https://github.com/kumasento/dfe-mlir
MLIR dialect for data-flow engine design
https://github.com/kumasento/dfe-mlir
data-flow fpga mlir
Last synced: 26 days ago
JSON representation
MLIR dialect for data-flow engine design
- Host: GitHub
- URL: https://github.com/kumasento/dfe-mlir
- Owner: kumasento
- License: mit
- Created: 2020-07-13T21:37:52.000Z (almost 5 years ago)
- Default Branch: master
- Last Pushed: 2020-07-25T16:55:05.000Z (over 4 years ago)
- Last Synced: 2025-02-02T02:25:02.429Z (3 months ago)
- Topics: data-flow, fpga, mlir
- Language: C++
- Homepage:
- Size: 106 KB
- Stars: 4
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# dfe-mlir
MLIR dialect for data-flow engine design
## Core Features
* A `maxj` dialect that can concisely represent most data-flow designs achievable by MaxJ (e.g., multi-kernel and LMem designs). We can perform optimization on it, and designs described by it can be finally translated to valid MaxJ code.
## Install
Please make sure your GCC and CMake satisfies the latest requirements from LLVM.
```shell
git clone --recursive https://github.com/kumasento/dfe-mlir
# build llvm
cd dfe-mlir/lib/llvm-project
mkdir build
cd build
cmake \
-DLLVM_ENABLE_PROJECTS="clang;clang-tools-extra;lld;mlir" \
-DLLVM_TARGETS_TO_BUILD="X86" \
-DLLVM_ENABLE_ASSERTIONS=ON \
-DLLVM_BUILD_LLVM_DYLIB=ON \
-DLLVM_INCLUDE_TESTS=OFF -G "Ninja" ../llvm
ninja -j$(nproc)
# build dfe-mlir
cd ../../../
mkdir build
cd build
cmake -G "Ninja" ..
ninja
```
## Test
### The `maxj` Dialect