Search papers, labs, and topics across Lattice.
This paper introduces LACE, a multi-agent framework that leverages large language models to streamline the implementation and validation of Domain-specific Instruction Set Architecture eXtensions (ISAX) in RISC-V cores. By translating natural language ISAX intents into a compact intermediate representation and performing retrieval-guided RTL edits, LACE significantly enhances the accuracy of instruction generation while minimizing integration efforts. The framework achieves a remarkable 72.8% pass@1 generation accuracy across four embedded RISC-V cores, marking a substantial improvement over existing methods.
LACE transforms the RISC-V instruction extension process, achieving a 72.8% accuracy in generating ISAX implementations from natural language intents, a leap from near-zero accuracy.
Domain-specific Instruction Set Architecture eXtensions (ISAX) are widely adopted in the RISC-V ecosystem to accelerate emerging workloads, but implementing and validating ISAXes across different cores remains slow and fragmented. Existing frameworks still require per-core interface adaptation, and differential testing often breaks once either the microarchitecture or the ISAX changes. We present LACE, an LLM-aided multi-agent workflow that translates natural-language ISAX intents into a compact two-level IR (operation-level and HDL task-level), performs retrieval-guided localized RTL edits over large repositories, and closes the loop with a compiler-agnostic riscv-formal checking flow (assuming RVFI availability or instrumentation). Across four embedded RISC-V cores, LACE raises pass@1 generation accuracy from near-zero to 72.8\% within our evaluation setup, while improving code localization and reducing integration rework. The code of LACE is available at https://github.com/UMN-ZhaoLab/LACE.
