ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-19 20:40:07 +00:00
Code Issues Packages Projects Releases Wiki Activity

Temporary disable part of dynamic op conv instances (#1630)

Browse Source 
* Temporary disable part of dynamic op conv instances

* fix

[ROCm/composable_kernel commit: 4f1fdbb6e3]
This commit is contained in:
Bartłomiej Kocot
2024-11-04 22:34:17 +01:00
committed by GitHub
parent 232d1462a1
commit 357cab6560
9 changed files with 36 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_dynamic_op/xdl/device_grouped_conv2d_fwd_xdl_dynamic_op_nhwgc_gkyxc_nhwgk_bf16_instance.cpp
View File

@@ -31,6 +31,8 @@ void add_device_grouped_conv2d_fwd_xdl_dynamic_op_nhwgc_gkyxc_nhwgk_bf16_instanc
Tuple<>,
NHWGK,
ConvFwdDefault>{});
#if 0 // Enable with dynamic op optimizations (at now generating a lot of virtual functions cause
// long compilation time)
add_device_operation_instances(
instances,
device_grouped_conv_fwd_xdl_dynamic_op_bf16_instances<2,
@@ -47,6 +49,7 @@ void add_device_grouped_conv2d_fwd_xdl_dynamic_op_nhwgc_gkyxc_nhwgk_bf16_instanc
Tuple<>,
NHWGK,
ConvFwd1x1S1P0>{});
#endif
}
} // namespace instance

3
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_dynamic_op/xdl/device_grouped_conv2d_fwd_xdl_dynamic_op_nhwgc_gkyxc_nhwgk_f16_instance.cpp
View File

@@ -31,6 +31,8 @@ void add_device_grouped_conv2d_fwd_xdl_dynamic_op_nhwgc_gkyxc_nhwgk_f16_instance
Tuple<>,
NHWGK,
ConvFwdDefault>{});
#if 0 // Enable with dynamic op optimizations (at now generating a lot of virtual functions cause
// long compilation time)
add_device_operation_instances(
instances,
device_grouped_conv_fwd_xdl_dynamic_op_f16_instances<2,
@@ -47,6 +49,7 @@ void add_device_grouped_conv2d_fwd_xdl_dynamic_op_nhwgc_gkyxc_nhwgk_f16_instance
Tuple<>,
NHWGK,
ConvFwd1x1S1P0>{});
#endif
}
} // namespace instance

3
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_dynamic_op/xdl/device_grouped_conv2d_fwd_xdl_dynamic_op_nhwgc_gkyxc_nhwgk_f32_instance.cpp
View File

@@ -31,6 +31,8 @@ void add_device_grouped_conv2d_fwd_xdl_dynamic_op_nhwgc_gkyxc_nhwgk_f32_instance
Tuple<>,
NHWGK,
ConvFwdDefault>{});
#if 0 // Enable with dynamic op optimizations (at now generating a lot of virtual functions cause
// long compilation time)
add_device_operation_instances(
instances,
device_grouped_conv_fwd_xdl_dynamic_op_f32_instances<2,
@@ -47,6 +49,7 @@ void add_device_grouped_conv2d_fwd_xdl_dynamic_op_nhwgc_gkyxc_nhwgk_f32_instance
Tuple<>,
NHWGK,
ConvFwd1x1S1P0>{});
#endif
}
} // namespace instance

3
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_dynamic_op/xdl/device_grouped_conv2d_fwd_xdl_dynamic_op_nhwgc_gkyxc_nhwgk_int8_instance.cpp
View File

@@ -30,6 +30,8 @@ void add_device_grouped_conv2d_fwd_xdl_dynamic_op_nhwgc_gkyxc_nhwgk_int8_instanc
Tuple<>,
NHWGK,
ConvFwdDefault>{});
#if 0 // Enable with dynamic op optimizations (at now generating a lot of virtual functions cause
// long compilation time)
add_device_operation_instances(
instances,
device_grouped_conv_fwd_xdl_dynamic_op_int8_instances<2,
@@ -46,6 +48,7 @@ void add_device_grouped_conv2d_fwd_xdl_dynamic_op_nhwgc_gkyxc_nhwgk_int8_instanc
Tuple<>,
NHWGK,
ConvFwd1x1S1P0>{});
#endif
}
} // namespace instance

3
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd_dynamic_op/xdl/device_grouped_conv3d_fwd_xdl_dynamic_op_ndhwgc_gkzyxc_ndhwgk_bf16_instance.cpp
View File

@@ -31,6 +31,8 @@ void add_device_grouped_conv3d_fwd_xdl_dynamic_op_ndhwgc_gkzyxc_ndhwgk_bf16_inst
Tuple<>,
NDHWGK,
ConvFwdDefault>{});
#if 0 // Enable with dynamic op optimizations (at now generating a lot of virtual functions cause
// long compilation time)
add_device_operation_instances(
instances,
device_grouped_conv_fwd_xdl_dynamic_op_bf16_instances<3,
@@ -47,6 +49,7 @@ void add_device_grouped_conv3d_fwd_xdl_dynamic_op_ndhwgc_gkzyxc_ndhwgk_bf16_inst
Tuple<>,
NDHWGK,
ConvFwd1x1S1P0>{});
#endif
}
} // namespace instance

3
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd_dynamic_op/xdl/device_grouped_conv3d_fwd_xdl_dynamic_op_ndhwgc_gkzyxc_ndhwgk_f16_instance.cpp
View File

@@ -31,6 +31,8 @@ void add_device_grouped_conv3d_fwd_xdl_dynamic_op_ndhwgc_gkzyxc_ndhwgk_f16_insta
Tuple<>,
NDHWGK,
ConvFwdDefault>{});
#if 0 // Enable with dynamic op optimizations (at now generating a lot of virtual functions cause
// long compilation time)
add_device_operation_instances(
instances,
device_grouped_conv_fwd_xdl_dynamic_op_f16_instances<3,
@@ -47,6 +49,7 @@ void add_device_grouped_conv3d_fwd_xdl_dynamic_op_ndhwgc_gkzyxc_ndhwgk_f16_insta
Tuple<>,
NDHWGK,
ConvFwd1x1S1P0>{});
#endif
}
} // namespace instance

3
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd_dynamic_op/xdl/device_grouped_conv3d_fwd_xdl_dynamic_op_ndhwgc_gkzyxc_ndhwgk_f32_instance.cpp
View File

@@ -31,6 +31,8 @@ void add_device_grouped_conv3d_fwd_xdl_dynamic_op_ndhwgc_gkzyxc_ndhwgk_f32_insta
Tuple<>,
NDHWGK,
ConvFwdDefault>{});
#if 0 // Enable with dynamic op optimizations (at now generating a lot of virtual functions cause
// long compilation time)
add_device_operation_instances(
instances,
device_grouped_conv_fwd_xdl_dynamic_op_f32_instances<3,
@@ -47,6 +49,7 @@ void add_device_grouped_conv3d_fwd_xdl_dynamic_op_ndhwgc_gkzyxc_ndhwgk_f32_insta
Tuple<>,
NDHWGK,
ConvFwd1x1S1P0>{});
#endif
}
} // namespace instance

3
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd_dynamic_op/xdl/device_grouped_conv3d_fwd_xdl_dynamic_op_ndhwgc_gkzyxc_ndhwgk_int8_instance.cpp
View File

@@ -30,6 +30,8 @@ void add_device_grouped_conv3d_fwd_xdl_dynamic_op_ndhwgc_gkzyxc_ndhwgk_int8_inst
Tuple<>,
NDHWGK,
ConvFwdDefault>{});
#if 0 // Enable with dynamic op optimizations (at now generating a lot of virtual functions cause
// long compilation time)
add_device_operation_instances(
instances,
device_grouped_conv_fwd_xdl_dynamic_op_int8_instances<3,
@@ -46,6 +48,7 @@ void add_device_grouped_conv3d_fwd_xdl_dynamic_op_ndhwgc_gkzyxc_ndhwgk_int8_inst
Tuple<>,
NDHWGK,
ConvFwd1x1S1P0>{});
#endif
}
} // namespace instance