- Existing integral and floating casts involving Long and Double types are only efficient on AVX512DQ targets.
- Intel E-Core Xerons only support AVX2 feature.

Optimize these casts using newly proposed Lowering phase.