Improve split operator by oneDNN reorder primitive

[bgawrych]

Description

Split operation can be done by multiple reorder primitives with properly set strides

Benchmark script:

import gc
import mxnet
import mxnet.gluon.nn as nn
import mxnet.numpy as np
import time


dims = [128, 512, 1024]
print("shape;axis;time")
for ndim in range(2):
   for dim1 in dims:
     for dim2 in dims:
        shape = (dim1, dim2) if ndim == 0 else (8, dim1, dim2)
        a = np.random.uniform(-1.0, 1.0, shape).astype(np.float32)
        for axis in range(2 + ndim):
            for section in (2, 4, 8):
                tic = time.time()
                for i in range(100):
                    out = np.array_split(a, axis=axis, indices_or_sections=section)
                    [o.wait_to_read() for o in out]
                toc = time.time()
                gc.collect()
                print(f"{shape};{axis};{section};{toc-tic}")

[mxnet-bot]

Hey @bgawrych , Thanks for submitting the PR
All tests are already queued to run once. If tests fail, you can trigger one or more tests again with the following commands:

• To trigger all jobs: @mxnet-bot run ci [all]
• To trigger specific jobs: @mxnet-bot run ci [job1, job2]

---

CI supported jobs: [unix-cpu, website, sanity, windows-gpu, edge, unix-gpu, miscellaneous, windows-cpu, centos-cpu, clang, centos-gpu]

---

Note:
Only following 3 categories can trigger CI :PR Author, MXNet Committer, Jenkins Admin.
All CI tests must pass before the PR can be merged.

[bgawrych]

@mxnet-bot run ci [centos-gpu]

[mxnet-bot]

Jenkins CI successfully triggered : [centos-gpu]

[bgawrych]

@mxnet-bot run ci [unix-gpu, windows-gpu]

[mxnet-bot]

Jenkins CI successfully triggered : [unix-gpu, windows-gpu]

[bgawrych]

@mxnet-bot run ci [windows-gpu]

[mxnet-bot]

Jenkins CI successfully triggered : [windows-gpu]

[bgawrych]

@mxnet-bot run ci [windows-gpu]

[mxnet-bot]

Jenkins CI successfully triggered : [windows-gpu]

[bgawrych]

@mxnet-bot run ci [windows-gpu]

[mxnet-bot]

Jenkins CI successfully triggered : [windows-gpu]

[bartekkuncer]

LGTM. Please change onednn to oneDNN in any future change.