Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in / Register
Toggle navigation
Menu
Open sidebar
gaoqiong
composable_kernel_ROCM
Commits
e0a26328
Commit
e0a26328
authored
Jan 14, 2025
by
Mateusz Ozga
Browse files
Instances updated
parent
1a0af70d
Changes
1
Hide whitespace changes
Inline
Side-by-side
Showing
1 changed file
with
95 additions
and
95 deletions
+95
-95
library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp
...wd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp
+95
-95
No files found.
library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp
View file @
e0a26328
...
@@ -54,24 +54,24 @@ using device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances = std::tuple<
...
@@ -54,24 +54,24 @@ using device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances = std::tuple<
//#########################################| Spatial| | | | | | | | Operation| Operation| Operation| Specialization| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| MBlock_MPerBlock| NWaveNPerXdl| Scheduler| Version|
//#########################################| Spatial| | | | | | | | Operation| Operation| Operation| Specialization| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| MBlock_MPerBlock| NWaveNPerXdl| Scheduler| Version|
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| | | |
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| | | |
// generic instance
// generic instance
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
64
,
16
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
2
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
1
,
Scheduler
,
PipelineVersion
>
,
// instances for small conv.K and conv.C
// instances for small conv.K and conv.C
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
32
,
16
,
4
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
1
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
1
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
32
,
64
,
16
,
4
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
32
,
64
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
256
,
128
,
16
,
4
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
64
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
256
,
128
,
32
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
64
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
256
,
16
,
4
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
64
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
256
,
32
,
8
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
64
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
128
,
16
,
4
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
128
,
32
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
128
,
16
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
128
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
64
,
16
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
64
,
128
,
16
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
64
,
128
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
64
,
16
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
64
,
16
,
4
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
64
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
64
,
128
,
16
,
4
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
64
,
128
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
32
,
16
,
4
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
32
,
128
,
16
,
4
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
32
,
128
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
32
,
16
,
4
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
32
,
64
,
16
,
4
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
32
,
64
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
// clang-format on
// clang-format on
>
;
>
;
...
@@ -89,26 +89,26 @@ using device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances = std::tuple<
...
@@ -89,26 +89,26 @@ using device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances = std::tuple<
//#########################################| Spatial| | | | | | | | Operation| Operation| Operation| Specialization| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| MBlock_MPerBlock| NWaveNPerXdl| Scheduler| Version|
//#########################################| Spatial| | | | | | | | Operation| Operation| Operation| Specialization| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| MBlock_MPerBlock| NWaveNPerXdl| Scheduler| Version|
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| | | |
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| | | |
// generic instance
// generic instance
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
2
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
2
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
2
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
2
,
Scheduler
,
PipelineVersion
>
,
// instance for small conv.K
// instance for small conv.K
// for fp16 conv.K and conv.C must be divisible by 2
// for fp16 conv.K and conv.C must be divisible by 2
// since half_t atomic_add require scalar_per_x_vector % 2 == 0
// since half_t atomic_add require scalar_per_x_vector % 2 == 0
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
2
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
2
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
2
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
32
,
64
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
2
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
32
,
64
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
256
,
128
,
32
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
256
,
128
,
32
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
256
,
32
,
8
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
256
,
32
,
8
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
128
,
32
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
128
,
32
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
128
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
128
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
64
,
128
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
64
,
128
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
64
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
64
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
64
,
128
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
64
,
128
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
32
,
128
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
32
,
128
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
32
,
64
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
32
,
64
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
// clang-format on
// clang-format on
>
;
>
;
...
@@ -127,23 +127,23 @@ using device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances = std
...
@@ -127,23 +127,23 @@ using device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances = std
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| | | |
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| | | |
// generic instance
// generic instance
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
1
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
1
,
Scheduler
,
PipelineVersion
>
,
// // instance for small conv.K
// // instance for small conv.K
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
1
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
1
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
32
,
64
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
32
,
64
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
256
,
128
,
32
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
256
,
128
,
32
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
256
,
32
,
8
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
256
,
32
,
8
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
128
,
32
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
128
,
32
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
128
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
128
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
64
,
128
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
64
,
128
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
64
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
64
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
64
,
128
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
64
,
128
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
32
,
128
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
32
,
128
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
32
,
64
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
32
,
64
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
,
Scheduler
,
PipelineVersion
>
// clang-format on
// clang-format on
>
;
>
;
...
@@ -161,26 +161,26 @@ using device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_instances = std::tuple<
...
@@ -161,26 +161,26 @@ using device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_instances = std::tuple<
//#########################################| Spatial| | | | | | | | Operation| Operation| Operation| Specialization| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| MBlock_MPerBlock| NWaveNPerXdl| Scheduler| Version|
//#########################################| Spatial| | | | | | | | Operation| Operation| Operation| Specialization| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| MBlock_MPerBlock| NWaveNPerXdl| Scheduler| Version|
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| | | |
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| | | |
// generic instance
// generic instance
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
2
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
2
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
2
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
2
,
Scheduler
,
PipelineVersion
>
,
// // instance for small conv.K
// // instance for small conv.K
// // for bf16 conv.K and conv.C must be divisible by 2
// // for bf16 conv.K and conv.C must be divisible by 2
// // since half_t atomic_add require scalar_per_x_vector % 2 == 0
// // since half_t atomic_add require scalar_per_x_vector % 2 == 0
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
2
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
2
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
2
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
32
,
64
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
2
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
32
,
64
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
256
,
128
,
32
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
256
,
128
,
32
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
256
,
32
,
8
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
256
,
32
,
8
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
128
,
32
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
128
,
32
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
128
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
128
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
64
,
128
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
64
,
128
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
64
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
64
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
64
,
128
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
64
,
128
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
32
,
128
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
32
,
128
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
32
,
64
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
BF16
,
BF16
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
32
,
64
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
>
//clang-format on
//clang-format on
>
;
>
;
...
@@ -199,26 +199,26 @@ using device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_comp_bf8_f8_instances = s
...
@@ -199,26 +199,26 @@ using device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_comp_bf8_f8_instances = s
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| | | | | |
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| | | | | |
#if defined CK_ENABLE_FP8 && defined CK_ENABLE_BF8
#if defined CK_ENABLE_FP8 && defined CK_ENABLE_BF8
// generic instance
// generic instance
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
2
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
2
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
2
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
2
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
// // instance for small conv.K
// // instance for small conv.K
// // for fp16 conv.K and conv.C must be divisible by 2
// // for fp16 conv.K and conv.C must be divisible by 2
// // since half_t atomic_add require scalar_per_x_vector % 2 == 0
// // since half_t atomic_add require scalar_per_x_vector % 2 == 0
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
2
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
2
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
2
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
32
,
64
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
2
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
32
,
64
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
256
,
128
,
32
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
256
,
128
,
32
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
256
,
32
,
8
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
256
,
32
,
8
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
128
,
32
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
128
,
32
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
128
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
128
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
64
,
128
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
64
,
128
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
64
,
32
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
64
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
128
,
64
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
64
,
128
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
256
,
64
,
128
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
128
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
32
,
128
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
128
,
32
,
128
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
1
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
64
,
32
,
32
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
,
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
32
,
64
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
DeviceGroupedConvBwdWeight_Xdl_CShuffle
<
NDimSpatial
,
ALayout
,
BLayout
,
ELayout
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvSpec
,
64
,
32
,
64
,
32
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
1
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
,
Scheduler
,
PipelineVersion
,
BF8
,
F8
>
#endif
#endif
// clang-format on
// clang-format on
>
;
>
;
...
...
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
.
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment