarPerVector_GemmM = 1;
......
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