Arm backend: Add CanonicalizeViewCopyPermutePass (pytorch#20276)

This pass replaces PostponePermuteOpBelowSqueezeOrUnsqueezeLikeView
using a more general condition for which permute-view combinations may
be
fused, see docstrings.

+ remove stale comments in mha test

Signed-off-by: Adrian Lundell <adrian.lundell@arm.com>

Author: AdrianLundell

backends/arm/_passes/__init__.py

@@ -9,6 +9,7 @@
 from .accumulate_index_put_pass import AccumulateIndexPutPass  # noqa
 from .broadcast_args_pass import BroadcastArgsPass  # noqa
 from .canonicalize_gather_pass import CanonicalizeGatherPass  # noqa
+from .canonicalize_view_copy_permute_pass import CanonicalizeViewCopyPermutePass  # noqa
 from .cast_int64_pass import CastInt64BuffersToInt32Pass  # noqa
 from .cast_to_int32_pass import CastToInt32Pass  # noqa
 from .constant_folding_pass import ConstantFoldingPass  # noqa

backends/arm/_passes/arm_pass_manager.py

@@ -15,6 +15,7 @@
     AccumulateIndexPutPass,
     BroadcastArgsPass,
     CanonicalizeGatherPass,
+    CanonicalizeViewCopyPermutePass,
     CastInt64BuffersToInt32Pass,
     CastToInt32Pass,
     ComputeConstantOpsAOTPass,
@@ -29,7 +30,6 @@
     ConvertInt64OutputOpsToInt32Pass,
     ConvertMinMaxPass,
     ConvertMmToBmmPass,
-    ConvertPermuteSingletonToViewPass,
     ConvertSplitToSlicePass,
     ConvertSqueezesToViewPass,
     ConvertToClampPass,
@@ -165,9 +165,6 @@
 from executorch.backends.transforms.fuse_cascaded_transpose_or_permute_ops import (
     FuseCascadedTransposeOrPermuteOps,
 )
-from executorch.backends.transforms.postpone_permute_below_squeeze_view import (
-    PostponePermuteOpBelowSqueezeOrUnsqueezeLikeView,
-)
 
 from executorch.exir import ExportedProgram
 from executorch.exir._program_utils import _get_updated_graph_signature
@@ -616,9 +613,8 @@ def _tosa_pipeline(
                 RewritePadPass(),
                 FuseViewCopyTransformPass(),
                 RemovePermutesAroundElementwiseTosaOps(),
-                PostponePermuteOpBelowSqueezeOrUnsqueezeLikeView(),
+                CanonicalizeViewCopyPermutePass(),
                 FuseCascadedTransposeOrPermuteOps(),
-                ConvertPermuteSingletonToViewPass(),
                 RewriteHighRankSingletonPermutePass(),
                 DecomposePermuteForU55Pass(),
                 RewriteSlicePass(),