|
@@ -455,17 +455,17 @@ class trace: |
|
|
|
|
|
|
|
|
:param enable_hwcd4: whether to use NHWCD4 data layout. This is faster on some |
|
|
:param enable_hwcd4: whether to use NHWCD4 data layout. This is faster on some |
|
|
OpenCL backend. |
|
|
OpenCL backend. |
|
|
:param enable_nchw88: whether to use NCHW4 data layout. it currently |
|
|
|
|
|
|
|
|
:param enable_nchw88: whether to use NCHW88 data layout. it currently |
|
|
used in X86 AVX backend. |
|
|
used in X86 AVX backend. |
|
|
:param enable_nchw44: whether to use NCHW4 data layout. it currently |
|
|
|
|
|
|
|
|
:param enable_nchw44: whether to use NCHW44 data layout. it currently |
|
|
used in arm backend. |
|
|
used in arm backend. |
|
|
:param enable_nchw44_dot: whether to use NCHW4 data layout. it currently |
|
|
|
|
|
|
|
|
:param enable_nchw44_dot: whether to use NCHW44_dot data layout. it currently |
|
|
used in armv8.2+dotprod backend. |
|
|
used in armv8.2+dotprod backend. |
|
|
:param enable_nchw4: whether to use NCHW4 data layout. it currently |
|
|
:param enable_nchw4: whether to use NCHW4 data layout. it currently |
|
|
used in nvidia backend(based on cudnn). |
|
|
used in nvidia backend(based on cudnn). |
|
|
:param enable_nchw32 whether to use NCHW32 data layout. it currently |
|
|
|
|
|
|
|
|
:param enable_nchw32: whether to use NCHW32 data layout. it currently |
|
|
used in nvidia backend with tensorcore(based on cudnn). |
|
|
used in nvidia backend with tensorcore(based on cudnn). |
|
|
:param enable_chwn4 whether to use CHWN4 data layout. it currently |
|
|
|
|
|
|
|
|
:param enable_chwn4: whether to use CHWN4 data layout. it currently |
|
|
used in nvidia backend with tensorcore. |
|
|
used in nvidia backend with tensorcore. |
|
|
|
|
|
|
|
|
:param enable_fuse_conv_bias_nonlinearity: whether to fuse conv+bias+nonlinearty |
|
|
:param enable_fuse_conv_bias_nonlinearity: whether to fuse conv+bias+nonlinearty |
|
|