Skip to content

Shard grouped conv bwd data instances #2925

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 6 commits into
base: develop
Choose a base branch
from
+1,463 −565
Open

Shard grouped conv bwd data instances #2925

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
0bdc767
Grouped Conv Bwd Data index calculation optimizations
bartekxk Sep 23, 2025
8596ce2
fixes
bartekxk Sep 24, 2025
1cf47ee
Shard grouped conv bwd data instances
bartekxk Sep 25, 2025
b10281c
Merge
bartekxk Oct 7, 2025
8fcbc0d
remove added instances
bartekxk Oct 7, 2025
0f67b3b
fix
bartekxk Oct 7, 2025
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
118 changes: 59 additions & 59 deletions ...ry/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_backward_data.hpp
View file
Open in desktop

Large diffs are not rendered by default.

14 changes: 7 additions & 7 deletions ...e/ck/library/tensor_operation_instance/gpu/grouped_convolution_backward_data_bilinear.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2024-2025, Advanced Micro Devices, Inc. All rights reserved.

#pragma once

Expand All @@ -16,7 +16,7 @@ namespace device {
namespace instance {

#ifdef CK_ENABLE_FP16
void add_device_grouped_conv3d_bwd_data_xdl_bilinear_ndhwgk_gkzyxc_ndhwgc_f16_instances(
void add_device_grouped_conv3d_bwd_data_xdl_bilinear_ndhwgc_gkzyxc_ndhwgk_f16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
NDHWGK,
GKZYXC,
Expand All @@ -31,7 +31,7 @@ void add_device_grouped_conv3d_bwd_data_xdl_bilinear_ndhwgk_gkzyxc_ndhwgc_f16_in
Bilinear>>>& instances);
#endif
#ifdef CK_ENABLE_FP32
void add_device_grouped_conv3d_bwd_data_xdl_bilinear_ndhwgk_gkzyxc_ndhwgc_f32_instances(
void add_device_grouped_conv3d_bwd_data_xdl_bilinear_ndhwgc_gkzyxc_ndhwgk_f32_instances(
std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
NDHWGK,
GKZYXC,
Expand All @@ -46,7 +46,7 @@ void add_device_grouped_conv3d_bwd_data_xdl_bilinear_ndhwgk_gkzyxc_ndhwgc_f32_in
Bilinear>>>& instances);
#endif
#ifdef CK_ENABLE_BF16
void add_device_grouped_conv3d_bwd_data_xdl_bilinear_ndhwgk_gkzyxc_ndhwgc_bf16_instances(
void add_device_grouped_conv3d_bwd_data_xdl_bilinear_ndhwgc_gkzyxc_ndhwgk_bf16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
NDHWGK,
GKZYXC,
Expand Down Expand Up @@ -115,7 +115,7 @@ struct DeviceOperationInstanceFactory<
is_same_v<OutDataType, F16> && is_same_v<ComputeTypeA, F16> &&
is_same_v<ComputeTypeB, F16>)
{
add_device_grouped_conv3d_bwd_data_xdl_bilinear_ndhwgk_gkzyxc_ndhwgc_f16_instances(
add_device_grouped_conv3d_bwd_data_xdl_bilinear_ndhwgc_gkzyxc_ndhwgk_f16_instances(
op_ptrs);
}
#endif
Expand All @@ -124,7 +124,7 @@ struct DeviceOperationInstanceFactory<
is_same_v<OutDataType, F32> && is_same_v<ComputeTypeA, F32> &&
is_same_v<ComputeTypeB, F32>)
{
add_device_grouped_conv3d_bwd_data_xdl_bilinear_ndhwgk_gkzyxc_ndhwgc_f32_instances(
add_device_grouped_conv3d_bwd_data_xdl_bilinear_ndhwgc_gkzyxc_ndhwgk_f32_instances(
op_ptrs);
}
#endif
Expand All @@ -133,7 +133,7 @@ struct DeviceOperationInstanceFactory<
is_same_v<OutDataType, BF16> && is_same_v<ComputeTypeA, BF16> &&
is_same_v<ComputeTypeB, BF16>)
{
add_device_grouped_conv3d_bwd_data_xdl_bilinear_ndhwgk_gkzyxc_ndhwgc_bf16_instances(
add_device_grouped_conv3d_bwd_data_xdl_bilinear_ndhwgc_gkzyxc_ndhwgk_bf16_instances(
op_ptrs);
}
#endif
Expand Down
14 changes: 7 additions & 7 deletions ...lude/ck/library/tensor_operation_instance/gpu/grouped_convolution_backward_data_scale.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2024-2025, Advanced Micro Devices, Inc. All rights reserved.

#pragma once

Expand All @@ -16,7 +16,7 @@ namespace device {
namespace instance {

#ifdef CK_ENABLE_FP16
void add_device_grouped_conv3d_bwd_data_xdl_scale_ndhwgk_gkzyxc_ndhwgc_f16_instances(
void add_device_grouped_conv3d_bwd_data_xdl_scale_ndhwgc_gkzyxc_ndhwgk_f16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
NDHWGK,
GKZYXC,
Expand All @@ -31,7 +31,7 @@ void add_device_grouped_conv3d_bwd_data_xdl_scale_ndhwgk_gkzyxc_ndhwgc_f16_insta
Scale>>>& instances);
#endif
#ifdef CK_ENABLE_FP32
void add_device_grouped_conv3d_bwd_data_xdl_scale_ndhwgk_gkzyxc_ndhwgc_f32_instances(
void add_device_grouped_conv3d_bwd_data_xdl_scale_ndhwgc_gkzyxc_ndhwgk_f32_instances(
std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
NDHWGK,
GKZYXC,
Expand All @@ -46,7 +46,7 @@ void add_device_grouped_conv3d_bwd_data_xdl_scale_ndhwgk_gkzyxc_ndhwgc_f32_insta
Scale>>>& instances);
#endif
#ifdef CK_ENABLE_BF16
void add_device_grouped_conv3d_bwd_data_xdl_scale_ndhwgk_gkzyxc_ndhwgc_bf16_instances(
void add_device_grouped_conv3d_bwd_data_xdl_scale_ndhwgc_gkzyxc_ndhwgk_bf16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
NDHWGK,
GKZYXC,
Expand Down Expand Up @@ -115,7 +115,7 @@ struct DeviceOperationInstanceFactory<
is_same_v<OutDataType, F16> && is_same_v<ComputeTypeA, F16> &&
is_same_v<ComputeTypeB, F16>)
{
add_device_grouped_conv3d_bwd_data_xdl_scale_ndhwgk_gkzyxc_ndhwgc_f16_instances(
add_device_grouped_conv3d_bwd_data_xdl_scale_ndhwgc_gkzyxc_ndhwgk_f16_instances(
op_ptrs);
}
#endif
Expand All @@ -124,7 +124,7 @@ struct DeviceOperationInstanceFactory<
is_same_v<OutDataType, F32> && is_same_v<ComputeTypeA, F32> &&
is_same_v<ComputeTypeB, F32>)
{
add_device_grouped_conv3d_bwd_data_xdl_scale_ndhwgk_gkzyxc_ndhwgc_f32_instances(
add_device_grouped_conv3d_bwd_data_xdl_scale_ndhwgc_gkzyxc_ndhwgk_f32_instances(
op_ptrs);
}
#endif
Expand All @@ -133,7 +133,7 @@ struct DeviceOperationInstanceFactory<
is_same_v<OutDataType, BF16> && is_same_v<ComputeTypeA, BF16> &&
is_same_v<ComputeTypeB, BF16>)
{
add_device_grouped_conv3d_bwd_data_xdl_scale_ndhwgk_gkzyxc_ndhwgc_bf16_instances(
add_device_grouped_conv3d_bwd_data_xdl_scale_ndhwgc_gkzyxc_ndhwgk_bf16_instances(
op_ptrs);
}
#endif
Expand Down
32 changes: 16 additions & 16 deletions ...clude/ck/library/tensor_operation_instance/gpu/grouped_convolution_backward_data_wmma.inc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,7 +10,7 @@ namespace instance {

// conv2d backward data
#ifdef CK_ENABLE_FP16
void add_device_grouped_conv2d_bwd_data_wmma_gnhwk_gkyxc_gnhwc_f16_instances(
void add_device_grouped_conv2d_bwd_data_wmma_gnhwc_gkyxc_gnhwk_f16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
GNHWK,
GKYXC,
Expand All @@ -24,7 +24,7 @@ void add_device_grouped_conv2d_bwd_data_wmma_gnhwk_gkyxc_gnhwc_f16_instances(
PassThrough,
PassThrough>>>& instances);

void add_device_grouped_conv2d_bwd_data_wmma_gnhwk_gkyxc_gnhwc_f16_1x1s1p0_instances(
void add_device_grouped_conv2d_bwd_data_wmma_gnhwc_gkyxc_gnhwk_f16_1x1s1p0_instances(
std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
GNHWK,
GKYXC,
Expand All @@ -38,7 +38,7 @@ void add_device_grouped_conv2d_bwd_data_wmma_gnhwk_gkyxc_gnhwc_f16_1x1s1p0_insta
PassThrough,
PassThrough>>>& instances);

void add_device_grouped_conv2d_bwd_data_wmma_nhwgk_gkyxc_nhwgc_f16_instances(
void add_device_grouped_conv2d_bwd_data_wmma_nhwgc_gkyxc_nhwgk_f16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
NHWGK,
GKYXC,
Expand All @@ -52,7 +52,7 @@ void add_device_grouped_conv2d_bwd_data_wmma_nhwgk_gkyxc_nhwgc_f16_instances(
PassThrough,
PassThrough>>>& instances);

void add_device_grouped_conv2d_bwd_data_wmma_nhwgk_gkyxc_nhwgc_f16_1x1s1p0_instances(
void add_device_grouped_conv2d_bwd_data_wmma_nhwgc_gkyxc_nhwgk_f16_1x1s1p0_instances(
std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
NHWGK,
GKYXC,
Expand All @@ -66,7 +66,7 @@ void add_device_grouped_conv2d_bwd_data_wmma_nhwgk_gkyxc_nhwgc_f16_1x1s1p0_insta
PassThrough,
PassThrough>>>& instances);

void add_device_grouped_conv3d_bwd_data_wmma_gndhwk_gkzyxc_gndhwc_f16_instances(
void add_device_grouped_conv3d_bwd_data_wmma_gndhwc_gkzyxc_gndhwk_f16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
GNDHWK,
GKZYXC,
Expand All @@ -80,7 +80,7 @@ void add_device_grouped_conv3d_bwd_data_wmma_gndhwk_gkzyxc_gndhwc_f16_instances(
PassThrough,
PassThrough>>>& instances);

void add_device_grouped_conv3d_bwd_data_wmma_gndhwk_gkzyxc_gndhwc_f16_1x1s1p0_instances(
void add_device_grouped_conv3d_bwd_data_wmma_gndhwc_gkzyxc_gndhwk_f16_1x1s1p0_instances(
std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
GNDHWK,
GKZYXC,
Expand All @@ -94,7 +94,7 @@ void add_device_grouped_conv3d_bwd_data_wmma_gndhwk_gkzyxc_gndhwc_f16_1x1s1p0_in
PassThrough,
PassThrough>>>& instances);

void add_device_grouped_conv3d_bwd_data_wmma_ndhwgk_gkzyxc_ndhwgc_f16_instances(
void add_device_grouped_conv3d_bwd_data_wmma_ndhwgc_gkzyxc_ndhwgk_f16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
NDHWGK,
GKZYXC,
Expand All @@ -108,7 +108,7 @@ void add_device_grouped_conv3d_bwd_data_wmma_ndhwgk_gkzyxc_ndhwgc_f16_instances(
PassThrough,
PassThrough>>>& instances);

void add_device_grouped_conv3d_bwd_data_wmma_ndhwgk_gkzyxc_ndhwgc_f16_1x1s1p0_instances(
void add_device_grouped_conv3d_bwd_data_wmma_ndhwgc_gkzyxc_ndhwgk_f16_1x1s1p0_instances(
std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
NDHWGK,
GKZYXC,
Expand All @@ -124,7 +124,7 @@ void add_device_grouped_conv3d_bwd_data_wmma_ndhwgk_gkzyxc_ndhwgc_f16_1x1s1p0_in
#endif

#ifdef CK_ENABLE_INT8
void add_device_grouped_conv2d_bwd_data_wmma_gnhwk_gkyxc_gnhwc_i8_instances(
void add_device_grouped_conv2d_bwd_data_wmma_gnhwc_gkyxc_gnhwk_i8_instances(
std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
GNHWK,
GKYXC,
Expand All @@ -138,7 +138,7 @@ void add_device_grouped_conv2d_bwd_data_wmma_gnhwk_gkyxc_gnhwc_i8_instances(
PassThrough,
PassThrough>>>& instances);

void add_device_grouped_conv2d_bwd_data_wmma_gnhwk_gkyxc_gnhwc_i8_1x1s1p0_instances(
void add_device_grouped_conv2d_bwd_data_wmma_gnhwc_gkyxc_gnhwk_i8_1x1s1p0_instances(
std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
GNHWK,
GKYXC,
Expand All @@ -152,7 +152,7 @@ void add_device_grouped_conv2d_bwd_data_wmma_gnhwk_gkyxc_gnhwc_i8_1x1s1p0_instan
PassThrough,
PassThrough>>>& instances);

void add_device_grouped_conv2d_bwd_data_wmma_nhwgk_gkyxc_nhwgc_i8_instances(
void add_device_grouped_conv2d_bwd_data_wmma_nhwgc_gkyxc_nhwgk_i8_instances(
std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
NHWGK,
GKYXC,
Expand All @@ -166,7 +166,7 @@ void add_device_grouped_conv2d_bwd_data_wmma_nhwgk_gkyxc_nhwgc_i8_instances(
PassThrough,
PassThrough>>>& instances);

void add_device_grouped_conv2d_bwd_data_wmma_nhwgk_gkyxc_nhwgc_i8_1x1s1p0_instances(
void add_device_grouped_conv2d_bwd_data_wmma_nhwgc_gkyxc_nhwgk_i8_1x1s1p0_instances(
std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
NHWGK,
GKYXC,
Expand All @@ -180,7 +180,7 @@ void add_device_grouped_conv2d_bwd_data_wmma_nhwgk_gkyxc_nhwgc_i8_1x1s1p0_instan
PassThrough,
PassThrough>>>& instances);

void add_device_grouped_conv3d_bwd_data_wmma_gndhwk_gkzyxc_gndhwc_i8_instances(
void add_device_grouped_conv3d_bwd_data_wmma_gndhwc_gkzyxc_gndhwk_i8_instances(
std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
GNDHWK,
GKZYXC,
Expand All @@ -194,7 +194,7 @@ void add_device_grouped_conv3d_bwd_data_wmma_gndhwk_gkzyxc_gndhwc_i8_instances(
PassThrough,
PassThrough>>>& instances);

void add_device_grouped_conv3d_bwd_data_wmma_gndhwk_gkzyxc_gndhwc_i8_1x1s1p0_instances(
void add_device_grouped_conv3d_bwd_data_wmma_gndhwc_gkzyxc_gndhwk_i8_1x1s1p0_instances(
std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
GNDHWK,
GKZYXC,
Expand All @@ -208,7 +208,7 @@ void add_device_grouped_conv3d_bwd_data_wmma_gndhwk_gkzyxc_gndhwc_i8_1x1s1p0_ins
PassThrough,
PassThrough>>>& instances);

void add_device_grouped_conv3d_bwd_data_wmma_ndhwgk_gkzyxc_ndhwgc_i8_instances(
void add_device_grouped_conv3d_bwd_data_wmma_ndhwgc_gkzyxc_ndhwgk_i8_instances(
std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
NDHWGK,
GKZYXC,
Expand All @@ -222,7 +222,7 @@ void add_device_grouped_conv3d_bwd_data_wmma_ndhwgk_gkzyxc_ndhwgc_i8_instances(
PassThrough,
PassThrough>>>& instances);

void add_device_grouped_conv3d_bwd_data_wmma_ndhwgk_gkzyxc_ndhwgc_i8_1x1s1p0_instances(
void add_device_grouped_conv3d_bwd_data_wmma_ndhwgc_gkzyxc_ndhwgk_i8_1x1s1p0_instances(
std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
NDHWGK,
GKZYXC,
Expand Down
Loading
Loading