How to workaround lack of permute + view?

[vedantroy]

I'm trying to implement the following (Pytorch equivalent):

# shape = [max_num_blocks_per_seq, head_size, block_size]
tensor = ...
# shape = [max_num_blocks_per_seq, block_size, head_size]
tensor = tensor.permute(0, 2, 1)
# shape = [max_num_blocks_per_seq * block_size, head_size]
tensor = tensor.view(-1, head_size)

However, Triton does not support the permute operator or view without re-ordering.
Is it possible to implement something like this in Triton without those 2 operators?