vLLM-CENT

vLLM-CENT compiles model-specific operations into a typed CENT program. The first frontend lowers one batch-one Llama decode block; the generic cent/ package contains no Llama or transformer definitions.

This project does not execute a model and does not require an emulator. Its output is an ordered sequence of instructions for the architecture described in the CENT paper.

Paper-faithful instruction layer

The instruction classes model all commands in Tables 2 and 3 with their full operand lists:

• Near-bank PUs: MAC_ABK, EW_MUL, and AF.
• PNM units: EXP, RED, ACC, and RISCV.
• CXL movement: SEND_CXL, RECV_CXL, and BCAST_CXL.
• Shared Buffer and DRAM movement: WR_SBK, RD_SBK, and WR_ABK.
• Global Buffer and PU movement: COPY_BKGB, COPY_GBBK, WR_BIAS, RD_MAC, and WR_GB.

Python fields use readable names while docstrings retain the paper terms. For example, CentMemoryAddress.row is the paper's RO, column is CO, and CentSharedBufferAddress.slot is used as Rs or Rd according to whether it is a source or destination. accumulation_register is Regid.

OPsize is represented by operation_size. One instruction causes the CENT decoder to generate that many micro-operations over consecutive 256-bit Shared Buffer slots and DRAM column regions. A five-burst transfer that fits in one row is therefore one instruction with operation_size=5, not five duplicate instructions.

render_text_program emits the Table 2/3 assembly. The package deliberately does not model SYNC, EOC, or RD_AF: a full-text search of the paper found no such instructions, and they belong to the old reference simulator's trace protocol rather than CENT's published ISA. CentProgram is therefore just an immutable, nonempty ordered tuple of real ISA instructions; finishing a builder does not append an artificial terminator.

Architecture boundary

models/llama/                  cent/
--------------------------    --------------------------
Llama dimensions              CENT hardware geometry
Llama tensor layout           complete typed CENT ISA
Llama lowering stages         hardware-aware validation
Llama block orchestration     generic program builder
                               assembly renderers
             \                 /
              CompileRequest
                    |
          compile_transformer_block
                    |
               CentProgram

A future model family belongs in its own models/<family>/ package and reuses the target types. compiler.py remains only a small family dispatcher.

Current public interface

• LlamaModelSpec describes one Llama block.
• CentHardwareSpec describes DRAM geometry, the 64KB Shared Buffer as 2,048 256-bit slots by default, PU accumulation-register capacity, and the target ABI's sigmoid AFid.
• CentBlockPlacementSpec assigns physical channels to one block.
• DecodeStepSpec provides current and reserved context lengths.
• CompileRequest combines the source model, target, placement, and decode step.
• compile_transformer_block returns a validated CentProgram.
• render_text_program emits paper-ISA assembly.

The Llama frontend currently lowers RMSNorm, Q/K/V projections, rotary embedding data movement and multiplication, KV-cache updates, attention-score GEMV, the existing two-pass softmax data path, value-cache GEMV, output projection, residual additions, and the gated SiLU feed-forward network.

Two paper-to-implementation gaps remain explicit rather than guessed:

• The paper names AFid but does not publish numeric function IDs. Rather than guessing, CentHardwareSpec.sigmoid_activation_function_id requires the caller to provide the target ABI's value. Tests use zero only as a documented fake target value.
• The paper explains that uncommon operations such as reciprocal and square root run through RISCV, but does not publish their program-counter entry addresses. Consequently the current softmax/RMSNorm lowering is not yet a complete executable paper algorithm; wiring those operations requires a target runtime ABI that supplies the correct PC values.

Run the tests

From this directory, either install the package in editable mode or expose its src directory:

python3 -m pip install -e .
python3 -m unittest discover -s tests -v

PYTHONPATH=src python3 -m unittest discover -s tests -v

The tests render all 18 paper instructions with distinct operands, validate DRAM/Shared Buffer/register boundaries, verify immutable static opcodes, test every nontrivial lowering helper directly, preserve exact small-model golden counts, and compile Llama 3 8B and 70B block dimensions.