W8A8/W4A8 inference on Apple Silicon — unlocking unused INT8 TensorOps in M5 for 1.2–1.9× faster LLM prefill, built as MLX custom primitives.
-
Updated
May 3, 2026 - Python
#
w4a8
Here are 5 public repositories matching this topic...
KV260 integration lane for PCCX™ v002 LLM IP-core bring-up, validation, and board/runtime evidence.
fpga transformer rtl xilinx systemverilog quantization bare-metal gemma inference-engine npu edge-ai hardware-accelerator kv260 kria llm ai-accelerator gemma3n w4a8 pccx pccx-v002
-
Updated
May 8, 2026 - Shell
PCCX™ specification, documentation, and ecosystem coordination hub for open AI accelerator IP.
docs fpga transformer rtl isa systemverilog computer-architecture quantization gemm inference-engine npu edge-ai hardware-accelerator kv-cache gemv llm ai-accelerator w4a8 pccx
-
Updated
May 8, 2026 - SystemVerilog
PCCX™ vision-v001 compatibility track for CNN inference planning and v002/Vision absorption review.
computer-vision fpga cnn rtl vision xilinx systemverilog quantization inference-engine npu opencore edge-ai hardware-accelerator kv260 kria ai-accelerator w4a8 pccx pccx-v002 pccx-vision
-
Updated
May 8, 2026
PCCX™ v002 IP-core package — board- and model-agnostic reusable RTL for LLM, Vision, Voice, and common subsystems.
fpga transformer rtl isa systemverilog quantization gemm sail ip-core npu opencore edge-ai hardware-accelerator kv-cache gemv llm ai-accelerator w4a8 pccx pccx-v002
-
Updated
May 8, 2026 - SystemVerilog
Improve this page
Add a description, image, and links to the w4a8 topic page so that developers can more easily learn about it.
Add this topic to your repo
To associate your repository with the w4a8 topic, visit your repo's landing page and select "manage topics."