Skip to content

GitLab

Commit 2603bb0f authored by Chao Liu's avatar Chao Liu
Browse files

tuning on vega 20

parent a9031464
Hide whitespace changes
Inline Side-by-side
Showing with 670 additions and 674 deletions
driver/device_convolution_implicit_gemm_v1_chwn_cyxk_khwn.hpp
View file @ 2603bb0f
......@@ -140,7 +140,7 @@ void device_convolution_implicit_gemm_v1_chwn_cyxk_khwn(InDesc,
constexpr index_t WeiBlockCopyDataPerRead_K = 4;
constexpr index_t OutThreadCopyDataPerWrite_N = 2;
#elif 1
#elif 0
// for 3x3, 34x34, v1r3, Pascal
// for 3x3, 28x28, v1r3, Pascal
// for 3x3, 14x14, v1r3, Pascal
......@@ -206,6 +206,8 @@ void device_convolution_implicit_gemm_v1_chwn_cyxk_khwn(InDesc,
constexpr index_t OutThreadCopyDataPerWrite_N = 1;
#elif 0
// for 3x3, 34x34, v1r1, Vega 20
constexpr index_t BlockSize = 256;
constexpr index_t NPerBlock = 16;
constexpr index_t KPerBlock = 128;
constexpr index_t CPerBlock = 4;
......@@ -227,16 +229,43 @@ void device_convolution_implicit_gemm_v1_chwn_cyxk_khwn(InDesc,
constexpr index_t GemmDataPerReadA = 4;
constexpr index_t GemmDataPerReadB = 4;
constexpr index_t InBlockCopy_ThreadPerDimC = 4;
constexpr index_t InBlockCopy_ThreadPerDimH = 4;
constexpr index_t InBlockCopy_ThreadPerDimW = 2;
constexpr index_t InBlockCopy_ThreadPerDimN = 8;
constexpr index_t InBlockCopyDataPerRead_N = 2;
using InBlockCopyClusterLengths_CHWN = Sequence<4, 4, 2, 8>;
constexpr index_t InBlockCopyDataPerRead_N = 2;
constexpr index_t WeiBlockCopyDataPerRead_K = 2;
constexpr index_t OutThreadCopyDataPerWrite_N = 4;
constexpr index_t WeiBlockCopyDataPerRead_K = 2;
constexpr index_t OutThreadCopyDataPerWrite_N = 4;
#elif 1
// for 3x3, 34x34, v1r3, Vega 20
constexpr index_t BlockSize = 256;
constexpr index_t NPerBlock = 16;
constexpr index_t KPerBlock = 128;
constexpr index_t CPerBlock = 8;
constexpr index_t HoPerBlock = 2;
constexpr index_t WoPerBlock = 4;
constexpr index_t NPerThread = 4;
constexpr index_t KPerThread = 8;
constexpr index_t HoPerThread = 1;
constexpr index_t WoPerThread = 2;
constexpr index_t GemmMPerThreadSubC = 4;
constexpr index_t GemmNPerThreadSubC = 4;
constexpr index_t GemmMLevel0Cluster = 4;
constexpr index_t GemmNLevel0Cluster = 4;
constexpr index_t GemmMLevel1Cluster = 4;
constexpr index_t GemmNLevel1Cluster = 2;
constexpr index_t GemmKPerThreadLoop = 1;
constexpr index_t GemmDataPerReadA = 4;
constexpr index_t GemmDataPerReadB = 4;
using InBlockCopyClusterLengths_CHWN = Sequence<8, 2, 4, 4>;
constexpr index_t InBlockCopyDataPerRead_N = 4;
constexpr index_t WeiBlockCopyDataPerRead_K = 4;
constexpr index_t OutThreadCopyDataPerWrite_N = 4;
#elif 0
// for 3x3, 56x56, v1r1, Pascal
constexpr index_t NPerBlock = 32;
......@@ -448,7 +477,7 @@ void device_convolution_implicit_gemm_v1_chwn_cyxk_khwn(InDesc,
GridwiseConvolutionImplicitGemm_v1r2_chwn_cyxk_khwn
#elif 1
GridwiseConvolutionImplicitGemm_v1r3_chwn_cyxk_khwn
#elif 0
#elif 1
GridwiseConvolutionImplicitGemm_v1r3_lds_double_buffer_chwn_cyxk_khwn
#endif
<GridSize,
......
driver/device_convolution_implicit_gemm_v1_nchw_cyxk_khwn.hpp
View file @ 2603bb0f
......@@ -182,7 +182,7 @@ void device_convolution_implicit_gemm_v1_nchw_cyxk_khwn(InDesc,
constexpr auto gridwise_conv =
#if 0
GridwiseConvolutionImplicitGemm_v1r2_nchw_cyxk_khwn
#elif 1
#elif 0
GridwiseConvolutionImplicitGemm_v1r3_nchw_cyxk_khwn
#elif 1
GridwiseConvolutionImplicitGemm_v1r3_lds_double_buffer_nchw_cyxk_khwn
......
driver/device_direct_convolution_2_vectorized_nchw_kcyx_nkhw.hpp
View file @ 2603bb0f
......@@ -52,7 +52,7 @@ void device_direct_convolution_2_vectorized_nchw_kcyx_nkhw(InDesc,
in_nchw_vec(n, c, h, w) =
vector_t::Pack(in_nchw(n, 2 * c, h, w), in_nchw(n, 2 * c + 1, h, w));
#elif 1
in_nchw_vec(n, c, h, w) = vector_t::Pack(in_nchw(n, 4 * c, h, w),
in_nchw_vec(n, c, h, w) = vector_t::Pack(in_nchw(n, 4 * c, h, w),
in_nchw(n, 4 * c + 1, h, w),
in_nchw(n, 4 * c + 2, h, w),
in_nchw(n, 4 * c + 3, h, w));
......@@ -114,37 +114,37 @@ void device_direct_convolution_2_vectorized_nchw_kcyx_nkhw(InDesc,
constexpr index_t BlockSize = 128;
#elif 0
// 3x3, 34x34, 128 thread, fp32, vector = 2
constexpr index_t NPerBlock = 2;
constexpr index_t KPerBlock = 32;
constexpr index_t CPerBlock = 2;
constexpr index_t NPerBlock = 2;
constexpr index_t KPerBlock = 32;
constexpr index_t CPerBlock = 2;
constexpr index_t HoPerBlock = 2;
constexpr index_t WoPerBlock = 32;
constexpr index_t NPerThread = 2;
constexpr index_t KPerThread = 4;
constexpr index_t CPerThread = 1;
constexpr index_t NPerThread = 2;
constexpr index_t KPerThread = 4;
constexpr index_t CPerThread = 1;
constexpr index_t HoPerThread = 2;
constexpr index_t WoPerThread = 2;
constexpr index_t InBlockCopyDataPerRead = 2;
constexpr index_t InBlockCopyDataPerRead = 2;
constexpr index_t WeiBlockCopyDataPerRead = 2;
constexpr index_t BlockSize = 128;
#elif 0
// 3x3, 34x34, 128 thread, int8, vector = 4
constexpr index_t NPerBlock = 2;
constexpr index_t KPerBlock = 32;
constexpr index_t CPerBlock = 8;
constexpr index_t NPerBlock = 2;
constexpr index_t KPerBlock = 32;
constexpr index_t CPerBlock = 8;
constexpr index_t HoPerBlock = 4;
constexpr index_t WoPerBlock = 32;
constexpr index_t NPerThread = 1;
constexpr index_t KPerThread = 8;
constexpr index_t CPerThread = 2;
constexpr index_t NPerThread = 1;
constexpr index_t KPerThread = 8;
constexpr index_t CPerThread = 2;
constexpr index_t HoPerThread = 4;
constexpr index_t WoPerThread = 2;
constexpr index_t InBlockCopyDataPerRead = 2;
constexpr index_t InBlockCopyDataPerRead = 2;
constexpr index_t WeiBlockCopyDataPerRead = 2;
constexpr index_t BlockSize = 128;
......
driver/driver.hip.cpp
View file @ 2603bb0f
......@@ -371,7 +371,7 @@ void host_winograd_3x3_convolution(const Tensor<TIn>& in_nchw,
std::size_t ho = HoPerTile * htile + j;
for(int i = 0; i < WoPerTile; ++i)
{
std::size_t wo = WoPerTile * wtile + i;
std::size_t wo = WoPerTile * wtile + i;
out_nkhw(n, k, ho, wo) = out_hold(n, k, htile, wtile, j, i);
}
}
......@@ -425,13 +425,13 @@ int main(int argc, char* argv[])
constexpr index_t WPad = 0;
#elif 0
// 3x3, 56x56
constexpr index_t N = 64;
constexpr index_t C = 64;
constexpr index_t N = 64;
constexpr index_t C = 64;
constexpr index_t HI = 56;
constexpr index_t WI = 56;
constexpr index_t K = 128;
constexpr index_t Y = 3;
constexpr index_t X = 3;
constexpr index_t K = 128;
constexpr index_t Y = 3;
constexpr index_t X = 3;
constexpr index_t HPad = 0;
constexpr index_t WPad = 0;
......@@ -603,9 +603,9 @@ int main(int argc, char* argv[])
device_direct_convolution_2_nchw_kcyx_nkhw
#elif 0
device_direct_convolution_2_vectorized_nchw_kcyx_nkhw
#elif 0
#elif 1
device_convolution_implicit_gemm_v1_chwn_cyxk_khwn
#elif 0
#elif 1
device_convolution_implicit_gemm_v1_nchw_cyxk_khwn
#elif 1
device_convolution_implicit_gemm_v1_nchw_cyxk_nkhw
......
src/include/ConstantTensorDescriptor.hip.hpp
View file @ 2603bb0f
......@@ -116,11 +116,7 @@ struct ConstantTensorDescriptor
static_for<0, nDim, 1>{}([&](auto IDim) {
constexpr index_t idim = IDim.Get();
#if DEVICE_BACKEND_HIP
id += __mul24(multi_id[idim], GetStride(IDim));
#else
id += multi_id[idim] * GetStride(IDim);
#endif
});
return id;
......
src/include/blockwise_3d_tensor_op.hip.hpp
View file @ 2603bb0f
......@@ -213,7 +213,6 @@ struct Blockwise3dTensorCopy3
#pragma unroll
for(index_t iloop_d2 = 0; iloop_d2 < nloop_d2; ++iloop_d2)
{
#pragma unroll
const index_t src_offset =
SrcDesc{}.Get1dIndex(iloop_d0 * thread_per_d0,
iloop_d1 * thread_per_d1,
......
src/include/blockwise_4d_tensor_op.hip.hpp
View file @ 2603bb0f
......@@ -341,10 +341,11 @@ struct BlockwiseChwnTensorCopyPadded
constexpr index_t NLoop = ref_desc.GetElementSize() / BlockSize;
const Float* p_src_tmp =
p_src + src_desc.Get1dIndex(c_block_data_begin,
(ho_block_data_begin + h_block_pad_low) - h_global_pad_low,
(wo_block_data_begin + w_block_pad_low) - w_global_pad_low,
n_block_data_begin);
p_src +
src_desc.Get1dIndex(c_block_data_begin,
(ho_block_data_begin + h_block_pad_low) - h_global_pad_low,
(wo_block_data_begin + w_block_pad_low) - w_global_pad_low,
n_block_data_begin);
#if 0
if(get_thread_local_1d_id() == 0)
......
src/include/blockwise_batched_gemm.hip.hpp
View file @ 2603bb0f
......@@ -404,8 +404,9 @@ struct BlockwiseBatchGemmBlockABlockBThreadCTransANormalBNormalC_V2
{
threadwise_matrix_copy(
c_thread_sub_mtx,
p_c_thread + c_thread_sub_mtx.Get1dIndex(m_repeat * MPerLevel1Cluster,
n_repeat * NPerLevel1Cluster),
p_c_thread +
c_thread_sub_mtx.Get1dIndex(m_repeat * MPerLevel1Cluster,
n_repeat * NPerLevel1Cluster),
c_block_mtx,
p_c_block +
c_block_mtx.Get1dIndex(m_repeat * MPerLevel1Cluster,
......
src/include/blockwise_direct_convolution.hip.hpp
View file @ 2603bb0f
......@@ -93,10 +93,11 @@ __device__ void blockwise_direct_convolution(InBlockDesc,
Float p_out_thread[out_thread_desc.GetElementSpace()];
threadwise_4d_tensor_copy(out_block_desc,
p_out_block + out_block_desc.Get1dIndex(n_thread_data_begin,
k_thread_data_begin,
ho_thread_data_begin,
wo_thread_data_begin),
p_out_block +
out_block_desc.Get1dIndex(n_thread_data_begin,
k_thread_data_begin,
ho_thread_data_begin,
wo_thread_data_begin),
out_thread_desc,
p_out_thread,
out_thread_desc.GetLengths());
......@@ -107,10 +108,11 @@ __device__ void blockwise_direct_convolution(InBlockDesc,
// threadwise convolution
threadwise_direct_convolution_2(
in_thread_block_desc,
p_in_block + in_block_desc.Get1dIndex(n_thread_data_begin,
c_thread_data_begin,
hi_thread_data_begin,
wi_thread_data_begin),
p_in_block +
in_block_desc.Get1dIndex(n_thread_data_begin,
c_thread_data_begin,
hi_thread_data_begin,
wi_thread_data_begin),
wei_thread_block_desc,
p_wei_block +
wei_block_desc.Get1dIndex(k_thread_data_begin, c_thread_data_begin, 0, 0),
......@@ -122,10 +124,11 @@ __device__ void blockwise_direct_convolution(InBlockDesc,
threadwise_4d_tensor_copy(out_thread_desc,
p_out_thread,
out_block_desc,
p_out_block + out_block_desc.Get1dIndex(n_thread_data_begin,
k_thread_data_begin,
ho_thread_data_begin,
wo_thread_data_begin),
p_out_block +
out_block_desc.Get1dIndex(n_thread_data_begin,
k_thread_data_begin,
ho_thread_data_begin,
wo_thread_data_begin),
out_thread_desc.GetLengths());
}
}
src/include/blockwise_nd_tensor_op.hip.hpp
View file @ 2603bb0f
......@@ -56,7 +56,7 @@ struct BlockwiseNdTensorCopyReorder_v3
"wrong! BlockSize is not big enough for ThreadPerDims!");
// sanity check: work division
static_for<0, nDim, 1>{}([](auto IDim) {
static_for<0, nDim, 1>{}([&](auto IDim) {
constexpr auto I = decltype(IDim){};
constexpr index_t src_len = src_lengths.Get(I);
constexpr index_t src_sub_len = src_sub_lengths.Get(I);
......@@ -220,7 +220,7 @@ struct BlockwiseNdTensorCopyReorder_v3
constexpr index_t dst_offset = DstDesc{}.Get1dIndex(dst_data_multi_id);
// write in the order of dst
// write in the order of dst
#if 1
threadwise_nd_tensor_copy_reorder_given_dst2src_v2(thread_tensor_desc,
p_clipboard + clipboard_offset,
......
src/include/gridwise_convolution_implicit_gemm_v1r1_chwn_cyxk_khwn.hip.hpp
View file @ 2603bb0f
......@@ -43,11 +43,12 @@ struct GridwiseConvolutionImplicitGemm_v1r1_chwn_cyxk_khwn
Float* const __restrict__ p_out_global) const
{
// be careful of this assertion
static_assert(NPerBlock % NPerThread == 0 && (GemmNPerThreadSubC <= NPerBlock &&
NPerBlock % GemmNPerThreadSubC == 0) ||
(GemmNPerThreadSubC >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0),
"wrong!");
static_assert(
NPerBlock % NPerThread == 0 &&
((GemmNPerThreadSubC <= NPerBlock && NPerBlock % GemmNPerThreadSubC == 0) ||
(GemmNPerThreadSubC >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0)),
"wrong!");
constexpr auto I0 = Number<0>{};
constexpr auto I1 = Number<1>{};
......@@ -219,8 +220,9 @@ struct GridwiseConvolutionImplicitGemm_v1r1_chwn_cyxk_khwn
threadwise_4d_tensor_set_zero(out_k_h_w_n_thread_desc, p_out_thread);
const Float* p_in_global_block_offset =
p_in_global + in_c_h_w_n_global_desc.Get1dIndex(
0, hi_block_data_begin, wi_block_data_begin, n_block_data_begin);
p_in_global +
in_c_h_w_n_global_desc.Get1dIndex(
0, hi_block_data_begin, wi_block_data_begin, n_block_data_begin);
const Float* p_wei_global_block_offset =
p_wei_global + wei_c_y_x_k_global_desc.Get1dIndex(0, 0, 0, k_block_data_begin);
......@@ -275,39 +277,40 @@ struct GridwiseConvolutionImplicitGemm_v1r1_chwn_cyxk_khwn
const index_t wo_thread_data_begin = c_thread_mtx_begin.col / NPerBlock;
const index_t n_thread_data_begin = c_thread_mtx_begin.col % NPerBlock;
static_if<GemmNPerThreadSubC <= NPerBlock>{}(
[&](auto f_dummy) { // f_dummy do nothing but perfect forwarding. Using this trick to
// make this lambda a generic lambda, so it won't be compiled until
// instantiated
static_assert((f_dummy(GemmNPerThreadSubC) <= NPerBlock &&
NPerBlock % GemmNPerThreadSubC == 0),
"wrong!");
// output is a 10d tensor
constexpr index_t N2 = GemmNPerThreadSubC;
constexpr index_t N1 = NPerBlock / N2;
constexpr index_t W2 = (GemmNLevel0Cluster * GemmNLevel1Cluster) /
f_dummy(NPerBlock / GemmNPerThreadSubC);
constexpr index_t W1 = WoPerBlock / W2;
constexpr index_t K2 = GemmMPerThreadSubC;
constexpr index_t K1 = KPerBlock / KPerThread;
constexpr auto out_10d_global_desc =
make_ConstantTensorDescriptor(Sequence<K / (K1 * K2),
K1,
K2,
Ho,
Wo / (W1 * W2),
W1,
W2,
N / f_dummy(N1 * N2),
N1,
N2>{});
constexpr auto out_10d_thread_desc = make_ConstantTensorDescriptor(
Sequence<KPerThread / K2, 1, K2, HoPerThread, 1, W1, 1, 1, 1, N2>{});
static_if<GemmNPerThreadSubC <= NPerBlock>{}([&](auto f_dummy) { // f_dummy do nothing but
// perfect forwarding.
// Using this trick to
// make this lambda a generic lambda, so it won't be compiled until
// instantiated
static_assert(
(f_dummy(GemmNPerThreadSubC) <= NPerBlock && NPerBlock % GemmNPerThreadSubC == 0),
"wrong!");
// output is a 10d tensor
constexpr index_t N2 = GemmNPerThreadSubC;
constexpr index_t N1 = NPerBlock / N2;
constexpr index_t W2 =
(GemmNLevel0Cluster * GemmNLevel1Cluster) / f_dummy(NPerBlock / GemmNPerThreadSubC);
constexpr index_t W1 = WoPerBlock / W2;
constexpr index_t K2 = GemmMPerThreadSubC;
constexpr index_t K1 = KPerBlock / KPerThread;
constexpr auto out_10d_global_desc =
make_ConstantTensorDescriptor(Sequence<K / (K1 * K2),
K1,
K2,
Ho,
Wo / (W1 * W2),
W1,
W2,
N / f_dummy(N1 * N2),
N1,
N2>{});
constexpr auto out_10d_thread_desc = make_ConstantTensorDescriptor(
Sequence<KPerThread / K2, 1, K2, HoPerThread, 1, W1, 1, 1, 1, N2>{});
#if 0
if(get_thread_local_1d_id() == 0 && get_block_1d_id() == 0)
......@@ -322,47 +325,37 @@ struct GridwiseConvolutionImplicitGemm_v1r1_chwn_cyxk_khwn
}
#endif
threadwise_nd_tensor_copy(out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global +
out_k_h_w_n_global_desc.Get1dIndex(
k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin,
n_block_data_begin + n_thread_data_begin),
out_10d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite_N>{});
})
.else_([&](auto f_dummy) {
static_assert(f_dummy(GemmNPerThreadSubC) >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0,
"wrong!");
// output is a 10d tensor
constexpr index_t N1 = NPerBlock;
constexpr index_t W3 = GemmNPerThreadSubC / NPerBlock;
constexpr index_t W2 = GemmNLevel0Cluster * GemmNLevel1Cluster;
constexpr index_t W1 = WoPerBlock / f_dummy(W2 * W3);
constexpr index_t K2 = GemmMPerThreadSubC;
constexpr index_t K1 = KPerBlock / KPerThread;
constexpr auto out_10d_global_desc =
make_ConstantTensorDescriptor(Sequence<K / (K1 * K2),
K1,
K2,
Ho,
Wo / (W1 * W2 * W3),
W1,
W2,
W3,
N / N1,
N1>{});
constexpr auto out_10d_thread_desc = make_ConstantTensorDescriptor(
Sequence<KPerThread / K2, 1, K2, HoPerThread, 1, W1, 1, W3, 1, N1>{});
threadwise_nd_tensor_copy(
out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global +
out_k_h_w_n_global_desc.Get1dIndex(k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin,
n_block_data_begin + n_thread_data_begin),
out_10d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite_N>{});
}).else_([&](auto f_dummy) {
static_assert(f_dummy(GemmNPerThreadSubC) >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0,
"wrong!");
// output is a 10d tensor
constexpr index_t N1 = NPerBlock;
constexpr index_t W3 = GemmNPerThreadSubC / NPerBlock;
constexpr index_t W2 = GemmNLevel0Cluster * GemmNLevel1Cluster;
constexpr index_t W1 = WoPerBlock / f_dummy(W2 * W3);
constexpr index_t K2 = GemmMPerThreadSubC;
constexpr index_t K1 = KPerBlock / KPerThread;
constexpr auto out_10d_global_desc = make_ConstantTensorDescriptor(
Sequence<K / (K1 * K2), K1, K2, Ho, Wo / (W1 * W2 * W3), W1, W2, W3, N / N1, N1>{});
constexpr auto out_10d_thread_desc = make_ConstantTensorDescriptor(
Sequence<KPerThread / K2, 1, K2, HoPerThread, 1, W1, 1, W3, 1, N1>{});
#if 0
if(get_thread_local_1d_id() == 0 && get_block_1d_id() == 0)
......@@ -382,17 +375,17 @@ struct GridwiseConvolutionImplicitGemm_v1r1_chwn_cyxk_khwn
}
#endif
threadwise_nd_tensor_copy(out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global +
out_k_h_w_n_global_desc.Get1dIndex(
k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin,
n_block_data_begin + n_thread_data_begin),
out_10d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite_N>{});
});
threadwise_nd_tensor_copy(
out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global +
out_k_h_w_n_global_desc.Get1dIndex(k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin,
n_block_data_begin + n_thread_data_begin),
out_10d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite_N>{});
});
}
};
src/include/gridwise_convolution_implicit_gemm_v1r2_chwn_cyxk_khwn.hip.hpp
View file @ 2603bb0f
......@@ -44,11 +44,12 @@ struct GridwiseConvolutionImplicitGemm_v1r2_chwn_cyxk_khwn
Float* const __restrict__ p_out_global) const
{
// be careful of this assertion
static_assert(NPerBlock % NPerThread == 0 && (GemmNPerThreadSubC <= NPerBlock &&
NPerBlock % GemmNPerThreadSubC == 0) ||
(GemmNPerThreadSubC >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0),
"wrong!");
static_assert(
NPerBlock % NPerThread == 0 &&
((GemmNPerThreadSubC <= NPerBlock && NPerBlock % GemmNPerThreadSubC == 0) ||
(GemmNPerThreadSubC >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0)),
"wrong!");
constexpr auto I0 = Number<0>{};
constexpr auto I1 = Number<1>{};
......@@ -125,8 +126,8 @@ struct GridwiseConvolutionImplicitGemm_v1r2_chwn_cyxk_khwn
constexpr auto out_k_h_w_n_thread_desc = make_ConstantTensorDescriptor(
Sequence<KPerThread, HoPerThread, WoPerThread, NPerThread>{});
// blockwise copy
// input: format is [C, Hi, Wi, N]
// blockwise copy
// input: format is [C, Hi, Wi, N]
#if 1
const auto blockwise_in_copy =
Blockwise4dTensorCopy1<BlockSize,
......@@ -228,8 +229,9 @@ struct GridwiseConvolutionImplicitGemm_v1r2_chwn_cyxk_khwn
#if 1
const Float* p_in_global_block_offset =
p_in_global + in_c_h_w_n_global_desc.Get1dIndex(
0, hi_block_data_begin, wi_block_data_begin, n_block_data_begin);
p_in_global +
in_c_h_w_n_global_desc.Get1dIndex(
0, hi_block_data_begin, wi_block_data_begin, n_block_data_begin);
const Float* p_wei_global_block_offset =
p_wei_global + wei_c_y_x_k_global_desc.Get1dIndex(0, 0, 0, k_block_data_begin);
......@@ -273,12 +275,12 @@ struct GridwiseConvolutionImplicitGemm_v1r2_chwn_cyxk_khwn
const Float* p_wei_global_block_offset =
p_wei_global + wei_c_y_x_k_global_desc.Get1dIndex(0, y, 0, k_block_data_begin);
for(index_t c_block_data_begin = 0; c_block_data_begin < C;
for(index_t
c_block_data_begin = 0;
c_block_data_begin < C;
c_block_data_begin += CPerBlock,
p_in_global_block_offset +=
CPerBlock * in_c_h_w_n_global_desc.GetStride(I0),
p_wei_global_block_offset +=
CPerBlock * wei_c_y_x_k_global_desc.GetStride(I0))
p_in_global_block_offset += CPerBlock * in_c_h_w_n_global_desc.GetStride(I0),
p_wei_global_block_offset += CPerBlock * wei_c_y_x_k_global_desc.GetStride(I0))
{
blockwise_in_copy.Run(p_in_global_block_offset, p_in_block);
......@@ -308,39 +310,40 @@ struct GridwiseConvolutionImplicitGemm_v1r2_chwn_cyxk_khwn
const index_t wo_thread_data_begin = c_thread_mtx_begin.col / NPerBlock;
const index_t n_thread_data_begin = c_thread_mtx_begin.col % NPerBlock;
static_if<GemmNPerThreadSubC <= NPerBlock>{}(
[&](auto f_dummy) { // f_dummy do nothing but perfect forwarding. Using this trick to
// make this lambda a generic lambda, so it won't be compiled until
// instantiated
static_assert((f_dummy(GemmNPerThreadSubC) <= NPerBlock &&
NPerBlock % GemmNPerThreadSubC == 0),
"wrong!");
// output is a 10d tensor
constexpr index_t N2 = GemmNPerThreadSubC;
constexpr index_t N1 = NPerBlock / N2;
constexpr index_t W2 = (GemmNLevel0Cluster * GemmNLevel1Cluster) /
f_dummy(NPerBlock / GemmNPerThreadSubC);
constexpr index_t W1 = WoPerBlock / W2;
constexpr index_t K2 = GemmMPerThreadSubC;
constexpr index_t K1 = KPerBlock / KPerThread;
constexpr auto out_10d_global_desc =
make_ConstantTensorDescriptor(Sequence<K / (K1 * K2),
K1,
K2,
Ho,
Wo / (W1 * W2),
W1,
W2,
N / f_dummy(N1 * N2),
N1,
N2>{});
constexpr auto out_10d_thread_desc = make_ConstantTensorDescriptor(
Sequence<KPerThread / K2, 1, K2, HoPerThread, 1, W1, 1, 1, 1, N2>{});
static_if<GemmNPerThreadSubC <= NPerBlock>{}([&](auto f_dummy) { // f_dummy do nothing but
// perfect forwarding.
// Using this trick to
// make this lambda a generic lambda, so it won't be compiled until
// instantiated
static_assert(
(f_dummy(GemmNPerThreadSubC) <= NPerBlock && NPerBlock % GemmNPerThreadSubC == 0),
"wrong!");
// output is a 10d tensor
constexpr index_t N2 = GemmNPerThreadSubC;
constexpr index_t N1 = NPerBlock / N2;
constexpr index_t W2 =
(GemmNLevel0Cluster * GemmNLevel1Cluster) / f_dummy(NPerBlock / GemmNPerThreadSubC);
constexpr index_t W1 = WoPerBlock / W2;
constexpr index_t K2 = GemmMPerThreadSubC;
constexpr index_t K1 = KPerBlock / KPerThread;
constexpr auto out_10d_global_desc =
make_ConstantTensorDescriptor(Sequence<K / (K1 * K2),
K1,
K2,
Ho,
Wo / (W1 * W2),
W1,
W2,
N / f_dummy(N1 * N2),
N1,
N2>{});
constexpr auto out_10d_thread_desc = make_ConstantTensorDescriptor(
Sequence<KPerThread / K2, 1, K2, HoPerThread, 1, W1, 1, 1, 1, N2>{});
#if 0
if(get_thread_local_1d_id() == 0 && get_block_1d_id() == 0)
......@@ -355,47 +358,37 @@ struct GridwiseConvolutionImplicitGemm_v1r2_chwn_cyxk_khwn
}
#endif
threadwise_nd_tensor_copy(out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global +
out_k_h_w_n_global_desc.Get1dIndex(
k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin,
n_block_data_begin + n_thread_data_begin),
out_10d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite_N>{});
})
.else_([&](auto f_dummy) {
static_assert(f_dummy(GemmNPerThreadSubC) >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0,
"wrong!");
// output is a 10d tensor
constexpr index_t N1 = NPerBlock;
constexpr index_t W3 = GemmNPerThreadSubC / NPerBlock;
constexpr index_t W2 = GemmNLevel0Cluster * GemmNLevel1Cluster;
constexpr index_t W1 = WoPerBlock / f_dummy(W2 * W3);
constexpr index_t K2 = GemmMPerThreadSubC;
constexpr index_t K1 = KPerBlock / KPerThread;
constexpr auto out_10d_global_desc =
make_ConstantTensorDescriptor(Sequence<K / (K1 * K2),
K1,
K2,
Ho,
Wo / (W1 * W2 * W3),
W1,
W2,
W3,
N / N1,
N1>{});
constexpr auto out_10d_thread_desc = make_ConstantTensorDescriptor(
Sequence<KPerThread / K2, 1, K2, HoPerThread, 1, W1, 1, W3, 1, N1>{});
threadwise_nd_tensor_copy(
out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global +
out_k_h_w_n_global_desc.Get1dIndex(k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin,
n_block_data_begin + n_thread_data_begin),
out_10d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite_N>{});
}).else_([&](auto f_dummy) {
static_assert(f_dummy(GemmNPerThreadSubC) >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0,
"wrong!");
// output is a 10d tensor
constexpr index_t N1 = NPerBlock;
constexpr index_t W3 = GemmNPerThreadSubC / NPerBlock;
constexpr index_t W2 = GemmNLevel0Cluster * GemmNLevel1Cluster;
constexpr index_t W1 = WoPerBlock / f_dummy(W2 * W3);
constexpr index_t K2 = GemmMPerThreadSubC;
constexpr index_t K1 = KPerBlock / KPerThread;
constexpr auto out_10d_global_desc = make_ConstantTensorDescriptor(
Sequence<K / (K1 * K2), K1, K2, Ho, Wo / (W1 * W2 * W3), W1, W2, W3, N / N1, N1>{});
constexpr auto out_10d_thread_desc = make_ConstantTensorDescriptor(
Sequence<KPerThread / K2, 1, K2, HoPerThread, 1, W1, 1, W3, 1, N1>{});
#if 0
if(get_thread_local_1d_id() == 0 && get_block_1d_id() == 0)
......@@ -415,17 +408,17 @@ struct GridwiseConvolutionImplicitGemm_v1r2_chwn_cyxk_khwn
}
#endif
threadwise_nd_tensor_copy(out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global +
out_k_h_w_n_global_desc.Get1dIndex(
k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin,
n_block_data_begin + n_thread_data_begin),
out_10d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite_N>{});
});
threadwise_nd_tensor_copy(
out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global +
out_k_h_w_n_global_desc.Get1dIndex(k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin,
n_block_data_begin + n_thread_data_begin),
out_10d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite_N>{});
});
}
};
src/include/gridwise_convolution_implicit_gemm_v1r2_nchw_cyxk_khwn.hip.hpp
View file @ 2603bb0f
......@@ -49,8 +49,11 @@ struct GridwiseConvolutionImplicitGemm_v1r2_nchw_cyxk_khwn
{
// be careful of this assertion
static_assert(
NPerThread <= NPerBlock && NPerBlock % NPerThread == 0,
"wrong! should satisfy: NPerThread <= NPerBlock && NPerBlock % NPerThread == 0");
NPerBlock % NPerThread == 0 &&
((GemmNPerThreadSubC <= NPerBlock && NPerBlock % GemmNPerThreadSubC == 0) ||
(GemmNPerThreadSubC >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0)),
"wrong!");
constexpr auto I0 = Number<0>{};
constexpr auto I1 = Number<1>{};
......@@ -262,12 +265,12 @@ struct GridwiseConvolutionImplicitGemm_v1r2_nchw_cyxk_khwn
const Float* p_wei_global_block_offset =
p_wei_global + wei_c_y_x_k_global_desc.Get1dIndex(0, y, 0, k_block_data_begin);
for(index_t c_block_data_begin = 0; c_block_data_begin < C;
for(index_t
c_block_data_begin = 0;
c_block_data_begin < C;
c_block_data_begin += CPerBlock,
p_in_global_block_offset +=
CPerBlock * in_n_c_h_w_global_desc.GetStride(I1),
p_wei_global_block_offset +=
CPerBlock * wei_c_y_x_k_global_desc.GetStride(I0))
p_in_global_block_offset += CPerBlock * in_n_c_h_w_global_desc.GetStride(I1),
p_wei_global_block_offset += CPerBlock * wei_c_y_x_k_global_desc.GetStride(I0))
{
Float p_in_clipboard[blockwise_in_copy_reorder.GetRegisterClipboardSize()];
Float p_wei_clipboard[blockwise_wei_copy.GetRegisterClipboardSize()];
......@@ -333,15 +336,16 @@ struct GridwiseConvolutionImplicitGemm_v1r2_nchw_cyxk_khwn
}
#endif
threadwise_10d_tensor_copy(out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global + out_k_h_w_n_global_desc.Get1dIndex(
k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin,
n_block_data_begin + n_thread_data_begin),
out_10d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite_N>{});
threadwise_10d_tensor_copy(
out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global +
out_k_h_w_n_global_desc.Get1dIndex(k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin,
n_block_data_begin + n_thread_data_begin),
out_10d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite_N>{});
}
};
src/include/gridwise_convolution_implicit_gemm_v1r3_chwn_cyxk_khwn.hip.hpp
View file @ 2603bb0f
......@@ -43,11 +43,12 @@ struct GridwiseConvolutionImplicitGemm_v1r3_chwn_cyxk_khwn
Float* const __restrict__ p_out_global) const
{
// be careful of this assertion
static_assert(NPerBlock % NPerThread == 0 && (GemmNPerThreadSubC <= NPerBlock &&
NPerBlock % GemmNPerThreadSubC == 0) ||
(GemmNPerThreadSubC >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0),
"wrong!");
static_assert(
NPerBlock % NPerThread == 0 &&
((GemmNPerThreadSubC <= NPerBlock && NPerBlock % GemmNPerThreadSubC == 0) ||
(GemmNPerThreadSubC >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0)),
"wrong!");
constexpr auto I0 = Number<0>{};
constexpr auto I1 = Number<1>{};
......@@ -212,10 +213,11 @@ struct GridwiseConvolutionImplicitGemm_v1r3_chwn_cyxk_khwn
// set threadwise output tensor to 0
threadwise_4d_tensor_set_zero(out_k_h_w_n_thread_desc, p_out_thread);
#if 1
#if 0
const Float* p_in_global_block_offset =
p_in_global + in_c_h_w_n_global_desc.Get1dIndex(
0, hi_block_data_begin, wi_block_data_begin, n_block_data_begin);
p_in_global +
in_c_h_w_n_global_desc.Get1dIndex(
0, hi_block_data_begin, wi_block_data_begin, n_block_data_begin);
const Float* p_wei_global_block_offset =
p_wei_global + wei_c_y_x_k_global_desc.Get1dIndex(0, 0, 0, k_block_data_begin);
......@@ -226,6 +228,7 @@ struct GridwiseConvolutionImplicitGemm_v1r3_chwn_cyxk_khwn
{
for(index_t y = 0; y < Y; ++y)
{
#pragma unroll
for(index_t x = 0; x < X; ++x)
{
blockwise_in_copy.Run(p_in_global_block_offset +
......@@ -287,39 +290,40 @@ struct GridwiseConvolutionImplicitGemm_v1r3_chwn_cyxk_khwn
const index_t wo_thread_data_begin = c_thread_mtx_begin.col / NPerBlock;
const index_t n_thread_data_begin = c_thread_mtx_begin.col % NPerBlock;
static_if<GemmNPerThreadSubC <= NPerBlock>{}(
[&](auto f_dummy) { // f_dummy do nothing but perfect forwarding. Using this trick to
// make this lambda a generic lambda, so it won't be compiled until
// instantiated
static_assert((f_dummy(GemmNPerThreadSubC) <= NPerBlock &&
NPerBlock % GemmNPerThreadSubC == 0),
"wrong!");
// output is a 10d tensor
constexpr index_t N2 = GemmNPerThreadSubC;
constexpr index_t N1 = NPerBlock / N2;
constexpr index_t W2 = (GemmNLevel0Cluster * GemmNLevel1Cluster) /
f_dummy(NPerBlock / GemmNPerThreadSubC);
constexpr index_t W1 = WoPerBlock / W2;
constexpr index_t K2 = GemmMPerThreadSubC;
constexpr index_t K1 = KPerBlock / KPerThread;
constexpr auto out_10d_global_desc =
make_ConstantTensorDescriptor(Sequence<K / (K1 * K2),
K1,
K2,
Ho,
Wo / (W1 * W2),
W1,
W2,
N / f_dummy(N1 * N2),
N1,
N2>{});
constexpr auto out_10d_thread_desc = make_ConstantTensorDescriptor(
Sequence<KPerThread / K2, 1, K2, HoPerThread, 1, W1, 1, 1, 1, N2>{});
static_if<GemmNPerThreadSubC <= NPerBlock>{}([&](auto f_dummy) { // f_dummy do nothing but
// perfect forwarding.
// Using this trick to
// make this lambda a generic lambda, so it won't be compiled until
// instantiated
static_assert(
(f_dummy(GemmNPerThreadSubC) <= NPerBlock && NPerBlock % GemmNPerThreadSubC == 0),
"wrong!");
// output is a 10d tensor
constexpr index_t N2 = GemmNPerThreadSubC;
constexpr index_t N1 = NPerBlock / N2;
constexpr index_t W2 =
(GemmNLevel0Cluster * GemmNLevel1Cluster) / f_dummy(NPerBlock / GemmNPerThreadSubC);
constexpr index_t W1 = WoPerBlock / W2;
constexpr index_t K2 = GemmMPerThreadSubC;
constexpr index_t K1 = KPerBlock / KPerThread;
constexpr auto out_10d_global_desc =
make_ConstantTensorDescriptor(Sequence<K / (K1 * K2),
K1,
K2,
Ho,
Wo / (W1 * W2),
W1,
W2,
N / f_dummy(N1 * N2),
N1,
N2>{});
constexpr auto out_10d_thread_desc = make_ConstantTensorDescriptor(
Sequence<KPerThread / K2, 1, K2, HoPerThread, 1, W1, 1, 1, 1, N2>{});
#if 0
if(get_thread_local_1d_id() == 0 && get_block_1d_id() == 0)
......@@ -334,47 +338,37 @@ struct GridwiseConvolutionImplicitGemm_v1r3_chwn_cyxk_khwn
}
#endif
threadwise_nd_tensor_copy(out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global +
out_k_h_w_n_global_desc.Get1dIndex(
k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin,
n_block_data_begin + n_thread_data_begin),
out_10d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite_N>{});
})
.else_([&](auto f_dummy) {
static_assert(f_dummy(GemmNPerThreadSubC) >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0,
"wrong!");
// output is a 10d tensor
constexpr index_t N1 = NPerBlock;
constexpr index_t W3 = GemmNPerThreadSubC / NPerBlock;
constexpr index_t W2 = GemmNLevel0Cluster * GemmNLevel1Cluster;
constexpr index_t W1 = WoPerBlock / f_dummy(W2 * W3);
constexpr index_t K2 = GemmMPerThreadSubC;
constexpr index_t K1 = KPerBlock / KPerThread;
constexpr auto out_10d_global_desc =
make_ConstantTensorDescriptor(Sequence<K / (K1 * K2),
K1,
K2,
Ho,
Wo / (W1 * W2 * W3),
W1,
W2,
W3,
N / N1,
N1>{});
constexpr auto out_10d_thread_desc = make_ConstantTensorDescriptor(
Sequence<KPerThread / K2, 1, K2, HoPerThread, 1, W1, 1, W3, 1, N1>{});
threadwise_nd_tensor_copy(
out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global +
out_k_h_w_n_global_desc.Get1dIndex(k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin,
n_block_data_begin + n_thread_data_begin),
out_10d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite_N>{});
}).else_([&](auto f_dummy) {
static_assert(f_dummy(GemmNPerThreadSubC) >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0,
"wrong!");
// output is a 10d tensor
constexpr index_t N1 = NPerBlock;
constexpr index_t W3 = GemmNPerThreadSubC / NPerBlock;
constexpr index_t W2 = GemmNLevel0Cluster * GemmNLevel1Cluster;
constexpr index_t W1 = WoPerBlock / f_dummy(W2 * W3);
constexpr index_t K2 = GemmMPerThreadSubC;
constexpr index_t K1 = KPerBlock / KPerThread;
constexpr auto out_10d_global_desc = make_ConstantTensorDescriptor(
Sequence<K / (K1 * K2), K1, K2, Ho, Wo / (W1 * W2 * W3), W1, W2, W3, N / N1, N1>{});
constexpr auto out_10d_thread_desc = make_ConstantTensorDescriptor(
Sequence<KPerThread / K2, 1, K2, HoPerThread, 1, W1, 1, W3, 1, N1>{});
#if 0
if(get_thread_local_1d_id() == 0 && get_block_1d_id() == 0)
......@@ -394,17 +388,17 @@ struct GridwiseConvolutionImplicitGemm_v1r3_chwn_cyxk_khwn
}
#endif
threadwise_nd_tensor_copy(out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global +
out_k_h_w_n_global_desc.Get1dIndex(
k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin,
n_block_data_begin + n_thread_data_begin),
out_10d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite_N>{});
});
threadwise_nd_tensor_copy(
out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global +
out_k_h_w_n_global_desc.Get1dIndex(k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin,
n_block_data_begin + n_thread_data_begin),
out_10d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite_N>{});
});
}
};
src/include/gridwise_convolution_implicit_gemm_v1r3_lds_double_buffer_chwn_cyxk_khwn.hip.hpp
View file @ 2603bb0f
......@@ -43,11 +43,12 @@ struct GridwiseConvolutionImplicitGemm_v1r3_lds_double_buffer_chwn_cyxk_khwn
Float* const __restrict__ p_out_global) const
{
// be careful of this assertion
static_assert(NPerBlock % NPerThread == 0 && (GemmNPerThreadSubC <= NPerBlock &&
NPerBlock % GemmNPerThreadSubC == 0) ||
(GemmNPerThreadSubC >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0),
"wrong!");
static_assert(
NPerBlock % NPerThread == 0 &&
((GemmNPerThreadSubC <= NPerBlock && NPerBlock % GemmNPerThreadSubC == 0) ||
(GemmNPerThreadSubC >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0)),
"wrong!");
constexpr auto I0 = Number<0>{};
constexpr auto I1 = Number<1>{};
......@@ -127,8 +128,8 @@ struct GridwiseConvolutionImplicitGemm_v1r3_lds_double_buffer_chwn_cyxk_khwn
constexpr auto out_k_h_w_n_thread_desc = make_ConstantTensorDescriptor(
Sequence<KPerThread, HoPerThread, WoPerThread, NPerThread>{});
// blockwise copy
// input: format is [C, Hi, Wi, N]
// blockwise copy
// input: format is [C, Hi, Wi, N]
#if 0
const auto blockwise_in_copy =
Blockwise4dTensorCopy1<BlockSize,
......@@ -349,39 +350,40 @@ struct GridwiseConvolutionImplicitGemm_v1r3_lds_double_buffer_chwn_cyxk_khwn
const index_t wo_thread_data_begin = c_thread_mtx_begin.col / NPerBlock;
const index_t n_thread_data_begin = c_thread_mtx_begin.col % NPerBlock;
static_if<GemmNPerThreadSubC <= NPerBlock>{}(
[&](auto f_dummy) { // f_dummy do nothing but perfect forwarding. Using this trick to
// make this lambda a generic lambda, so it won't be compiled until
// instantiated
static_assert((f_dummy(GemmNPerThreadSubC) <= NPerBlock &&
NPerBlock % GemmNPerThreadSubC == 0),
"wrong!");
// output is a 10d tensor
constexpr index_t N2 = GemmNPerThreadSubC;
constexpr index_t N1 = NPerBlock / N2;
constexpr index_t W2 = (GemmNLevel0Cluster * GemmNLevel1Cluster) /
f_dummy(NPerBlock / GemmNPerThreadSubC);
constexpr index_t W1 = WoPerBlock / W2;
constexpr index_t K2 = GemmMPerThreadSubC;
constexpr index_t K1 = KPerBlock / KPerThread;
constexpr auto out_10d_global_desc =
make_ConstantTensorDescriptor(Sequence<K / (K1 * K2),
K1,
K2,
Ho,
Wo / (W1 * W2),
W1,
W2,
N / f_dummy(N1 * N2),
N1,
N2>{});
constexpr auto out_10d_thread_desc = make_ConstantTensorDescriptor(
Sequence<KPerThread / K2, 1, K2, HoPerThread, 1, W1, 1, 1, 1, N2>{});
static_if<GemmNPerThreadSubC <= NPerBlock>{}([&](auto f_dummy) { // f_dummy do nothing but
// perfect forwarding.
// Using this trick to
// make this lambda a generic lambda, so it won't be compiled until
// instantiated
static_assert(
(f_dummy(GemmNPerThreadSubC) <= NPerBlock && NPerBlock % GemmNPerThreadSubC == 0),
"wrong!");
// output is a 10d tensor
constexpr index_t N2 = GemmNPerThreadSubC;
constexpr index_t N1 = NPerBlock / N2;
constexpr index_t W2 =
(GemmNLevel0Cluster * GemmNLevel1Cluster) / f_dummy(NPerBlock / GemmNPerThreadSubC);
constexpr index_t W1 = WoPerBlock / W2;
constexpr index_t K2 = GemmMPerThreadSubC;
constexpr index_t K1 = KPerBlock / KPerThread;
constexpr auto out_10d_global_desc =
make_ConstantTensorDescriptor(Sequence<K / (K1 * K2),
K1,
K2,
Ho,
Wo / (W1 * W2),
W1,
W2,
N / f_dummy(N1 * N2),
N1,
N2>{});
constexpr auto out_10d_thread_desc = make_ConstantTensorDescriptor(
Sequence<KPerThread / K2, 1, K2, HoPerThread, 1, W1, 1, 1, 1, N2>{});
#if 0
if(get_thread_local_1d_id() == 0 && get_block_1d_id() == 0)
......@@ -396,47 +398,37 @@ struct GridwiseConvolutionImplicitGemm_v1r3_lds_double_buffer_chwn_cyxk_khwn
}
#endif
threadwise_nd_tensor_copy(out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global +
out_k_h_w_n_global_desc.Get1dIndex(
k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin,
n_block_data_begin + n_thread_data_begin),
out_10d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite_N>{});
})
.else_([&](auto f_dummy) {
static_assert(f_dummy(GemmNPerThreadSubC) >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0,
"wrong!");
// output is a 10d tensor
constexpr index_t N1 = NPerBlock;
constexpr index_t W3 = GemmNPerThreadSubC / NPerBlock;
constexpr index_t W2 = GemmNLevel0Cluster * GemmNLevel1Cluster;
constexpr index_t W1 = WoPerBlock / f_dummy(W2 * W3);
constexpr index_t K2 = GemmMPerThreadSubC;
constexpr index_t K1 = KPerBlock / KPerThread;
constexpr auto out_10d_global_desc =
make_ConstantTensorDescriptor(Sequence<K / (K1 * K2),
K1,
K2,
Ho,
Wo / (W1 * W2 * W3),
W1,
W2,
W3,
N / N1,
N1>{});
constexpr auto out_10d_thread_desc = make_ConstantTensorDescriptor(
Sequence<KPerThread / K2, 1, K2, HoPerThread, 1, W1, 1, W3, 1, N1>{});
threadwise_nd_tensor_copy(
out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global +
out_k_h_w_n_global_desc.Get1dIndex(k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin,
n_block_data_begin + n_thread_data_begin),
out_10d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite_N>{});
}).else_([&](auto f_dummy) {
static_assert(f_dummy(GemmNPerThreadSubC) >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0,
"wrong!");
// output is a 10d tensor
constexpr index_t N1 = NPerBlock;
constexpr index_t W3 = GemmNPerThreadSubC / NPerBlock;
constexpr index_t W2 = GemmNLevel0Cluster * GemmNLevel1Cluster;
constexpr index_t W1 = WoPerBlock / f_dummy(W2 * W3);
constexpr index_t K2 = GemmMPerThreadSubC;
constexpr index_t K1 = KPerBlock / KPerThread;
constexpr auto out_10d_global_desc = make_ConstantTensorDescriptor(
Sequence<K / (K1 * K2), K1, K2, Ho, Wo / (W1 * W2 * W3), W1, W2, W3, N / N1, N1>{});
constexpr auto out_10d_thread_desc = make_ConstantTensorDescriptor(
Sequence<KPerThread / K2, 1, K2, HoPerThread, 1, W1, 1, W3, 1, N1>{});
#if 0
if(get_thread_local_1d_id() == 0 && get_block_1d_id() == 0)
......@@ -456,17 +448,17 @@ struct GridwiseConvolutionImplicitGemm_v1r3_lds_double_buffer_chwn_cyxk_khwn
}
#endif
threadwise_nd_tensor_copy(out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global +
out_k_h_w_n_global_desc.Get1dIndex(
k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin,
n_block_data_begin + n_thread_data_begin),
out_10d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite_N>{});
});
threadwise_nd_tensor_copy(
out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global +
out_k_h_w_n_global_desc.Get1dIndex(k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin,
n_block_data_begin + n_thread_data_begin),
out_10d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite_N>{});
});
}
};
src/include/gridwise_convolution_implicit_gemm_v1r3_lds_double_buffer_nchw_cyxk_khwn.hip.hpp
View file @ 2603bb0f
......@@ -47,11 +47,12 @@ struct GridwiseConvolutionImplicitGemm_v1r3_lds_double_buffer_nchw_cyxk_khwn
Float* const __restrict__ p_out_global) const
{
// be careful of this assertion
static_assert(NPerBlock % NPerThread == 0 && (GemmNPerThreadSubC <= NPerBlock &&
NPerBlock % GemmNPerThreadSubC == 0) ||
(GemmNPerThreadSubC >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0),
"wrong!");
static_assert(
NPerBlock % NPerThread == 0 &&
((GemmNPerThreadSubC <= NPerBlock && NPerBlock % GemmNPerThreadSubC == 0) ||
(GemmNPerThreadSubC >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0)),
"wrong!");
constexpr auto I0 = Number<0>{};
constexpr auto I1 = Number<1>{};
......@@ -349,39 +350,40 @@ struct GridwiseConvolutionImplicitGemm_v1r3_lds_double_buffer_nchw_cyxk_khwn
const index_t wo_thread_data_begin = c_thread_mtx_begin.col / NPerBlock;
const index_t n_thread_data_begin = c_thread_mtx_begin.col % NPerBlock;
static_if<GemmNPerThreadSubC <= NPerBlock>{}(
[&](auto f_dummy) { // f_dummy do nothing but perfect forwarding. Using this trick to
// make this lambda a generic lambda, so it won't be compiled until
// instantiated
static_assert((f_dummy(GemmNPerThreadSubC) <= NPerBlock &&
NPerBlock % GemmNPerThreadSubC == 0),
"wrong!");
// output is a 10d tensor
constexpr index_t N2 = GemmNPerThreadSubC;
constexpr index_t N1 = NPerBlock / N2;
constexpr index_t W2 = (GemmNLevel0Cluster * GemmNLevel1Cluster) /
f_dummy(NPerBlock / GemmNPerThreadSubC);
constexpr index_t W1 = WoPerBlock / W2;
constexpr index_t K2 = GemmMPerThreadSubC;
constexpr index_t K1 = KPerBlock / KPerThread;
constexpr auto out_10d_global_desc =
make_ConstantTensorDescriptor(Sequence<K / (K1 * K2),
K1,
K2,
Ho,
Wo / (W1 * W2),
W1,
W2,
N / f_dummy(N1 * N2),
N1,
N2>{});
constexpr auto out_10d_thread_desc = make_ConstantTensorDescriptor(
Sequence<KPerThread / K2, 1, K2, HoPerThread, 1, W1, 1, 1, 1, N2>{});
static_if<GemmNPerThreadSubC <= NPerBlock>{}([&](auto f_dummy) { // f_dummy do nothing but
// perfect forwarding.
// Using this trick to
// make this lambda a generic lambda, so it won't be compiled until
// instantiated
static_assert(
(f_dummy(GemmNPerThreadSubC) <= NPerBlock && NPerBlock % GemmNPerThreadSubC == 0),
"wrong!");
// output is a 10d tensor
constexpr index_t N2 = GemmNPerThreadSubC;
constexpr index_t N1 = NPerBlock / N2;
constexpr index_t W2 =
(GemmNLevel0Cluster * GemmNLevel1Cluster) / f_dummy(NPerBlock / GemmNPerThreadSubC);
constexpr index_t W1 = WoPerBlock / W2;
constexpr index_t K2 = GemmMPerThreadSubC;
constexpr index_t K1 = KPerBlock / KPerThread;
constexpr auto out_10d_global_desc =
make_ConstantTensorDescriptor(Sequence<K / (K1 * K2),
K1,
K2,
Ho,
Wo / (W1 * W2),
W1,
W2,
N / f_dummy(N1 * N2),
N1,
N2>{});
constexpr auto out_10d_thread_desc = make_ConstantTensorDescriptor(
Sequence<KPerThread / K2, 1, K2, HoPerThread, 1, W1, 1, 1, 1, N2>{});
#if 0
if(get_thread_local_1d_id() == 0 && get_block_1d_id() == 0)
......@@ -396,47 +398,37 @@ struct GridwiseConvolutionImplicitGemm_v1r3_lds_double_buffer_nchw_cyxk_khwn
}
#endif
threadwise_nd_tensor_copy(out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global +
out_k_h_w_n_global_desc.Get1dIndex(
k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin,
n_block_data_begin + n_thread_data_begin),
out_10d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite_N>{});
})
.else_([&](auto f_dummy) {
static_assert(f_dummy(GemmNPerThreadSubC) >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0,
"wrong!");
// output is a 10d tensor
constexpr index_t N1 = NPerBlock;
constexpr index_t W3 = GemmNPerThreadSubC / NPerBlock;
constexpr index_t W2 = GemmNLevel0Cluster * GemmNLevel1Cluster;
constexpr index_t W1 = WoPerBlock / f_dummy(W2 * W3);
constexpr index_t K2 = GemmMPerThreadSubC;
constexpr index_t K1 = KPerBlock / KPerThread;
constexpr auto out_10d_global_desc =
make_ConstantTensorDescriptor(Sequence<K / (K1 * K2),
K1,
K2,
Ho,
Wo / (W1 * W2 * W3),
W1,
W2,
W3,
N / N1,
N1>{});
constexpr auto out_10d_thread_desc = make_ConstantTensorDescriptor(
Sequence<KPerThread / K2, 1, K2, HoPerThread, 1, W1, 1, W3, 1, N1>{});
threadwise_nd_tensor_copy(
out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global +
out_k_h_w_n_global_desc.Get1dIndex(k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin,
n_block_data_begin + n_thread_data_begin),
out_10d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite_N>{});
}).else_([&](auto f_dummy) {
static_assert(f_dummy(GemmNPerThreadSubC) >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0,
"wrong!");
// output is a 10d tensor
constexpr index_t N1 = NPerBlock;
constexpr index_t W3 = GemmNPerThreadSubC / NPerBlock;
constexpr index_t W2 = GemmNLevel0Cluster * GemmNLevel1Cluster;
constexpr index_t W1 = WoPerBlock / f_dummy(W2 * W3);
constexpr index_t K2 = GemmMPerThreadSubC;
constexpr index_t K1 = KPerBlock / KPerThread;
constexpr auto out_10d_global_desc = make_ConstantTensorDescriptor(
Sequence<K / (K1 * K2), K1, K2, Ho, Wo / (W1 * W2 * W3), W1, W2, W3, N / N1, N1>{});
constexpr auto out_10d_thread_desc = make_ConstantTensorDescriptor(
Sequence<KPerThread / K2, 1, K2, HoPerThread, 1, W1, 1, W3, 1, N1>{});
#if 0
if(get_thread_local_1d_id() == 0 && get_block_1d_id() == 0)
......@@ -456,17 +448,17 @@ struct GridwiseConvolutionImplicitGemm_v1r3_lds_double_buffer_nchw_cyxk_khwn
}
#endif
threadwise_nd_tensor_copy(out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global +
out_k_h_w_n_global_desc.Get1dIndex(
k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin,
n_block_data_begin + n_thread_data_begin),
out_10d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite_N>{});
});
threadwise_nd_tensor_copy(
out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global +
out_k_h_w_n_global_desc.Get1dIndex(k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin,
n_block_data_begin + n_thread_data_begin),
out_10d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite_N>{});
});
}
};
src/include/gridwise_convolution_implicit_gemm_v1r3_nchw_cyxk_khwn.hip.hpp
View file @ 2603bb0f
......@@ -47,11 +47,12 @@ struct GridwiseConvolutionImplicitGemm_v1r3_nchw_cyxk_khwn
Float* const __restrict__ p_out_global) const
{
// be careful of this assertion
static_assert(NPerBlock % NPerThread == 0 && (GemmNPerThreadSubC <= NPerBlock &&
NPerBlock % GemmNPerThreadSubC == 0) ||
(GemmNPerThreadSubC >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0),
"wrong!");
static_assert(
NPerBlock % NPerThread == 0 &&
((GemmNPerThreadSubC <= NPerBlock && NPerBlock % GemmNPerThreadSubC == 0) ||
(GemmNPerThreadSubC >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0)),
"wrong!");
constexpr auto I0 = Number<0>{};
constexpr auto I1 = Number<1>{};
......@@ -223,8 +224,9 @@ struct GridwiseConvolutionImplicitGemm_v1r3_nchw_cyxk_khwn
#if 1
const Float* p_in_global_block_offset =
p_in_global + in_n_c_h_w_global_desc.Get1dIndex(
n_block_data_begin, 0, hi_block_data_begin, wi_block_data_begin);
p_in_global +
in_n_c_h_w_global_desc.Get1dIndex(
n_block_data_begin, 0, hi_block_data_begin, wi_block_data_begin);
const Float* p_wei_global_block_offset =
p_wei_global + wei_c_y_x_k_global_desc.Get1dIndex(0, 0, 0, k_block_data_begin);
......@@ -329,39 +331,40 @@ struct GridwiseConvolutionImplicitGemm_v1r3_nchw_cyxk_khwn
const index_t wo_thread_data_begin = c_thread_mtx_begin.col / NPerBlock;
const index_t n_thread_data_begin = c_thread_mtx_begin.col % NPerBlock;
static_if<GemmNPerThreadSubC <= NPerBlock>{}(
[&](auto f_dummy) { // f_dummy do nothing but perfect forwarding. Using this trick to
// make this lambda a generic lambda, so it won't be compiled until
// instantiated
static_assert((f_dummy(GemmNPerThreadSubC) <= NPerBlock &&
NPerBlock % GemmNPerThreadSubC == 0),
"wrong!");
// output is a 10d tensor
constexpr index_t N2 = GemmNPerThreadSubC;
constexpr index_t N1 = NPerBlock / N2;
constexpr index_t W2 = (GemmNLevel0Cluster * GemmNLevel1Cluster) /
f_dummy(NPerBlock / GemmNPerThreadSubC);
constexpr index_t W1 = WoPerBlock / W2;
constexpr index_t K2 = GemmMPerThreadSubC;
constexpr index_t K1 = KPerBlock / KPerThread;
constexpr auto out_10d_global_desc =
make_ConstantTensorDescriptor(Sequence<K / (K1 * K2),
K1,
K2,
Ho,
Wo / (W1 * W2),
W1,
W2,
N / f_dummy(N1 * N2),
N1,
N2>{});
constexpr auto out_10d_thread_desc = make_ConstantTensorDescriptor(
Sequence<KPerThread / K2, 1, K2, HoPerThread, 1, W1, 1, 1, 1, N2>{});
static_if<GemmNPerThreadSubC <= NPerBlock>{}([&](auto f_dummy) { // f_dummy do nothing but
// perfect forwarding.
// Using this trick to
// make this lambda a generic lambda, so it won't be compiled until
// instantiated
static_assert(
(f_dummy(GemmNPerThreadSubC) <= NPerBlock && NPerBlock % GemmNPerThreadSubC == 0),
"wrong!");
// output is a 10d tensor
constexpr index_t N2 = GemmNPerThreadSubC;
constexpr index_t N1 = NPerBlock / N2;
constexpr index_t W2 =
(GemmNLevel0Cluster * GemmNLevel1Cluster) / f_dummy(NPerBlock / GemmNPerThreadSubC);
constexpr index_t W1 = WoPerBlock / W2;
constexpr index_t K2 = GemmMPerThreadSubC;
constexpr index_t K1 = KPerBlock / KPerThread;
constexpr auto out_10d_global_desc =
make_ConstantTensorDescriptor(Sequence<K / (K1 * K2),
K1,
K2,
Ho,
Wo / (W1 * W2),
W1,
W2,
N / f_dummy(N1 * N2),
N1,
N2>{});
constexpr auto out_10d_thread_desc = make_ConstantTensorDescriptor(
Sequence<KPerThread / K2, 1, K2, HoPerThread, 1, W1, 1, 1, 1, N2>{});
#if 0
if(get_thread_local_1d_id() == 0 && get_block_1d_id() == 0)
......@@ -376,47 +379,37 @@ struct GridwiseConvolutionImplicitGemm_v1r3_nchw_cyxk_khwn
}
#endif
threadwise_nd_tensor_copy(out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global +
out_k_h_w_n_global_desc.Get1dIndex(
k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin,
n_block_data_begin + n_thread_data_begin),
out_10d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite_N>{});
})
.else_([&](auto f_dummy) {
static_assert(f_dummy(GemmNPerThreadSubC) >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0,
"wrong!");
// output is a 10d tensor
constexpr index_t N1 = NPerBlock;
constexpr index_t W3 = GemmNPerThreadSubC / NPerBlock;
constexpr index_t W2 = GemmNLevel0Cluster * GemmNLevel1Cluster;
constexpr index_t W1 = WoPerBlock / f_dummy(W2 * W3);
constexpr index_t K2 = GemmMPerThreadSubC;
constexpr index_t K1 = KPerBlock / KPerThread;
constexpr auto out_10d_global_desc =
make_ConstantTensorDescriptor(Sequence<K / (K1 * K2),
K1,
K2,
Ho,
Wo / (W1 * W2 * W3),
W1,
W2,
W3,
N / N1,
N1>{});
constexpr auto out_10d_thread_desc = make_ConstantTensorDescriptor(
Sequence<KPerThread / K2, 1, K2, HoPerThread, 1, W1, 1, W3, 1, N1>{});
threadwise_nd_tensor_copy(
out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global +
out_k_h_w_n_global_desc.Get1dIndex(k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin,
n_block_data_begin + n_thread_data_begin),
out_10d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite_N>{});
}).else_([&](auto f_dummy) {
static_assert(f_dummy(GemmNPerThreadSubC) >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0,
"wrong!");
// output is a 10d tensor
constexpr index_t N1 = NPerBlock;
constexpr index_t W3 = GemmNPerThreadSubC / NPerBlock;
constexpr index_t W2 = GemmNLevel0Cluster * GemmNLevel1Cluster;
constexpr index_t W1 = WoPerBlock / f_dummy(W2 * W3);
constexpr index_t K2 = GemmMPerThreadSubC;
constexpr index_t K1 = KPerBlock / KPerThread;
constexpr auto out_10d_global_desc = make_ConstantTensorDescriptor(
Sequence<K / (K1 * K2), K1, K2, Ho, Wo / (W1 * W2 * W3), W1, W2, W3, N / N1, N1>{});
constexpr auto out_10d_thread_desc = make_ConstantTensorDescriptor(
Sequence<KPerThread / K2, 1, K2, HoPerThread, 1, W1, 1, W3, 1, N1>{});
#if 0
if(get_thread_local_1d_id() == 0 && get_block_1d_id() == 0)
......@@ -436,17 +429,17 @@ struct GridwiseConvolutionImplicitGemm_v1r3_nchw_cyxk_khwn
}
#endif
threadwise_nd_tensor_copy(out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global +
out_k_h_w_n_global_desc.Get1dIndex(
k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin,
n_block_data_begin + n_thread_data_begin),
out_10d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite_N>{});
});
threadwise_nd_tensor_copy(
out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global +
out_k_h_w_n_global_desc.Get1dIndex(k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin,
n_block_data_begin + n_thread_data_begin),
out_10d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite_N>{});
});
}
};
src/include/gridwise_convolution_implicit_gemm_v1r3_nchw_cyxk_nkhw.hip.hpp
View file @ 2603bb0f
......@@ -47,11 +47,12 @@ struct GridwiseConvolutionImplicitGemm_v1r3_nchw_cyxk_nkhw
Float* const __restrict__ p_out_global) const
{
// be careful of this assertion
static_assert(NPerBlock % NPerThread == 0 && (GemmNPerThreadSubC <= NPerBlock &&
NPerBlock % GemmNPerThreadSubC == 0) ||
(GemmNPerThreadSubC >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0),
"wrong!");
static_assert(
NPerBlock % NPerThread == 0 &&
((GemmNPerThreadSubC <= NPerBlock && NPerBlock % GemmNPerThreadSubC == 0) ||
(GemmNPerThreadSubC >= NPerBlock && NPerThread == NPerBlock &&
GemmNPerThreadSubC % NPerThread == 0)),
"wrong!");
constexpr auto I0 = Number<0>{};
constexpr auto I1 = Number<1>{};
......@@ -223,8 +224,9 @@ struct GridwiseConvolutionImplicitGemm_v1r3_nchw_cyxk_nkhw
#if 1
const Float* p_in_global_block_offset =
p_in_global + in_n_c_h_w_global_desc.Get1dIndex(
n_block_data_begin, 0, hi_block_data_begin, wi_block_data_begin);
p_in_global +
in_n_c_h_w_global_desc.Get1dIndex(
n_block_data_begin, 0, hi_block_data_begin, wi_block_data_begin);
const Float* p_wei_global_block_offset =
p_wei_global + wei_c_y_x_k_global_desc.Get1dIndex(0, 0, 0, k_block_data_begin);
......@@ -409,14 +411,15 @@ struct GridwiseConvolutionImplicitGemm_v1r3_nchw_cyxk_nkhw
out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global + out_n_k_h_w_global_desc.Get1dIndex(
n_block_data_begin + n_thread_data_begin,
k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin),
p_out_global +
out_n_k_h_w_global_desc.Get1dIndex(
n_block_data_begin + n_thread_data_begin,
k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin),
out_10d_thread_desc.GetLengths(),
map_out_global2thread);
// Number<OutThreadCopyDataPerWrite_W>{});
// Number<OutThreadCopyDataPerWrite_W>{});
#endif
})
.else_([&](auto f_dummy) {
......@@ -500,14 +503,15 @@ struct GridwiseConvolutionImplicitGemm_v1r3_nchw_cyxk_nkhw
out_10d_thread_desc,
p_out_thread,
out_10d_global_desc,
p_out_global + out_n_k_h_w_global_desc.Get1dIndex(
n_block_data_begin + n_thread_data_begin,
k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin),
p_out_global +
out_n_k_h_w_global_desc.Get1dIndex(
n_block_data_begin + n_thread_data_begin,
k_block_data_begin + k_thread_data_begin,
ho_block_data_begin + ho_thread_data_begin,
wo_block_data_begin + wo_thread_data_begin),
out_10d_thread_desc.GetLengths(),
map_out_global2thread);
// Number<OutThreadCopyDataPerWrite_W>{});
// Number<OutThreadCopyDataPerWrite_W>{});
#endif
});
}
......
src/include/gridwise_convolution_implicit_gemm_v2_chwn_cyxk_khwn_lds_double_buffer.hip.hpp
View file @ 2603bb0f
......@@ -365,13 +365,14 @@ struct GridwiseConvolutionImplicitGemm_v2_chwn_cyxk_khwn_lds_double_buffer
constexpr auto out_kb_global_desc = make_ConstantTensorDescriptor(Sequence<K, B>{});
threadwise_6d_tensor_copy(out_6d_thread_desc,
p_out_thread,
out_6d_global_desc,
p_out_global + out_kb_global_desc.Get1dIndex(
k_thread_data_begin, b_thread_data_begin),
out_6d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite>{});
threadwise_6d_tensor_copy(
out_6d_thread_desc,
p_out_thread,
out_6d_global_desc,
p_out_global +
out_kb_global_desc.Get1dIndex(k_thread_data_begin, b_thread_data_begin),
out_6d_thread_desc.GetLengths(),
Number<OutThreadCopyDataPerWrite>{});
}
else
{
......
src/include/gridwise_direct_convolution_1.hip.hpp
View file @ 2603bb0f
......@@ -113,10 +113,11 @@ __global__ void gridwise_direct_convolution_1(const Float* const __restrict__ p_
c_block_work_begin += CPerBlock)
{
// copy input tensor to LDS
blockwise_in_copy.Run(p_in_global + in_global_desc.Get1dIndex(n_block_work_begin,
c_block_work_begin,
hi_block_work_begin,
wi_block_work_begin),
blockwise_in_copy.Run(p_in_global +
in_global_desc.Get1dIndex(n_block_work_begin,
c_block_work_begin,
hi_block_work_begin,
wi_block_work_begin),
p_in_block);
// copy weight tensor to LDS
......@@ -143,9 +144,9 @@ __global__ void gridwise_direct_convolution_1(const Float* const __restrict__ p_
}
// copy output tensor from LDS to device mem
blockwise_out_copy.Run(p_out_block,
p_out_global + out_global_desc.Get1dIndex(n_block_work_begin,
k_block_work_begin,
ho_block_work_begin,
wo_block_work_begin));
blockwise_out_copy.Run(
p_out_block,
p_out_global +
out_global_desc.Get1dIndex(
n_block_work_begin, k_block_work_begin, ho_block_work_begin, wo_block_work_begin));
}
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