[PERFORMANCE] Improve shuffle implementation

[anko-intel]

Description

Improve performance of random.shuffle implementation.

Results on c6i.8xlarge instance ( Intel(R) Xeon(R) Platinum 8375C CPU @ 2.90GHz ) of the follwing script

import mxnet as mx
from benchmark.opperf.utils.benchmark_utils import run_performance_test

shapes = [
    (  10, 1),
    ( 32, 9999),
    (4096,   1),
    (4096,  500),
    (4096, 4096),
    (4096, 1024, 32)]

print("\nnot-in-place (mx.nd.random.shuffle):")
for shape in shapes:
    s_res = run_performance_test(mx.nd.random.shuffle, run_backward=False, inputs=[{"data": (shape)}])
    print(s_res)

print("\nin-place (mx.np.random.shuffle):")
for shape in shapes:
    s_res = run_performance_test(mx.np.random.shuffle, run_backward=False, inputs=[{"x": (shape)}])
    print(s_res)

shows following improvments:

• not-in-place (mx.nd.random.shuffle):

Shape	Before [ms]	After [ms]	Improvement
(10, 1)}	0.0038	0.0024	37%
(32, 9999)	0.168	0.0795	53%
(4096, 1)}	0.0463	0.0328	29%
(4096, 500)	1.1972	0.6511	46%
(4096, 4096)	15.5678	8.8016	43%
(4096, 1024, 32)	139.1404	74.0476	47%

• in-place (mx.np.random.shuffle):

Shape	Before [ms]	After [ms]	Improvement
(10, 1)}	0.0023	0.0022	4%
(32, 9999)	0.0852	0.0406	52%
(4096, 1)}	0.0436	0.0404	7%
(4096, 500)	0.6643	0.4317	35%
(4096, 4096)	6.778	3.7478	45%
(4096, 1024, 32)	89.1249	41.931	53%

Checklist

Essentials

• PR's title starts with a category (e.g. [BUGFIX], [MODEL], [TUTORIAL], [FEATURE], [DOC], etc)
• Changes are complete (i.e. I finished coding on this PR)
• All changes have test coverage
• Code is well-documented

[mxnet-bot]

Hey @anko-intel , 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: [edge, centos-gpu, sanity, unix-cpu, windows-gpu, unix-gpu, clang, website, miscellaneous, centos-cpu, windows-cpu]

---

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.

[anko-intel]

following error occurs after first comit:
/work/mxnet/src/operator/random/shuffle_op.cc:82:18: error: 'void* memcpy(void*, const void*, size_t)' writing to an object of type 'class mshadow::half::half_t' with no trivial copy-assignment; use copy-assignment or copy-initialization instead [-Werror=class-memaccess]

[anko-intel]

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

[mxnet-bot]

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

[anko-intel]

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

[mxnet-bot]

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

[anko-intel]

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

[mxnet-bot]

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

[anko-intel]

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

[mxnet-bot]

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

[mozga-intel]

LGTM!