Skip to content

Migrate backward warp kernel arguments to use PTA_B #4825

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
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Migrate backward warp kernel arguments to use PTA_B #4825

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
8 changes: 4 additions & 4 deletions fbgemm_gpu/codegen/training/backward/embedding_backward_split_kernel_warp_template.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -137,7 +137,7 @@ batch_index_select_dim0_codegen_backward_kernel_warp_per_row(
const bool enable_optimizer_offloading,
{%- endif %}
{%- if is_index_select %}
const at::PackedTensorAccessor32<int64_t, 1, at::RestrictPtrTraits> grad_offsets,
const pta::PackedTensorAccessor32<int64_t, 1, at::RestrictPtrTraits> grad_offsets,
const bool permute_output_dim_0_1
{%- else %}
{{ args.split_kernel_args | replace_pta_namespace() | join(",\n ") }}
Expand Down Expand Up @@ -436,7 +436,7 @@ batch_index_select_dim0_codegen_backward_kernel_warp_per_row
const bool enable_optimizer_offloading,
{%- endif %}
{%- if is_index_select %}
const at::PackedTensorAccessor32<int64_t, 1, at::RestrictPtrTraits> grad_offsets,
const pta::PackedTensorAccessor32<int64_t, 1, at::RestrictPtrTraits> grad_offsets,
const bool permute_output_dim_0_1
{%- else %}
{{ args.split_kernel_args_no_defaults |
Expand Down Expand Up @@ -606,7 +606,7 @@ hip_split_embedding{{ ndesc }}_backward_codegen_{{ optimizer }}_{{ wdesc }}{{ vd
const int32_t max_D,
const int32_t max_vecs_per_thread,
{%- if is_index_select %}
const at::PackedTensorAccessor32<int64_t, 1, at::RestrictPtrTraits> grad_offsets,
const pta::PackedTensorAccessor32<int64_t, 1, at::RestrictPtrTraits> grad_offsets,
const bool permute_output_dim_0_1
{%- else %}
{{ args.split_kernel_args | replace_pta_namespace() | join(",\n ") }}
Expand Down Expand Up @@ -771,7 +771,7 @@ hip_split_embedding{{ ndesc }}_backward_codegen_{{ optimizer }}_{{ wdesc }}{{ vd
const int32_t max_D,
const int32_t max_vecs_per_thread,
{%- if is_index_select %}
const at::PackedTensorAccessor32<int64_t, 1, at::RestrictPtrTraits> grad_offsets,
const pta::PackedTensorAccessor32<int64_t, 1, at::RestrictPtrTraits> grad_offsets,
const bool permute_output_dim_0_1
{%- else %}
{{ args.split_kernel_args_no_defaults | replace_pta_namespace() | join(",\n ") | replace("cache_t", cache_type) }}
Expand Down
8 changes: 3 additions & 5 deletions fbgemm_gpu/codegen/training/backward/embedding_backward_split_template.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -220,7 +220,7 @@ batch_index_select_dim0_codegen_backward_kernel_warp_per_row(
const bool enable_optimizer_offloading,
{%- endif %}
{%- if is_index_select %}
const at::PackedTensorAccessor32<int64_t, 1, at::RestrictPtrTraits> grad_offsets,
const pta::PackedTensorAccessor32<int64_t, 1, at::RestrictPtrTraits> grad_offsets,
const bool permute_output_dim_0_1
{%- else %}
{{ args.split_kernel_args | replace_pta_namespace() | join(",\n ") }}
Expand Down Expand Up @@ -292,7 +292,7 @@ hip_split_embedding{{ ndesc }}_backward_codegen_{{ optimizer }}_{{ wdesc }}{{ vd
const int32_t max_D,
const int32_t max_vecs_per_thread,
{%- if is_index_select %}
const at::PackedTensorAccessor32<int64_t, 1, at::RestrictPtrTraits> grad_offsets,
const pta::PackedTensorAccessor32<int64_t, 1, at::RestrictPtrTraits> grad_offsets,
const bool permute_output_dim_0_1
{%- else %}
{{ args.split_kernel_args | replace_pta_namespace() | join(",\n ") }}
Expand Down Expand Up @@ -966,7 +966,6 @@ Tensor {{ embedding_cuda_op }}(
{%- endif %}

DISPATCH_OPTIMAL_KERNEL(max_D, [&] {

auto long_run_ids = at::empty({indices.numel()}, sorted_linear_indices_run_lengths.options());
auto num_long_run_ids = at::zeros({1}, indices.options().dtype(at::kInt));

Expand All @@ -982,7 +981,6 @@ Tensor {{ embedding_cuda_op }}(
at::empty({indices.numel()}, sorted_linear_indices_run_lengths.options());
}


auto num_really_long_run_ids = at::zeros({1}, indices.options().dtype(at::kInt));
auto grad_accum_counter = at::empty(
use_deterministic_algorithms ? 0 : (indices.numel() / max_segment_length_per_cta),
Expand Down Expand Up @@ -1292,7 +1290,7 @@ Tensor {{ embedding_cuda_op }}(
enable_optimizer_offloading,
{%- endif %}
{%- if is_index_select %}
grad_offsets.packed_accessor32<int64_t, 1, at::RestrictPtrTraits>(),
PTA_B(grad_offsets, int64_t, 1, 32),
permute_output_dim_0_1
{%- else %}
{{ args.split_kernel_arg_constructors | make_pta_acc_builder_format() | join(",\n ") }}
Expand Down
Loading