[Good First Issue] [ARM]: Implement CPU plugin just-in-time emitter for Power operation

[a-sidorova]

Context

JIT Emitters are part of code generation feature (a.k.a. tensor compiler) that automatically produces highly-efficient optimized fused subgraph binary code. Each emitter implements specific operation from low level OpenVINO dialect.

Prerequisites

Recommended to use ARM CPU based platform for development (e.g. Mac, Raspberry Pi etc). Emulators (e.g. QEMU) is still an option, but not that convenient, especially for final performance evaluation.

What needs to be done?

• CPU Plugin has optimization for Power operation with scalar second input - converts such Power ops to CPU-opset op PowerStatic. Currently, CPU Plugin has JIT Emitter only for PowerStatic op with scalar second input for ARM64 SIMD platforms. Need to implement JIT emitter jit_power_dynamic_emitter for eltwise algorithm PowerDynamic - Power op with non-scalar second inputs. Use OpenVINO CPU plugin JIT emitters documentation and Arm A-profile A64 Instruction Set Architecture documentation for details.
  Hint: the emitter should emit binary call for the STL function powf as it's done in jit_power_dynamic_emitter for x64 platforms.
• Remove limitation condition from ARM64 JIT Executor.

Before emitter implementation, please, modify tests to be sure that developed functionality is covered by test:

• Add jit kernel check in eltwise test, the test must fail.

Tests

Tests are disabled in default build, so ensure to add -DENABLE_TESTS=ON into cmake command.

GoogleTest is used for testing. CPU functional test target is ov_cpu_func_tests. You can use GoogleTest filter:

./bin/[platform]/[build_type]/ov_cpu_func_tests --gtest_filter="*smoke*Eltwise*Pow*"

Examples

• PR with Abs operation support: [CPU] [ARM64] jit abs #23692
• Implementation of jit_power_dynamic_emitter for x64 platforms: link.
• Implementation of jit_power_static_emitter for ARM64 platforms that already emits binary function call of powf in some corner cases: link.

Resources

• Contribution guide - start here!
• What is OpenVINO?
• CPU plugin JIT emitters
• Blog post on contributing to OpenVINO
• User documentation
• Intel DevHub Discord channel - engage in discussions, ask questions and talk to OpenVINO developers
• How to link your Pull Request to an issue

Contact points

@a-sidorova, @dmitry-gorokhov

[jpy794]

Hi, can I take this?

[geeky33]

.take

[github-actions]

Thank you for looking into this issue! Please let us know if you have any questions or require any help.

[a-sidorova]

@geeky33 I see that @jpy794 was first who was interested in the issue - but unfortunately didn't write .take to automatically be assignee. Also I see that he already prepared the PR #29412 which looks pretty cool. So I'm going to reassign it to @jpy794 😊

@geeky33 you have already taken many issues (including one more ARM-related issue). So I believe that you don't mind. if you have some objections, feel free to write them