ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-12 09:16:52 +00:00
Code Issues Packages Projects Releases Wiki Activity

update to clang-format-10

Browse Source
This commit is contained in:
Chao Liu
2021-07-30 16:37:00 -05:00
parent bd27ed6c38
commit 82fae390fb
10 changed files with 38 additions and 50 deletions
Download Patch File Download Diff File Expand all files Collapse all files

26
composable_kernel/include/tensor_description/dynamic_tensor_descriptor.hpp
View File

@@ -33,13 +33,11 @@ struct DynamicTensorDescriptor
__host__ __device__ static constexpr index_t GetNumOfHiddenDimension()
{
constexpr auto all_low_dim_ids =
unpack([](auto&&... xs) constexpr { return merge_sequences(xs...); },
LowerDimensionIdss{});
constexpr auto all_low_dim_ids = unpack(
[](auto&&... xs) constexpr { return merge_sequences(xs...); }, LowerDimensionIdss{});
constexpr auto all_up_dim_ids =
unpack([](auto&&... xs) constexpr { return merge_sequences(xs...); },
UpperDimensionIdss{});
constexpr auto all_up_dim_ids = unpack(
[](auto&&... xs) constexpr { return merge_sequences(xs...); }, UpperDimensionIdss{});
constexpr auto all_dim_ids = merge_sequences(all_low_dim_ids, all_up_dim_ids);
@@ -347,22 +345,22 @@ transform_dynamic_tensor_descriptor(const OldTensorDescriptor& old_tensor_desc,
constexpr auto up_dim_numbers_scan = merge_sequences(
Sequence<0>{}, inclusive_scan_sequence(up_dim_numbers, math::plus<index_t>{}, Number<0>{}));
constexpr auto up_dim_hidden_idss =
generate_tuple([ old_hidden_dim_number, up_dim_numbers_scan ](auto i) constexpr {
constexpr auto up_dim_hidden_idss = generate_tuple(
[ old_hidden_dim_number, up_dim_numbers_scan ](auto i) constexpr {
return
typename arithmetic_sequence_gen<old_hidden_dim_number + up_dim_numbers_scan[i],
old_hidden_dim_number + up_dim_numbers_scan[i + 1],
1>::type{};
},
Number<num_new_transform>{});
Number<num_new_transform>{});
// new visible dimension's hidden ids
constexpr auto unordered_new_visible_dim_hidden_ids =
unpack([](auto... xs) constexpr { return merge_sequences(xs...); }, up_dim_hidden_idss);
constexpr auto unordered_new_visible_dim_hidden_ids = unpack(
[](auto... xs) constexpr { return merge_sequences(xs...); }, up_dim_hidden_idss);
constexpr auto new_visible_dim_unordered2ordered =
unpack([](auto... xs) constexpr { return merge_sequences(xs...); },
NewUpperDimensionNewVisibleIdss{});
constexpr auto new_visible_dim_unordered2ordered = unpack(
[](auto... xs) constexpr { return merge_sequences(xs...); },
NewUpperDimensionNewVisibleIdss{});
constexpr auto new_visible_dim_hidden_ids =
unordered_new_visible_dim_hidden_ids.ReorderGivenOld2New(new_visible_dim_unordered2ordered);

22
composable_kernel/include/tensor_description/tensor_adaptor.hpp
View File

@@ -106,13 +106,13 @@ struct TensorAdaptor
__host__ __device__ static constexpr index_t GetNumOfHiddenDimension()
{
constexpr auto all_low_dim_ids =
unpack([](auto&&... xs) constexpr { return merge_sequences(xs...); },
LowerDimensionHiddenIdss{});
constexpr auto all_low_dim_ids = unpack(
[](auto&&... xs) constexpr { return merge_sequences(xs...); },
LowerDimensionHiddenIdss{});
constexpr auto all_up_dim_ids =
unpack([](auto&&... xs) constexpr { return merge_sequences(xs...); },
UpperDimensionHiddenIdss{});
constexpr auto all_up_dim_ids = unpack(
[](auto&&... xs) constexpr { return merge_sequences(xs...); },
UpperDimensionHiddenIdss{});
constexpr auto all_dim_ids = merge_sequences(all_low_dim_ids, all_up_dim_ids);
@@ -418,13 +418,11 @@ __host__ __device__ constexpr auto make_single_stage_tensor_adaptor(const Transf
"wrong!");
// sanity check on LowerDimensionOldTopIdss and UpperDimensionNewTopIdss
constexpr auto all_low_dim_old_top_ids =
unpack([](auto&&... xs) constexpr { return merge_sequences(xs...); },
LowerDimensionOldTopIdss{});
constexpr auto all_low_dim_old_top_ids = unpack(
[](auto&&... xs) constexpr { return merge_sequences(xs...); }, LowerDimensionOldTopIdss{});
constexpr auto all_up_dim_new_top_ids =
unpack([](auto&&... xs) constexpr { return merge_sequences(xs...); },
UpperDimensionNewTopIdss{});
constexpr auto all_up_dim_new_top_ids = unpack(
[](auto&&... xs) constexpr { return merge_sequences(xs...); }, UpperDimensionNewTopIdss{});
static_assert(is_valid_sequence_map<decltype(all_low_dim_old_top_ids)>::value &&
is_valid_sequence_map<decltype(all_up_dim_new_top_ids)>::value,

1
composable_kernel/include/tensor_operation/blockwise_gemm_dlops_v3.hpp
View File

@@ -152,7 +152,6 @@ struct BlockwiseGemmDlops_km_kn_m0m1n0n1_v3
static_for<0, EPerBlock, EPerThreadLoop>{}([&](auto e_begin) {
static_for<0, KPerThread, KPerThreadSubC>{}([&](auto k_begin) {
a_thread_copy_.Run(a_block_mtx,
make_tuple(e_begin, k_begin),
a_block_buf,

2
composable_kernel/include/tensor_operation/threadwise_contraction_dlops.hpp
View File

@@ -87,7 +87,6 @@ struct ThreadwiseGemmDlops_km0m1_kn0n1_m0m1n0n1
static_for<0, TM1, 1>{}([&](auto tm1) {
static_for<0, TN0, 1>{}([&](auto tn0) {
static_for<0, TN1, 1>{}([&](auto tn1) {
constexpr index_t a_offset =
AThreadDesc_TK0_TM0_TM1_TK1{}.CalculateOffset(
a_origin_idx + make_multi_index(tk, tm0, tm1));
@@ -192,7 +191,6 @@ struct ThreadwiseContractionDlops_A_TK0_TM0_TM1_TK1_B_TK0_TN0_TN1_TK1_C_TM0_TM1_
static_for<0, TM1, 1>{}([&](auto tm1) {
static_for<0, TN0, 1>{}([&](auto tn0) {
static_for<0, TN1, 1>{}([&](auto tn1) {
vector_type<FloatA, TK1> a_vec;
vector_type<FloatB, TK1> b_vec;

1
composable_kernel/include/tensor_operation/threadwise_gemm_dlops_v3.hpp
View File

@@ -136,7 +136,6 @@ struct ThreadwiseGemmDlops_km_kn_mn_v3
{
static_for<0, H, 1>{}([&](auto h) {
static_for<0, W, 1>{}([&](auto w) {
constexpr index_t b_offset =
BDesc{}.CalculateOffset(b_origin_idx + make_tuple(e, 0, h, w));

3
composable_kernel/include/utility/data_type_enum.hpp
View File

@@ -4,7 +4,8 @@
namespace ck {
// this enumerate should be synchronized with include/miopen.h
typedef enum {
typedef enum
{
Half = 0,
Float = 1,
Int32 = 2,