Views
No views yet
| Parameter | Value |
|---|---|
| Cores | 128 |
| Neurons per core | 1,024 |
| Total neurons | 131,072 |
| Neuron model | Leaky Integrate-and-Fire (16-bit fixed-point) |
| Synapse pool | 131,072 entries per core |
| Learning | STDP, 14-opcode programmable learning ISA |
| Network-on-Chip | Configurable XY mesh with multicast |
| Host interface | UART (FPGA) / AXI-Lite (F2) / PCIe MMIO |
| Management | RV32IM RISC-V cluster |
| Multi-chip | Chip link with routing table |
| Clock | 62.5 MHz |
catalyst-n1/
rtl/ Verilog sources (core, NoC, memory, host, RISC-V)
tb/ Testbenches (unit, integration, regression)
sdk/ Python SDK with CPU, GPU, and FPGA backends
fpga/ FPGA build files (Arty A7, AWS F2, Kria K26)
sim/ Simulation scripts and visualization
Makefile Compile and run simulation1# Compile and run basic simulation
2make sim
3
4# Run full regression
5bash run_regression.sh
6
7# Run a single testbench
8iverilog -g2012 -DSIMULATION -o out.vvp \
9 rtl/sram.v rtl/spike_fifo.v rtl/uart_tx.v rtl/uart_rx.v \
10 rtl/scalable_core_v2.v rtl/neuromorphic_mesh.v \
11 rtl/host_interface.v rtl/neuromorphic_top.v rtl/sync_tree.v \
12 rtl/rv32i_core.v rtl/mmio_bridge.v rtl/rv32im_cluster.v \
13 tb/tb_p24_final.v
14vvp out.vvp
15
16# View waveforms (requires GTKWave)
17make wavessdk/README.md for full documentation.1cd sdk
2pip install -e .1import neurocore as nc
2
3net = nc.Network()
4inp = net.population(100, params={'threshold': 1000, 'leak': 10}, label='input')
5hid = net.population(50, params={'threshold': 1000, 'leak': 5}, label='hidden')
6out = net.population(10, params={'threshold': 1000, 'leak': 5}, label='output')
7
8net.connect(inp, hid, weight=500, probability=0.3)
9net.connect(hid, out, weight=400, probability=0.5)
10
11sim = nc.Simulator()
12sim.deploy(net)
13
14for t in range(100):
15 sim.inject(inp, neuron_ids=[0, 5, 10], current=1500)
16 sim.step()
17
18result = sim.get_result()
19result.raster_plot(show=True)1# Vivado batch build
2vivado -mode batch -source fpga/build_vivado.tclfpga/arty_a7.xdc. Top module: fpga/fpga_top.v.1# Build on F2 build instance
2cd fpga/f2
3bash run_build.shfpga/f2/cl_neuromorphic.sv. Host driver: fpga/f2_host.py.vivado -mode batch -source fpga/kria/build_kria.tclfpga/kria/kria_neuromorphic.v.1cd sdk
2python benchmarks/shd_train.py --data-dir benchmarks/data/shd --epochs 200
3python benchmarks/shd_deploy.py --checkpoint benchmarks/shd_model.pt --data-dir benchmarks/data/shdsdk/benchmarks/: DVS gesture recognition, XOR classification, temporal patterns, scaling, stress tests.