New SDOT Kernels in KleidiAI #11931
Description
🚀 The feature, motivation and pitch
KleidiAI introduced the following new SDOT kernels:
- matmul_clamp_f32_qai8dxp4x4_qsi4c32p4x4_16x4_neon_dotprod
- matmul_clamp_f32_qai8dxp1x4_qsi4c32p4x4_1x4_neon_dotprod
which provide significantly larger performance coverage of ARM CPUs. We will bring these into XNNPACK
Alternatives
No response
Additional context
No response
RFC (Optional)
No response