diff options
| author | Patrik Gustavsson <patrik.gustavsson@arm.com> | 2020-08-13 13:41:05 +0200 |
|---|---|---|
| committer | Fredrik Knutsson <fredrik.knutsson.hunnebo@gmail.com> | 2020-08-14 10:49:15 +0000 |
| commit | 458a208c44f70a9848f1e8e2e91f28ce3641c48f (patch) | |
| tree | 37f23561f75d61746383dafc987b411646baaed8 /ethosu/vela/tensor.py | |
| parent | be733cf04bb262d4eee791d76f01cecd64ff9255 (diff) | |
| download | ethos-u-vela-458a208c44f70a9848f1e8e2e91f28ce3641c48f.tar.gz | |
MLBEDSW-2570 Avoid usage of NHCWB16 for some cases
Avoid usage of NHCWB16 when Stack/Pack/Concat is performed in axis 3,
and the "concat start" of each slice to be combined is not a multiple
of 16.
Signed-off-by: Patrik Gustavsson <patrik.gustavsson@arm.com>
Change-Id: If3f7b4a3424be3c86fc2dc48e8649ce4c4f49485
Diffstat (limited to 'ethosu/vela/tensor.py')
| -rw-r--r-- | ethosu/vela/tensor.py | 4 |
1 files changed, 4 insertions, 0 deletions
diff --git a/ethosu/vela/tensor.py b/ethosu/vela/tensor.py index 35749709..ecca0e0e 100644 --- a/ethosu/vela/tensor.py +++ b/ethosu/vela/tensor.py @@ -300,6 +300,7 @@ class Tensor: "npu_tensor", "equivalence_id", "resampling_mode", + "avoid_NHCWB16", ) AllocationQuantum = 16 @@ -346,6 +347,8 @@ class Tensor: self.block_traversal = TensorBlockTraversal.Default self.resampling_mode = resampling_mode.NONE + self.avoid_NHCWB16 = False + def element_size(self): if self.element_size_bytes == 0: return self.dtype.size_in_bits() / 8 @@ -380,6 +383,7 @@ class Tensor: res.resampling_mode = self.resampling_mode res.copy_compressed_weight_info(self) + res.avoid_NHCWB16 = self.avoid_NHCWB16 return res def clone_into_fast_storage(self, arch): |