ARROW-11194: [Rust] Enable packed_simd for aarch64

[nevi-me]

packed_simd has support for aarch64 for the functions that we are using.

We can't test this feature yet, as we don't have aarch64 targets on our Rust CI yet. I however tested this on an ARM Mac device.
I perceive the build risk to be low, as I'm only enabling aarch64 in addition to the already used x86_64 target in packed_simd.

[github-actions]

https://issues.apache.org/jira/browse/ARROW-11194

[nevi-me]

@alamb @jorgecarleitao I see decent performance increases in benchmarks, but because my device gets hot running the benchmarks, I'm opting to view them as not too reliable.

[codecov-io]

Codecov Report

Merging #9148 (181a78d) into master (08cccd6) will decrease coverage by 0.01%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master    #9148      +/-   ##
==========================================
- Coverage   81.81%   81.80%   -0.02%     
==========================================
  Files         214      214              
  Lines       51373    51383      +10     
==========================================
+ Hits        42033    42034       +1     
- Misses       9340     9349       +9     

Impacted Files	Coverage Δ
rust/arrow/src/buffer.rs	97.72% <ø> (ø)
rust/arrow/src/compute/kernels/aggregate.rs	75.00% <ø> (ø)
rust/arrow/src/compute/kernels/arithmetic.rs	89.83% <ø> (ø)
rust/arrow/src/compute/kernels/comparison.rs	95.91% <ø> (ø)
rust/arrow/src/datatypes.rs	78.75% <ø> (ø)
rust/arrow/src/util/bit_util.rs	100.00% <ø> (ø)
rust/datafusion/src/datasource/csv.rs	65.00% <0.00%> (-16.25%)	⬇️
rust/datafusion/src/datasource/parquet.rs	94.77% <0.00%> (-1.44%)	⬇️
rust/parquet/src/encodings/encoding.rs	95.43% <0.00%> (+0.19%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 08cccd6...181a78d. Read the comment docs.

[jorgecarleitao]

LGTM.

[mqy]

#8891 added simd_x86.
How about add simd: { any(simd_x86, simd_aarch64) } in this PR?

[nevi-me]

How about merging simd_x86 and simd_aarch64 together, on the presumption that as long as we use the same library for both (packed_simd_2 now, stdsimd later), we don't need to split the architectures?

    cfg_aliases! {
        simd: { all(any(target_arch = "x86", target_arch = "x86_64", target_arch = "aarch64"), feature = "simd") },
    }

[mqy]

How about merging simd_x86 and simd_aarch64 together, on the presumption that as long as we use the same library for both (packed_simd_2 now, stdsimd later), we don't need to split the architectures?

    cfg_aliases! {
        simd: { all(any(target_arch = "x86", target_arch = "x86_64", target_arch = "aarch64"), feature = "simd") },
    }

+1 because the codes make no difference for all targets

[mqy]

List some possible TODO/discussion about the SIMD feature:

• Is it make sense to enable the SIMD feature by default [1]? I noticed that servo/pathfinder enables SIMD by default.
• Change static configure to dynamic CPU feature detection, core::arch dynamic-cpu-feature-detection explains.

[1] Quote from https://en.wikipedia.org/wiki/SIMD:
Small-scale (64 or 128 bits) SIMD became popular on general-purpose CPUs in the early 1990s and continued through 1997 and later with Motion Video Instructions (MVI) for Alpha. SIMD instructions can be found, to one degree or another, on most CPUs

[nevi-me]

• Is it make sense to enable the SIMD feature by default

No, because we want stable support by default, so simd is opt-in (ARROW-6717)

• Change static configure to dynamic CPU feature detection

ARROW-6410 mainly deals with this.

I'd opt not to document anything on this PR, as we have activity on JIRA documenting our use of SIMD.
In the long run, we'd likely want SIMD to become stable, so that we can more freely implement SIMD versions of our code, without the nightly constraint.

[mqy]

@nevi-me got it, thanks!

[paddyhoran]

LGTM

[alamb]

Seems like a good change to me

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 81.80%. Comparing base (08cccd6) to head (181a78d).

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #9148      +/-   ##
==========================================
- Coverage   81.81%   81.80%   -0.02%     
==========================================
  Files         214      214              
  Lines       51373    51383      +10     
==========================================
+ Hits        42033    42034       +1     
- Misses       9340     9349       +9     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.