wyooyw/CIMCompiler

DSL and Compiler for Digital SRAM-CIM Architecture

Environment

Build llvm.

The commit-id I use is 0977504537b4dd945fd91fe11eb1a3165297e64a

The script I use to build llvm is:

mkdir build
cd build
cmake -G Ninja ../llvm \
   -DCMAKE_BUILD_TYPE=Release \
   -DCMAKE_C_FLAGS_RELEASE="-O3" \
   -DCMAKE_CXX_FLAGS_RELEASE="-O3" \
   -DLLVM_ENABLE_PROJECTS="mlir;lld"\
   -DLLVM_BUILD_EXAMPLES=ON \
   -DLLVM_TARGETS_TO_BUILD="Native" \
   -DLLVM_ENABLE_ASSERTIONS=ON \
   -DLLVM_USE_SPLIT_DWARF=ON \
   -DCMAKE_C_COMPILER=clang \
   -DCMAKE_CXX_COMPILER=clang++ \
   -DLLVM_ENABLE_LLD=ON \
   -DLLVM_CCACHE_BUILD=ON \
   -DLLVM_ENABLE_RTTI=ON
cmake --build . --target check-mlir -j8

Quick Start

Build the compiler

bash build.sh

Run the test

bash test.sh

Use the compiler

Compile .cim file

bash run.sh

Compile Neural Network

bash run_model.sh

Model files are not pushed yet.

Operator Template

Operator templates are written in CIM-DSL and Jinja-Template. Here are the convolution operator templates:

• Dense Conv2d: op/dense/dense_conv2d_group/code_template.cim
• Value Sparse Conv2d: op/value_sparse/value_sparse_group_longer/code_template.cim
• Bit Sparse Conv2d: op/bit_sparse/bit_sparse_conv2d_group/code_template.cim
• Value&Bit Sparse Conv2d: op/value_bit_sparse/value_bit_sparse_base/code_template.cim