Skip to content

optimize phi::CPUPlace in kernels/impl/ #77204

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

Merged
luotao1 merged 1 commit into from
Jan 6, 2026
Merged

optimize phi::CPUPlace in kernels/impl/ #77204

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
2 changes: 1 addition & 1 deletion paddle/phi/kernels/impl/accuracy_check_kernel_impl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -251,7 +251,7 @@ struct AccuracyCheckFunctor<phi::GPUContext, T> {
in_data, other_data, rtol, atol, equal_nan, num, out_data);

DenseTensor out_cpu;
Copy(dev_ctx, *output, phi::CPUPlace(), true, &out_cpu);
Copy(dev_ctx, *output, CPUPlace(), true, &out_cpu);
auto data_ptr = out_cpu.data<bool>();

PADDLE_ENFORCE_EQ(*data_ptr,
Expand Down
7 changes: 3 additions & 4 deletions paddle/phi/kernels/impl/average_accumulates_kernel_impl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -48,18 +48,17 @@ void AverageAccumulatesKernel(const Context& dev_ctx,
// int64_t num_accumulates = 0;
// int64_t old_num_accumulates = 0;

auto num_updates_cpu =
phi::memory_utils::Alloc(phi::CPUPlace(), sizeof(int64_t));
auto num_updates_cpu = phi::memory_utils::Alloc(CPUPlace(), sizeof(int64_t));
int64_t* num_updates_cpu_ptr =
reinterpret_cast<int64_t*>(num_updates_cpu->ptr());

auto num_accumulates_cpu =
phi::memory_utils::Alloc(phi::CPUPlace(), sizeof(int64_t));
phi::memory_utils::Alloc(CPUPlace(), sizeof(int64_t));
int64_t* num_accumulates_cpu_ptr =
reinterpret_cast<int64_t*>(num_accumulates_cpu->ptr());

auto old_num_accumulates_cpu =
phi::memory_utils::Alloc(phi::CPUPlace(), sizeof(int64_t));
phi::memory_utils::Alloc(CPUPlace(), sizeof(int64_t));
int64_t* old_num_accumulates_cpu_ptr =
reinterpret_cast<int64_t*>(old_num_accumulates_cpu->ptr());

Expand Down
4 changes: 2 additions & 2 deletions paddle/phi/kernels/impl/beam_search_decode_kernel_impl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -49,7 +49,7 @@ struct BeamSearchDecodeFunctor {
if (tensor_on_gpu_) {
dev_ctx->Wait();
}
Copy(*dev_ctx, step_id, phi::CPUPlace(), false, &out);
Copy(*dev_ctx, step_id, CPUPlace(), false, &out);
dev_ctx->Wait();
}

Expand All @@ -72,7 +72,7 @@ struct BeamSearchDecodeFunctor {
if (tensor_on_gpu_) {
dev_ctx->Wait();
}
Copy(*dev_ctx, step_score, phi::CPUPlace(), false, &out);
Copy(*dev_ctx, step_score, CPUPlace(), false, &out);
dev_ctx->Wait();
}

Expand Down
2 changes: 1 addition & 1 deletion paddle/phi/kernels/impl/clip_by_norm_kernel_impl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -45,7 +45,7 @@ void ClipByNormFunctor(const Context& dev_ctx,
temp + (static_cast<T>(1) - temp) * max_norm / (x_norm + epsilon);
Eigen::array<int, 1> one_dim{{1}};
Eigen::DSizes<int, 1> m_dsize(input->numel());
if (dev_ctx.GetPlace() == phi::CPUPlace()) {
if (dev_ctx.GetPlace() == CPUPlace()) {
out.device(*place) = x * scaling.reshape(one_dim).eval().broadcast(m_dsize);
} else {
out.device(*place) = x * scaling.reshape(one_dim).broadcast(m_dsize);
Expand Down
2 changes: 1 addition & 1 deletion paddle/phi/kernels/impl/determinant_grad_kernel_impl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -92,7 +92,7 @@ inline bool CheckMatrixInvertible(const Context& dev_ctx,

// copy to host
DenseTensor cpu_tensor;
Copy<Context>(dev_ctx, dev_tensor, phi::CPUPlace(), false, &cpu_tensor);
Copy<Context>(dev_ctx, dev_tensor, CPUPlace(), false, &cpu_tensor);

// if founded zero, the matrix is not invertible
// else the matrix is invertible
Expand Down
4 changes: 2 additions & 2 deletions paddle/phi/kernels/impl/fetch_impl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,7 +26,7 @@ void FetchKernel(const Context& dev_ctx,
if (!x.IsInitialized()) {
return;
}
Copy(dev_ctx, x, phi::CPUPlace(), true, out);
Copy(dev_ctx, x, CPUPlace(), true, out);
}

template <typename T, typename Context>
Expand All @@ -35,7 +35,7 @@ void FetchArrayKernel(const Context& dev_ctx,
TensorArray* out) {
out->resize(x.size());
for (size_t i = 0; i < x.size(); ++i) {
Copy(dev_ctx, x[i], phi::CPUPlace(), true, &(out->at(i)));
Copy(dev_ctx, x[i], CPUPlace(), true, &(out->at(i)));
}
}

Expand Down
2 changes: 1 addition & 1 deletion paddle/phi/kernels/impl/gru_unit_kernel_impl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,7 +35,7 @@ void ActCompute(
} else if (act_type == tanh) {
funcs::TanhFunctor<T>()(d, x, y);
} else if (act_type == relu) {
if (place == phi::CPUPlace())
if (place == CPUPlace())
funcs::ReluCPUFunctor<T>()(d, x, y);
else
funcs::ReluCUDAFunctor<T>()(d, x, y);
Expand Down
2 changes: 1 addition & 1 deletion paddle/phi/kernels/impl/im2sequence_kernel_impl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -50,7 +50,7 @@ void Im2SequenceKernel(const Context& dev_ctx,
const DenseTensor* img_real_size = y.get_ptr();

DenseTensor cpu_shape_tensor;
Copy(dev_ctx, *img_real_size, phi::CPUPlace(), true, &cpu_shape_tensor);
Copy(dev_ctx, *img_real_size, CPUPlace(), true, &cpu_shape_tensor);
std::vector<int> img_real_h;
std::vector<int> img_real_w;
std::vector<int> output_height;
Expand Down
2 changes: 1 addition & 1 deletion paddle/phi/kernels/impl/isclose_kernel_impl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -267,7 +267,7 @@ struct GetTensorValue<phi::GPUContext, T> {
T value;
const auto gpu_place = dev_ctx.GetPlace();
memory_utils::Copy(
phi::CPUPlace(), &value, gpu_place, data, sizeof(T), dev_ctx.stream());
CPUPlace(), &value, gpu_place, data, sizeof(T), dev_ctx.stream());
return value;
}
};
Expand Down
3 changes: 1 addition & 2 deletions paddle/phi/kernels/impl/lamb_kernel_impl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -186,8 +186,7 @@ void ComputeImpl(const Context& dev_ctx,
// Diff from here

if (dev_ctx.GetPlace().GetType() == AllocationType::GPU &&
beta1_pow.place() == phi::CPUPlace() &&
beta2_pow.place() == phi::CPUPlace()) {
beta1_pow.place() == CPUPlace() && beta2_pow.place() == CPUPlace()) {
LambMomentREGUpdateFunctor<T, IsMultiPrecision> moment_update_functor(
weight_decay,
beta1,
Expand Down
2 changes: 1 addition & 1 deletion paddle/phi/kernels/impl/lod_reset_kernel_impl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -56,7 +56,7 @@ void LodResetKernel(const Context& dev_ctx,
auto* lod = lod_t->data<int>();
DenseTensor lod_cpu;
if (lod_t->place().GetType() == phi::AllocationType::GPU) {
Copy(dev_ctx, *lod_t, phi::CPUPlace(), true, &lod_cpu);
Copy(dev_ctx, *lod_t, CPUPlace(), true, &lod_cpu);
lod = lod_cpu.data<int>();
}
level0 = std::vector<int>(lod, lod + lod_t->numel());
Expand Down
4 changes: 2 additions & 2 deletions paddle/phi/kernels/impl/lstsq_kernel_impl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -233,7 +233,7 @@ inline void BatchedOrmqr<GPUContext, float>(const GPUContext& dev_ctx,

// check the error info
int info_h;
memory_utils::Copy(phi::CPUPlace(),
memory_utils::Copy(CPUPlace(),
&info_h,
dev_ctx.GetPlace(),
info_d,
Expand Down Expand Up @@ -302,7 +302,7 @@ inline void BatchedOrmqr<GPUContext, double>(const GPUContext& dev_ctx,

// check the error info
int info_h;
memory_utils::Copy(phi::CPUPlace(),
memory_utils::Copy(CPUPlace(),
&info_h,
dev_ctx.GetPlace(),
info_d,
Expand Down
2 changes: 1 addition & 1 deletion paddle/phi/kernels/impl/lu_kernel_impl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -478,7 +478,7 @@ void Unpack_Pivot(const Context& dev_ctx,
auto prank = Pdimvec.size();
auto Pnum = dims[prank - 1];
DenseTensor Pivot_cpu;
phi::CPUPlace cpu;
CPUPlace cpu;
Copy(dev_ctx, Pivot, cpu, false, &Pivot_cpu);
auto pdataptr = Pivot_cpu.data<int32_t>();
Pdimvec[prank - 1] = h;
Expand Down
2 changes: 1 addition & 1 deletion paddle/phi/kernels/impl/numel_kernel_impl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,7 +25,7 @@ void NumelKernel(const Context& dev_ctx,
auto place = dev_ctx.GetPlace();
auto out_data = dev_ctx.template Alloc<int64_t>(out);

if (place == phi::CPUPlace()) {
if (place == CPUPlace()) {
out_data[0] = input.numel();
} else {
DenseTensor cpu_tensor;
Expand Down
2 changes: 1 addition & 1 deletion paddle/phi/kernels/impl/sequence_mask_kernel_impl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -78,7 +78,7 @@ void SequenceMaskKernel(const Context& dev_ctx,
bool is_gpu_place = dev_ctx.GetPlace().GetType() == AllocationType::GPU;
if (is_gpu_place) {
DenseTensor temp;
Copy(dev_ctx, *max_len_tensor.get_ptr(), phi::CPUPlace(), false, &temp);
Copy(dev_ctx, *max_len_tensor.get_ptr(), CPUPlace(), false, &temp);
maxlen = *temp.data<int32_t>();
} else {
maxlen = *max_len_tensor.get_ptr()->data<int32_t>();
Expand Down
2 changes: 1 addition & 1 deletion paddle/phi/kernels/impl/sequence_pool_kernel_impl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -65,7 +65,7 @@ void SequencePoolKernel(const Context& dev_ctx,

// Do not create index buffer for inference mode
if (pooltype == "MAX" &&
(is_test == false || (dev_ctx.GetPlace() == phi::CPUPlace()) == false)) {
(is_test == false || (dev_ctx.GetPlace() == CPUPlace()) == false)) {
index = max_index;
index->Resize({dims});
dev_ctx.template Alloc<int32_t>(index);
Expand Down
12 changes: 6 additions & 6 deletions paddle/phi/kernels/impl/warpctc_kernel_impl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -276,8 +276,8 @@ void WarpctcKernel(const Context& dev_ctx,

DenseTensor logits_length_cpu;
DenseTensor labels_length_cpu;
Copy(dev_ctx, *logits_length, phi::CPUPlace(), false, &logits_length_cpu);
Copy(dev_ctx, *labels_length, phi::CPUPlace(), false, &labels_length_cpu);
Copy(dev_ctx, *logits_length, CPUPlace(), false, &logits_length_cpu);
Copy(dev_ctx, *labels_length, CPUPlace(), false, &labels_length_cpu);

logits_lod.push_back(0);
label_lod.push_back(0);
Expand Down Expand Up @@ -358,7 +358,7 @@ void WarpctcKernel(const Context& dev_ctx,
T* pad_value_data = dev_ctx.template HostAlloc<T>(&cpu_pad_value);
*pad_value_data = static_cast<T>(0);
DenseTensor pad_value;
if (dev_ctx.GetPlace() == phi::CPUPlace()) {
if (dev_ctx.GetPlace() == CPUPlace()) {
pad_value = cpu_pad_value;
} else {
Copy(dev_ctx, cpu_pad_value, dev_ctx.GetPlace(), true, &pad_value);
Expand Down Expand Up @@ -401,7 +401,7 @@ void WarpctcKernel(const Context& dev_ctx,
lod.push_back(label_lod);
warpctc_label.set_lod(lod);

if (dev_ctx.GetPlace() == phi::CPUPlace()) {
if (dev_ctx.GetPlace() == CPUPlace()) {
funcs::UnpaddingDenseTensorFunctor<Context, int>()(
dev_ctx,
label,
Expand All @@ -424,10 +424,10 @@ void WarpctcKernel(const Context& dev_ctx,
0 /*lod_level*/,
false /*norm_by_times*/,
funcs::kBatchLengthWidth);
Copy(dev_ctx, gpu_label, phi::CPUPlace(), true, &warpctc_label);
Copy(dev_ctx, gpu_label, CPUPlace(), true, &warpctc_label);
}
} else {
Copy(dev_ctx, label, phi::CPUPlace(), true, &warpctc_label);
Copy(dev_ctx, label, CPUPlace(), true, &warpctc_label);
}

const int* warpctc_label_data = warpctc_label.data<int>();
Expand Down
Loading