[sgbihu]

Details:

• Clamp the lower bound to INT32_MIN for i32 GPU Clamp operations before converting bounds to float. This prevents out-of-range convert_int values when the original lower bound is below the i32 range, such as INT64_MIN.
• Add CPU and GPU functional test cases using ActivationLayerTest for i32 and i64 inputs with lower bounds INT64_MIN, -2147483904, and INT32_MIN.

Tickets:

• CVS-199222

AI Assistance:

• AI assistance used: no / yes
• If yes, summarize how AI was used and what human validation was performed (build/tests/manual checks).

[Copilot]

🟢 Approval recommended

The GPU Clamp lower-bound change has no active review issues blocking approval.

No open issues 🎉

What changed in this PR

This PR prevents out-of-range lower-bound conversion in GPU Clamp operations when i64 inputs are converted to i32 internally.

Changes:

• Restrict the GPU i32 Clamp lower bound to at least INT32_MIN.
• Add CPU and GPU regression cases for i32 and i64 inputs with lower bounds at INT64_MIN, below INT32_MIN, and at INT32_MIN.

File	Description
src/​plugins/​intel_gpu/​tests/​functional/​shared_tests_instances/​single_layer_tests/​activation.cpp	Adds six GPU Clamp lower-bound regression cases.
src/​plugins/​intel_gpu/​src/​plugin/​ops/​unary.cpp	Floors the i32 Clamp lower bound at INT32_MIN before conversion.
src/​plugins/​intel_cpu/​tests/​functional/​shared_tests_instances/​single_layer_tests/​activation.cpp	Adds matching CPU Clamp lower-bound regression cases.

_{Review effort: Balanced}

[e-ddykim]

LGTM