Skip to content

CI测试 update kernels funcs #2515

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
co63oc wants to merge 3 commits into
base: develop
Choose a base branch
from
Open

CI测试 update kernels funcs #2515

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
2db3a8b
update kernels funcs
co63oc Apr 22, 2026
2377508
ci
co63oc Apr 22, 2026
796dcab
ci
co63oc Apr 22, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
26 changes: 13 additions & 13 deletions backends/mlu/kernels/funcs/elementwise_utils.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -120,11 +120,11 @@ inline void GetReduceAxesAndDstDims(const int axis,

template <typename T>
void MLUOpTensorKernel(const Context& dev_ctx,
const phi::DenseTensor& x,
const phi::DenseTensor& y,
const DenseTensor& x,
const DenseTensor& y,
int axis,
const cnnlOpTensorDesc_t op_tensor_type,
phi::DenseTensor* out) {
DenseTensor* out) {
PADDLE_ENFORCE_EQ((op_tensor_type == CNNL_OP_TENSOR_ADD) ||
(op_tensor_type == CNNL_OP_TENSOR_SUB) ||
(op_tensor_type == CNNL_OP_TENSOR_MUL),
Expand Down Expand Up @@ -241,10 +241,10 @@ inline void MLUBinary<POW>(const Context& dev_ctx,

template <BINARY_FUNCTOR Functor, typename T>
void MLUBinaryOp(const Context& dev_ctx,
const phi::DenseTensor& x,
const phi::DenseTensor& y,
const DenseTensor& x,
const DenseTensor& y,
int axis,
phi::DenseTensor* out) {
DenseTensor* out) {
dev_ctx.template Alloc<T>(out);
Tensor x_t, y_t;
x_t = x;
Expand Down Expand Up @@ -319,8 +319,8 @@ inline void MLUUnary<RECIPROCAL>(const Context& dev_ctx,

template <UNARY_FUNCTOR Functor, typename Tin, typename Tout = Tin>
void MLUUnaryOp(const Context& dev_ctx,
const phi::DenseTensor& x,
phi::DenseTensor* out) {
const DenseTensor& x,
DenseTensor* out) {
dev_ctx.template Alloc<Tout>(out);

MLUCnnlTensorDesc x_desc(x, CNNL_LAYOUT_ARRAY, ToCnnlDataType<Tin>());
Expand All @@ -342,12 +342,12 @@ enum MINMAX_GRAD_FUNCTOR {
};
template <MINMAX_GRAD_FUNCTOR Functor, typename Tin, typename Tout = Tin>
void MLUMinMaxGradHelper(const Context& dev_ctx,
const phi::DenseTensor& x,
const phi::DenseTensor& y,
const phi::DenseTensor& dout,
const DenseTensor& x,
const DenseTensor& y,
const DenseTensor& dout,
int axis,
phi::DenseTensor* dx,
phi::DenseTensor* dy) {
DenseTensor* dx,
DenseTensor* dy) {
const auto& x_dims = x.dims();
const auto& y_dims = y.dims();
axis =
Expand Down
6 changes: 3 additions & 3 deletions backends/mlu/kernels/funcs/logic_op.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,10 +21,10 @@ namespace custom_kernel {

template <typename Context>
void MLULogicOp(const Context& dev_ctx,
const phi::DenseTensor& x,
const phi::DenseTensor& y,
const DenseTensor& x,
const DenseTensor& y,
const std::string& logic_name,
phi::DenseTensor* out) {
DenseTensor* out) {
dev_ctx.template Alloc<bool>(out);

MLUCnnlTensorDesc input_x(x, CNNL_LAYOUT_ARRAY, ToCnnlDataType(x.dtype()));
Expand Down
4 changes: 4 additions & 0 deletions backends/mlu/kernels/funcs/mlu_baseop.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,6 +25,10 @@
namespace custom_kernel {

using Tensor = phi::DenseTensor;
using DenseTensor = phi::DenseTensor;
using DenseTensorMeta = phi::DenseTensorMeta;
using Scalar = phi::Scalar;
using DDim = phi::DDim;
using Context = phi::CustomContext;
using DataType = phi::DataType;
using DataLayout = phi::DataLayout;
Expand Down
32 changes: 15 additions & 17 deletions backends/mlu/kernels/funcs/mlu_funcs.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,9 +25,9 @@ namespace custom_kernel {
*/
template <typename Context>
inline void TensorCopy(const Context& dev_ctx,
const phi::DenseTensor& src,
const DenseTensor& src,
bool blocking,
phi::DenseTensor* dst,
DenseTensor* dst,
const phi::Place& dst_place = phi::CustomPlace()) {
dev_ctx.Wait();
auto* src_ptr = src.data();
Expand Down Expand Up @@ -103,7 +103,7 @@ template <typename T>
inline void TensorFromVector(const phi::CustomContext& ctx,
const std::vector<T>& src,
const phi::CustomContext& dev_ctx,
phi::DenseTensor* dst) {
DenseTensor* dst) {
auto dst_place = dev_ctx.GetPlace();
C_Device_st device{dst_place.GetDeviceId()};
auto src_ptr = static_cast<const void*>(src.data());
Expand All @@ -128,7 +128,7 @@ template <>
inline void TensorFromVector<bool>(const phi::CustomContext& ctx,
const std::vector<bool>& src,
const phi::CustomContext& dev_ctx,
phi::DenseTensor* dst) {
DenseTensor* dst) {
// vector<bool> has no data() member, use array instead.
// See details:
// https://stackoverflow.com/questions/46115669/why-does-stdvectorbool-have-no-data/46115714
Expand Down Expand Up @@ -166,7 +166,7 @@ template <typename T>
inline void TensorFromVector(const phi::CustomContext& ctx,
const std::vector<T>& src,
const phi::CPUContext& dev_ctx,
phi::DenseTensor* dst) {
DenseTensor* dst) {
auto dst_place = dev_ctx.GetPlace();
C_Device_st device{dst_place.GetDeviceId()};
auto src_ptr = static_cast<const void*>(src.data());
Expand All @@ -191,7 +191,7 @@ template <>
inline void TensorFromVector<bool>(const phi::CustomContext& ctx,
const std::vector<bool>& src,
const phi::CPUContext& dev_ctx,
phi::DenseTensor* dst) {
DenseTensor* dst) {
auto dst_place = dev_ctx.GetPlace();
PADDLE_THROW(phi::errors::Unimplemented(
"TensorFromVector on %s is not supported.", dst_place));
Expand All @@ -202,7 +202,7 @@ void TensorFromArray(const phi::CustomContext& ctx,
const T* src,
const size_t& array_size,
const phi::CustomContext& dev_ctx,
phi::DenseTensor* dst) {
DenseTensor* dst) {
auto dst_place = dev_ctx.GetPlace();
C_Device_st device{dst_place.GetDeviceId()};
auto src_ptr = static_cast<const void*>(src);
Expand All @@ -227,7 +227,7 @@ void TensorFromArray(const phi::CustomContext& ctx,
*/
template <typename T>
inline void TensorToVector(const phi::CustomContext& ctx,
const phi::DenseTensor& src,
const DenseTensor& src,
const phi::CustomContext& dev_ctx,
std::vector<T>* dst) {
auto src_ptr = static_cast<const void*>(src.data<T>());
Expand All @@ -251,7 +251,7 @@ inline void TensorToVector(const phi::CustomContext& ctx,

template <>
inline void TensorToVector<bool>(const phi::CustomContext& ctx,
const phi::DenseTensor& src,
const DenseTensor& src,
const phi::CustomContext& dev_ctx,
std::vector<bool>* dst) {
auto src_ptr = static_cast<const void*>(src.data<bool>());
Expand Down Expand Up @@ -359,11 +359,10 @@ inline void ExtractNCDWH(const phi::DDim& dims,

template <typename T>
inline std::vector<T> get_new_data_from_tensor(
const phi::CustomContext& dev_ctx,
const phi::DenseTensor* new_data_tensor) {
const phi::CustomContext& dev_ctx, const DenseTensor* new_data_tensor) {
std::vector<T> vec_new_data;
auto place = new_data_tensor->place();
phi::DenseTensor cpu_starts_tensor;
DenseTensor cpu_starts_tensor;
if (place.GetType() == phi::AllocationType::CUSTOM) {
// if tensor on CUSTOM place, do memcpy to host
cpu_starts_tensor.Resize(new_data_tensor->dims());
Expand All @@ -381,22 +380,21 @@ inline std::vector<T> get_new_data_from_tensor(
}

template <typename T>
inline phi::DenseTensor ReshapeToMatrix(const phi::DenseTensor& src,
T num_col_dims) {
inline DenseTensor ReshapeToMatrix(const DenseTensor& src, T num_col_dims) {
int rank = src.dims().size();
PADDLE_ENFORCE_GE(
rank,
2,
phi::errors::InvalidArgument(
"'ReshapeToMatrix()' is only used for flatten high rank "
"tensors to matrixs. The dimensions of phi::DenseTensor must be "
"tensors to matrixs. The dimensions of DenseTensor must be "
"greater or equal than 2. "
"But received dimensions of phi::DenseTensor is %d",
"But received dimensions of DenseTensor is %d",
rank));
if (rank == 2) {
return src;
}
phi::DenseTensor res;
DenseTensor res;
res = src;
res.Resize(phi::flatten_to_2d(src.dims(), num_col_dims));
return res;
Expand Down
2 changes: 1 addition & 1 deletion backends/mlu/kernels/funcs/range_op.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,7 +24,7 @@ void ArangeRawKernel(const Context& dev_ctx,
const T start_value,
const T end_value,
const T step_value,
phi::DenseTensor* out) {
DenseTensor* out) {
int64_t size = 0;
GetSize(start_value, end_value, step_value, &size);

Expand Down
4 changes: 2 additions & 2 deletions backends/mlu/kernels/funcs/reduce_op.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,12 +21,12 @@ namespace custom_kernel {

template <typename T, typename Context>
void MLUReduceOp(const Context& dev_ctx,
const phi::DenseTensor& x,
const DenseTensor& x,
const std::vector<int64_t>& axes,
bool keep_dim,
bool reduce_all,
const std::string& reduce_name,
phi::DenseTensor* out) {
DenseTensor* out) {
dev_ctx.template Alloc<T>(out);
if (x.dims().size() == 0) {
TensorCopy(dev_ctx, x, true, out);
Expand Down
Loading