FPGA Accelerate your Applications with OpenCAPI Acceleration...
summary
OC-Accel is an open-source OpenCAPI platform that turns C/C++ HLS kernels into FPGA accelerators on Xilinx/AlphaData cards, wiring logic to host memory coherently over OpenCAPI. A delivered image-filter example compiles, simulates and runs on Power9, with measured transfers of ~18 GB/s host→FPGA and ~21 GB/s FPGA→host for a 256 MB blob (local card RAM ~12 GB/s). A PSI detector case ramps acquisition from 4 MP to 10 MP at 2.2 kHz in months, and a Python front-end drives deployment and testing.
AI-written summary · about
chapters (12)
- 0:00Introduction to OpenCAPI
- 2:53What is OpenCAPI
- 7:11FPGA technology overview
- 10:04GPU vs FPGA acceleration
- 13:30OCXL framework introduction
- 14:13OCXL workflow and examples
- 17:23Setting up development tools
- 18:46Simulating FPGA logic
- 19:30Image filter example
- 21:57Python integration with OCXL
- 22:38Philippe’s test case
- 24:28Q&A session
related talks
14mAn Open-Source CVA6S+ High-Performance, Cache-Coherent Cluster for 64b Automotive MPUs
40mKeynote: Open Source FPGA Tooling
20mOpen-Source FPGAs: Tools, Flows, and Visualization
07mFPGA and open hardware
25mHow open is OpenPOWER?
16mSoftware-driven Hardware in the Data Center: LibreBMC, Renode, RowHammer...
21mOpen Standard Software Acceleration for RISC-V
30mRISC-V: Enabling Open Physical AI