ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-18 03:49:41 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add support for GKCYX grouped conv fwd (#2015)

Browse Source 
* Add support for GKCYX grouped conv fwd

* fixes

* fix

* changelog

* Fixes

[ROCm/composable_kernel commit: 54c81a1fcf]
This commit is contained in:
Bartłomiej Kocot
2025-03-26 21:13:38 +01:00
committed by GitHub
parent ba16351a03
commit f967fd7296
39 changed files with 1005 additions and 570 deletions
Download Patch File Download Diff File Expand all files Collapse all files

40
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/CMakeLists.txt
View File

@@ -15,6 +15,10 @@ add_instance_library(device_grouped_conv2d_fwd_instance
xdl/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f16_instance.cpp
xdl/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f32_instance.cpp
xdl/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_int8_instance.cpp
# NGCHW, GKCYX, NGKHW
xdl/device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_bf16_instance.cpp
xdl/device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_f16_instance.cpp
xdl/device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_f32_instance.cpp
# large tensor
# NHWGC, GKYXC, NHWGK
xdl/large_tensor/device_grouped_conv2d_fwd_xdl_large_tensor_nhwgc_gkyxc_nhwgk_bf16_instance.cpp
@@ -27,11 +31,10 @@ add_instance_library(device_grouped_conv2d_fwd_instance
xdl/merged_groups/device_grouped_conv2d_fwd_xdl_merged_groups_nhwgc_gkyxc_nhwgk_f16_instance.cpp
xdl/merged_groups/device_grouped_conv2d_fwd_xdl_merged_groups_nhwgc_gkyxc_nhwgk_f32_instance.cpp
xdl/merged_groups/device_grouped_conv2d_fwd_xdl_merged_groups_nhwgc_gkyxc_nhwgk_int8_instance.cpp
# NGCHW, GKYXC, NGKHW
xdl/merged_groups/device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkyxc_ngkhw_bf16_instance.cpp
xdl/merged_groups/device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkyxc_ngkhw_f16_instance.cpp
xdl/merged_groups/device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkyxc_ngkhw_f32_instance.cpp
xdl/merged_groups/device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkyxc_ngkhw_int8_instance.cpp
# NGCHW, GKCYX, NGKHW
xdl/merged_groups/device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkcyx_ngkhw_bf16_instance.cpp
xdl/merged_groups/device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkcyx_ngkhw_f16_instance.cpp
xdl/merged_groups/device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkcyx_ngkhw_f32_instance.cpp
#mem
# NHWGC, GKYXC, NHWGK
xdl/mem/device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_bf16_mem_intra_instance.cpp
@@ -43,27 +46,24 @@ add_instance_library(device_grouped_conv2d_fwd_instance
xdl/mem/device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_f16_mem_inter_instance.cpp
xdl/mem/device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_f32_mem_inter_instance.cpp
xdl/mem/device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_int8_mem_inter_instance.cpp
# NGCHW, GKYXC, NGKHW
xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_bf16_mem_intra_instance.cpp
xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f16_mem_intra_instance.cpp
xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f32_mem_intra_instance.cpp
xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_int8_mem_intra_instance.cpp
# NGCHW, GKYXC, NGKHW
xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_bf16_mem_inter_instance.cpp
xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f16_mem_inter_instance.cpp
xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f32_mem_inter_instance.cpp
xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_int8_mem_inter_instance.cpp
# NGCHW, GKCYX, NGKHW
xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_bf16_mem_intra_instance.cpp
xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_f16_mem_intra_instance.cpp
xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_f32_mem_intra_instance.cpp
# NGCHW, GKCYX, NGKHW
xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_bf16_mem_inter_instance.cpp
xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_f16_mem_inter_instance.cpp
xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_f32_mem_inter_instance.cpp
#comp
# NHWGC, GKYXC, NHWGK
xdl/comp/device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_bf16_comp_instance.cpp
xdl/comp/device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_f16_comp_instance.cpp
xdl/comp/device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_f32_comp_instance.cpp
xdl/comp/device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_int8_comp_instance.cpp
# NGCHW, GKYXC, NGKHW
xdl/comp/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_bf16_comp_instance.cpp
xdl/comp/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f16_comp_instance.cpp
xdl/comp/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f32_comp_instance.cpp
xdl/comp/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_int8_comp_instance.cpp
# NGCHW, GKCYX, NGKHW
xdl/comp/device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_bf16_comp_instance.cpp
xdl/comp/device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_f16_comp_instance.cpp
xdl/comp/device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_f32_comp_instance.cpp
#dl
# GNHWC, GKYXC, GNHWK
dl/device_grouped_conv2d_fwd_dl_gnhwc_gkyxc_gnhwk_f16_instance.cpp

12
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_bf16_comp_instance.cpp → library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_bf16_comp_instance.cpp
View File

@@ -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.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_comp_instance.hpp"
@@ -10,10 +10,10 @@ namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_bf16_comp_instances(
void add_device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_bf16_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
BF16,
@@ -28,7 +28,7 @@ void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_bf16_comp_instances(
instances,
device_grouped_conv_fwd_xdl_bf16_comp_instances<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwdDefault>{});
@@ -39,7 +39,7 @@ void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_bf16_comp_instances(
instances,
device_grouped_conv_fwd_xdl_bf16_comp_instances_part2<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwdDefault>{});
@@ -51,7 +51,7 @@ void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_bf16_comp_instances(
instances,
device_grouped_conv_fwd_xdl_bf16_comp_instances_2x<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwdDefault>{});

12
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f16_comp_instance.cpp → library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_f16_comp_instance.cpp
View File

@@ -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.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_comp_instance.hpp"
@@ -10,10 +10,10 @@ namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f16_comp_instances(
void add_device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_f16_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
F16,
@@ -28,7 +28,7 @@ void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f16_comp_instances(
instances,
device_grouped_conv_fwd_xdl_f16_comp_instances<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwdDefault>{});
@@ -39,7 +39,7 @@ void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f16_comp_instances(
instances,
device_grouped_conv_fwd_xdl_f16_comp_instances_part2<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwdDefault>{});
@@ -51,7 +51,7 @@ void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f16_comp_instances(
instances,
device_grouped_conv_fwd_xdl_f16_comp_instances_2x<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwdDefault>{});

8
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f32_comp_instance.cpp → library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_f32_comp_instance.cpp
View File

@@ -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.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_comp_instance.hpp"
@@ -9,10 +9,10 @@ namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f32_comp_instances(
void add_device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_f32_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
F32,
@@ -27,7 +27,7 @@ void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f32_comp_instances(
instances,
device_grouped_conv_fwd_xdl_f32_comp_instances<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwdDefault>{});

64
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_int8_comp_instance.cpp
View File

@@ -1,64 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_comp_instance.hpp"
#include "ck/host_utility/device_prop.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_int8_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKYXC,
Empty_Tuple,
NGKHW,
int8_t,
int8_t,
Empty_Tuple,
int8_t,
PassThrough,
PassThrough,
PassThrough>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_xdl_int8_comp_instances<2,
NGCHW,
GKYXC,
Empty_Tuple,
NGKHW,
ConvFwdDefault>{});
if(ck::get_device_name() != "gfx950")
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_xdl_int8_comp_instances_part2<2,
NGCHW,
GKYXC,
Empty_Tuple,
NGKHW,
ConvFwdDefault>{});
}
if(ck::get_device_name() == "gfx950")
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_xdl_int8_comp_instances_2x<2,
NGCHW,
GKYXC,
Empty_Tuple,
NGKHW,
ConvFwdDefault>{});
}
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

38
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_bf16_instance.cpp Normal file
View File

@@ -0,0 +1,38 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_bf16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKCYX,
Empty_Tuple,
NGKHW,
BF16,
BF16,
Empty_Tuple,
BF16,
PassThrough,
PassThrough,
PassThrough>>>& instances)
{
add_device_operation_instances(instances,
device_grouped_conv_fwd_xdl_bf16_instances<2,
NGCHW,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwdDefault>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

38
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_f16_instance.cpp Normal file
View File

@@ -0,0 +1,38 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_f16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKCYX,
Empty_Tuple,
NGKHW,
F16,
F16,
Empty_Tuple,
F16,
PassThrough,
PassThrough,
PassThrough>>>& instances)
{
add_device_operation_instances(instances,
device_grouped_conv_fwd_xdl_f16_instances<2,
NGCHW,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwdDefault>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

38
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_f32_instance.cpp Normal file
View File

@@ -0,0 +1,38 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_f32_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKCYX,
Empty_Tuple,
NGKHW,
F32,
F32,
Empty_Tuple,
F32,
PassThrough,
PassThrough,
PassThrough>>>& instances)
{
add_device_operation_instances(instances,
device_grouped_conv_fwd_xdl_f32_instances<2,
NGCHW,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwdDefault>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

8
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_bf16_mem_inter_instance.cpp → library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_bf16_mem_inter_instance.cpp
View File

@@ -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.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_mem_instance.hpp"
@@ -9,10 +9,10 @@ namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_bf16_mem_inter_instances(
void add_device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_bf16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
BF16,
@@ -26,7 +26,7 @@ void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_bf16_mem_inter_instance
add_device_operation_instances(instances,
device_grouped_conv_fwd_xdl_bf16_mem_instances<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwdDefault,

8
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_bf16_mem_intra_instance.cpp → library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_bf16_mem_intra_instance.cpp
View File

@@ -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.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_mem_instance.hpp"
@@ -9,10 +9,10 @@ namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_bf16_mem_intra_instances(
void add_device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_bf16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
BF16,
@@ -26,7 +26,7 @@ void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_bf16_mem_intra_instance
add_device_operation_instances(instances,
device_grouped_conv_fwd_xdl_bf16_mem_instances<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwdDefault,

8
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f16_mem_inter_instance.cpp → library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_f16_mem_inter_instance.cpp
View File

@@ -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.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_mem_instance.hpp"
@@ -9,10 +9,10 @@ namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f16_mem_inter_instances(
void add_device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_f16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
F16,
@@ -26,7 +26,7 @@ void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f16_mem_inter_instances
add_device_operation_instances(instances,
device_grouped_conv_fwd_xdl_f16_mem_instances<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwdDefault,

8
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f16_mem_intra_instance.cpp → library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_f16_mem_intra_instance.cpp
View File

@@ -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.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_mem_instance.hpp"
@@ -9,10 +9,10 @@ namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f16_mem_intra_instances(
void add_device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_f16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
F16,
@@ -26,7 +26,7 @@ void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f16_mem_intra_instances
add_device_operation_instances(instances,
device_grouped_conv_fwd_xdl_f16_mem_instances<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwdDefault,

8
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f32_mem_inter_instance.cpp → library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_f32_mem_inter_instance.cpp
View File

@@ -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.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_mem_instance.hpp"
@@ -9,10 +9,10 @@ namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f32_mem_inter_instances(
void add_device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_f32_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
F32,
@@ -26,7 +26,7 @@ void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f32_mem_inter_instances
add_device_operation_instances(instances,
device_grouped_conv_fwd_xdl_f32_mem_instances<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwdDefault,

8
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f32_mem_intra_instance.cpp → library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_f32_mem_intra_instance.cpp
View File

@@ -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.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_mem_instance.hpp"
@@ -9,10 +9,10 @@ namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f32_mem_intra_instances(
void add_device_grouped_conv2d_fwd_xdl_ngchw_gkcyx_ngkhw_f32_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
F32,
@@ -26,7 +26,7 @@ void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f32_mem_intra_instances
add_device_operation_instances(instances,
device_grouped_conv_fwd_xdl_f32_mem_instances<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwdDefault,

39
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_int8_mem_inter_instance.cpp
View File

@@ -1,39 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_mem_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_int8_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKYXC,
Empty_Tuple,
NGKHW,
int8_t,
int8_t,
Empty_Tuple,
int8_t,
PassThrough,
PassThrough,
PassThrough>>>& instances)
{
add_device_operation_instances(instances,
device_grouped_conv_fwd_xdl_int8_mem_instances<2,
NGCHW,
GKYXC,
Empty_Tuple,
NGKHW,
ConvFwdDefault,
Interwave>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

39
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/mem/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_int8_mem_intra_instance.cpp
View File

@@ -1,39 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_mem_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_int8_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKYXC,
Empty_Tuple,
NGKHW,
int8_t,
int8_t,
Empty_Tuple,
int8_t,
PassThrough,
PassThrough,
PassThrough>>>& instances)
{
add_device_operation_instances(instances,
device_grouped_conv_fwd_xdl_int8_mem_instances<2,
NGCHW,
GKYXC,
Empty_Tuple,
NGKHW,
ConvFwdDefault,
Intrawave>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck

10
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/merged_groups/device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkyxc_ngkhw_bf16_instance.cpp → library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/merged_groups/device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkcyx_ngkhw_bf16_instance.cpp
View File

@@ -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.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_merged_groups_instance.hpp"
@@ -9,10 +9,10 @@ namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkyxc_ngkhw_bf16_instances(
void add_device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkcyx_ngkhw_bf16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
BF16,
@@ -27,7 +27,7 @@ void add_device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkyxc_ngkhw_bf16_inst
instances,
device_grouped_conv_fwd_xdl_merged_groups_bf16_instances<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwdDefault>{});
@@ -36,7 +36,7 @@ void add_device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkyxc_ngkhw_bf16_inst
instances,
device_grouped_conv_fwd_xdl_merged_groups_bf16_instances<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwd3x3>{});

10
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/merged_groups/device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkyxc_ngkhw_f16_instance.cpp → library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/merged_groups/device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkcyx_ngkhw_f16_instance.cpp
View File

@@ -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.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_merged_groups_instance.hpp"
@@ -9,10 +9,10 @@ namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkyxc_ngkhw_f16_instances(
void add_device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkcyx_ngkhw_f16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
F16,
@@ -27,7 +27,7 @@ void add_device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkyxc_ngkhw_f16_insta
instances,
device_grouped_conv_fwd_xdl_merged_groups_f16_instances<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwdDefault>{});
@@ -36,7 +36,7 @@ void add_device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkyxc_ngkhw_f16_insta
instances,
device_grouped_conv_fwd_xdl_merged_groups_f16_instances<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwd3x3>{});

10
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/merged_groups/device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkyxc_ngkhw_f32_instance.cpp → library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/merged_groups/device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkcyx_ngkhw_f32_instance.cpp
View File

@@ -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.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_merged_groups_instance.hpp"
@@ -9,10 +9,10 @@ namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkyxc_ngkhw_f32_instances(
void add_device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkcyx_ngkhw_f32_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
F32,
@@ -27,7 +27,7 @@ void add_device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkyxc_ngkhw_f32_insta
instances,
device_grouped_conv_fwd_xdl_merged_groups_f32_instances<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwdDefault>{});
@@ -36,7 +36,7 @@ void add_device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkyxc_ngkhw_f32_insta
instances,
device_grouped_conv_fwd_xdl_merged_groups_f32_instances<2,
NGCHW,
GKYXC,
GKCYX,
Empty_Tuple,
NGKHW,
ConvFwd3x3>{});

48
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/merged_groups/device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkyxc_ngkhw_int8_instance.cpp
View File

@@ -1,48 +0,0 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_merged_groups_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_xdl_merged_groups_ngchw_gkyxc_ngkhw_int8_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,
GKYXC,
Empty_Tuple,
NGKHW,
int8_t,
int8_t,
Empty_Tuple,
int8_t,
PassThrough,
PassThrough,
PassThrough>>>& instances)
{
add_device_operation_instances(
instances,
device_grouped_conv_fwd_xdl_merged_groups_int8_instances<2,
NGCHW,
GKYXC,
Empty_Tuple,
NGKHW,
ConvFwdDefault>{});
add_device_operation_instances(
instances,
device_grouped_conv_fwd_xdl_merged_groups_int8_instances<2,
NGCHW,
GKYXC,
Empty_Tuple,
NGKHW,
ConvFwd3x3>{});
}
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck