Skip to content

OpenCAPI/omi_device_ice

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

License

Copyright 2019 International Business Machines

Licensed under the Apache License, Version 2.0 (the "License"); you may not use the files in this repository except in compliance with the License. You may obtain a copy of the License at http://www.apache.org/licenses/LICENSE-2.0

The patent license granted to you in Section 3 of the License, as applied to the "Work," hereby includes implementations of the Work in physical form.

Unless required by applicable law or agreed to in writing, the reference design distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.

The background Specification upon which this is based is managed by and available from the OpenCAPI Consortium. More information can be found at https://opencapi.org.

Purpose

This ICE design contains the FPGA code of an OMI host. It has all the features needed to configure, initialize, train and test an OMI DDIMM through OMI links.

OMI stands for Open Memory Interface. Check OMI web site at https://openmemoryinterface.org

Friendly documentation providing reference documentation, design and enablement code can also be found at https://opencapi.github.io/omi-doc/

Hardware requirements

As of today, this ICE design targets only 1 board:

  • Gemini board which is a lab experimentation board (not for distribution)
    • AMD/Xilinx FPGA XCZU19EG
    • 1 port (x8 OMI links) enabled in this design

Any new board or additional ports can be easily added. Board specific files can be found in ~/ice/src/board/

I2C communication with these boards can be done with a Rasberry Pi.

AMD/Xilinx Vivado 2021.2 is used to build the FPGA binary images.

Code is designed for AMD/Xilinx UltraScale+ FPGA family.

More about enablement on https://opencapi.github.io/omi-doc/blocs/enablement

FPGA Build Scripts

The /build directory contains the various scripts needed to synthesize and implement RTL, as well as generate a bitstream, for an FPGA using the Fire design. (it was targeting fire (host) and ice (device) but only fire is available yet)

To quickly get started, the sequence is ($design is fire ):

cd build
./run $design synthesize.tcl
./run $design implement.tcl
./run $design gen_bitstream.tcl

The following command launches for the entire process, with some optimization choices:

./run_ice.sh

Main Scripts and Directories

run: Takes a TCL script file and a design as an argument, and runs that TCL script on that design using Vivado 2018.3 in the terminal (no GUI). The command output is written to $design/$command.log, and also colorized and printed to the screen. Output files are saved in the $design/ directory, including a design checkpoint (.dcp file).

run_ice.sh: Runs the main synthesize, then implement using strategy 17, and gen_bitstream with implementation strategy result 17. This number will be adapted depending on the design and the WNS result displayed at the end of the place and route phase. If run on a machine with a bsub binary found (aka LSF is installed), then the commands are dispatched via LSF. Additionally, 16 implementation strategies are launched in parallel, and the timing results are checked on completed runs. The first completed implementation strategy that meets timing is used for the bitstream. Also in this mode, the "important" output files are copied to the $design_deploy/ directory.

ice/ : Contain all output files from Vivado. Each command outputs to a log stored in the top-level directory, and other files are stored in a sub-directory per command.

Vivado TCL Scripts

synthesize.tcl: Load libraries and read RTL and constraint files, recompile IP (if needed), synthesize the design, and add the ILA IP.

insert_ila.tcl: Helper script for synthesize.tcl that adds all signals marked with the mark_debug to a correctly clocked Integrated Logic Analyzer (ILA) (Trace Array in IBM speak).

implement.tcl: Implements a synthesized design, and creates various reports. Takes an argument to select the implementation strategy used (uses strategy 1 if not given).

gen_bitstream.tcl: Generate the bistream and debug probes for an implemented design. Takes an argument to select the implementation strategy used (uses strategy 1 if not given).

edit_ip.tcl: Open all the IP in src/ip/ for editing in the Vivado GUI (not in a project). Also used to add new IP.

LSF Wrappers and Configurations

bsub_run: Same as run, but run a command via bsub and print output to the screen.

bsub_batch: Same as run, but run a command via bsub and run in background.

build.lsf: LSF configuration used for all commands submitted via bsub.

clean_all.sh: clean the whole project removing all files generated by scripts.

clean_ice.sh: clean the whole ice project removing all files generated by scripts.

Work-In-Progress

These files are either a work-in-progress or stale, and should not be used.

waived_warnings.txt: List of warnings that are known and acceptable.

print_warnings_ice.sh: Print all the warnings in run.log that are not waived in waived_warnings.txt.

About

An example OMI Device FPGA with 2 DDR4 memory ports

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published