This tool serves as a general-purpose instruction set architecture metamodel. A parser for CoreDSL and an architecture generator for the instruction set simulator ETISS are currently also provided.
Please note: M2-ISA-R is in heavy development, things might change and break without notice. Please report issues on GitHub for any problems you encounter.
- Python 3.10+ with at least
pipandvenv
- Make a Python
venvsomewhere:python -m venv <path-to-venv> - Activate said
venv:source <path-to-venv>/bin/activate - Install M2-ISA-R:
pip install git+https://github.com/tum-ei-eda/M2-ISA-R.git
- Clone the repository, change into the cloned directory
- Make a Python
venv:python -m venv venv - Activate said
venv:source venv/bin/activate - Install the current checkout and its command-line tools:
pip install -e . - To install the published PyPI release instead:
pip install m2isar
Implicit truncation is an error by default. Use -Wno-error=implicit-trunc to
retain it as a warning, or -Wno-implicit-trunc to suppress it. Likewise,
-Werror -Wno-error=foo promotes enabled warnings except foo.
A later -Werror=foo promotes that warning again.
Run python -m unittest discover -s tests -v from the repository root.
See the test guide for feature-specific runs, optional MINRES
comparisons, and adding CDSL or Python metamodel regression cases.
M2-ISA-R consists of 3 components, two of which are exchangeable for different needs:
Frontend -> Metamodel -> Backend
The frontend transforms a model specification into M2-ISA-R's internal architecture model. This model can then be transformed again to an output format, e.g. models for an ISS. This repo provides CoreDSL 2 and CoreDSL 2 instruction-set frontends and several backends.
M2-ISA-R provides CoreDSL 2 frontends and backends for ETISS and experimental Spike plugins. Basic commands (run after installation):
- To parse a CoreDSL 2 description:
coredsl2_parser path/to/input/<top_level>.core_desc(writes the model underpath/to/input/gen_model/; use--helpfor validation, include path, and output options). - To generate an ETISS architecture:
etiss_writer path/to/input/gen_model/<top_level>.m2isarmodel - To generate a Spike extension:
spike_writer path/to/input/gen_model/<top_level>.m2isarmodel(see supported subset and build instructions).
The ETISS writer accepts --fill-mode empty|auto to choose whether architecture-specific implementation files are placeholders or populated from available model metadata. Pass GDB register-description mappings as --gdb-xml-descr <core-name>:<xml-file>... to generate the corresponding gdbserver register mapping and XML support. See ETISS writer options.
- Frontends:
coredsl2parses CoreDSL 2 core descriptions;coredsl2_setparses CoreDSL 2 instruction-set descriptions. - Transforms: type inference (
infer_types), operand inference (infer_operands), behavior validation (validate_behav), instruction encoding planning/application (encode_instructions), loop lowering (lower_loops), and merging always blocks (merge_always_blocks). - Backends and tools: ETISS architecture generation (
etiss), experimental Spike extension generation (spike), CoreDSL 2 and instruction-set export (coredsl2,coredsl2_set), model browsing (viewer), disassembly (disass), trace-model generation (trace_gen), ISA manual generation (isa_manual), and coverage processing (coverage).
Installed command-line entry points include coredsl2_parser, etiss_writer, spike_writer, m2isar_encode, m2isar_viewer, m2isar2lcov, and trace_gen. Some frontends, transforms, and backends are library modules without a dedicated installed command.
For parsers, see the CoreDSL 2 frontend and the CoreDSL 2 set frontend. For the ETISS architecture writer, see m2isar/backends/etiss.