NPU and Vector DSP Acceleration for High-Frequency Message Processing

Leveraging dedicated hardware accelerators to unpack, decrypt, and parse 1,000+ packets per second.

Heterogeneous Automotive SoCs Architecture & Design

Reference die floorplan for heterogeneous automotive SoCs, mapping NPU, vector DSP, and general-purpose CPU clusters against the V2X message ingestion pipeline, including on-chip interconnect topology and memory bandwidth partitioning.

Production Implementation & Verification

Production silicon validation covering sustained decode throughput above 1,000 packets per second, thermal throttling behavior under worst-case burst traffic, and AEC-Q100 Grade 2 qualification results.

Dedicated vector DSP cores decrypt and parse IEEE 1609.2 signed BSM/CPM payloads at line rate without touching the application CPU.
On-chip NPUs run quantized INT8 object-classification models fed directly by the V2X message decoder, cutting perception latency by half.
Hardware crypto accelerators verify ECDSA P-256 signatures for 1,000+ messages per second while maintaining certificate revocation list lookups in SRAM.
AMBA CHI-based mesh interconnects isolate safety-critical ASIL-D cores from best-effort infotainment workloads on the same die.

Explore Domain Portfolio