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

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.

@mxnet-bot run ci [all]

Jenkins CI successfully triggered : [centos-cpu, miscellaneous, centos-gpu, windows-gpu, unix-gpu, sanity, clang, edge, website, windows-cpu, unix-cpu]

Adopt PackedFunc based FFI on imperative invoke to reduce overhead.

Have you measured the improvement for this PR? Previously I thought you intend to convert the remaining operators to the new FFI, rendering the generic imperative invoke unneeded?

Adopt PackedFunc based FFI on imperative invoke to reduce overhead.

Have you measured the improvement for this PR? Previously I thought you intend to convert the remaining operators to the new FFI, rendering the generic imperative invoke unneeded?

Yes, some operators that takes no stringified params as invocation input will benefit a bit from it.
copyto

setup = """
from mxnet import np, npx
npx.set_np()
a = np.zeros((2,2))
b = np.ones((2,2))
"""
stmt = """
c = a.copyto(b)
"""
timer = timeit.Timer(setup=setup,
                     stmt=stmt)
num_repeat = 1000
print(min(timer.repeat(repeat=10, number=num_repeat)) / num_repeat)

legacy
5.821948400000032e-05
New FFI
5.494140500000011e-05

npx.gather_nd

setup = """
from mxnet import np, npx
npx.set_np()
a = np.zeros((2,2))
b = np.ones((2,2))
"""
stmt = """
c = npx.gather_nd(a, b)
"""
timer = timeit.Timer(setup=setup,
                     stmt=stmt)
num_repeat = 1000
print(min(timer.repeat(repeat=10, number=num_repeat)) / num_repeat)

legacy
8.739984900000009e-05
New FFI
7.992273099999991e-05