Clean warnings in `make sw` target

Snitch Cluster

This repository hosts the hardware and software for the Snitch cluster and its generator. Snitch is a high-efficiency compute cluster platform focused on floating-point workloads. It is developed as part of the PULP project, a joint effort between ETH Zurich and the University of Bologna.

Getting Started

To get started, check out the getting started guide.

Content

What can you expect to find in this repository?

The RISC-V Snitch integer core. This can be useful stand-alone if you are just interested in re-using the core for your project, e.g., as a tiny control core or you want to make a peripheral smart. The sky is the limit.
The Snitch cluster. A highly configurable cluster containing one to many integer cores with optional floating-point capabilities as well as our custom ISA extensions Xssr, Xfrep, and Xdma.
A runtime and example applications for the Snitch cluster.
RTL simulation environments for Verilator, Questa Advanced Simulator, and VCS, as well as configurations for our Banshee system simulator

This code was previously hosted in the Snitch monorepo and was spun off into its own repository to simplify maintenance and dependency handling. Note that our Snitch-based manycore system Occamy has also moved.

Tool Requirements

verilator >= v4.1
bender >= v0.27.0

License

Snitch is being made available under permissive open source licenses.

The following files are released under Apache License 2.0 (Apache-2.0) see LICENSE:

sw/
util/

The following files are released under Solderpad v0.51 (SHL-0.51) see hw/LICENSE:

hw/

The sw/deps directory references submodules that come with their own licenses. See the respective folder for the licenses used.

sw/deps/

Publications

If you use the Snitch cluster or its extensions in your work, you can cite us:

Snitch: A tiny Pseudo Dual-Issue Processor for Area and Energy Efficient Execution of Floating-Point Intensive Workloads

@article{zaruba2020snitch,
  title={Snitch: A tiny Pseudo Dual-Issue Processor for Area and Energy Efficient Execution of Floating-Point Intensive Workloads},
  author={Zaruba, Florian and Schuiki, Fabian and Hoefler, Torsten and Benini, Luca},
  journal={IEEE Transactions on Computers},
  year={2020},
  publisher={IEEE}
}

Stream semantic registers: A lightweight risc-v isa extension achieving full compute utilization in single-issue cores

@article{schuiki2020stream,
  title={Stream semantic registers: A lightweight risc-v isa extension achieving full compute utilization in single-issue cores},
  author={Schuiki, Fabian and Zaruba, Florian and Hoefler, Torsten and Benini, Luca},
  journal={IEEE Transactions on Computers},
  volume={70},
  number={2},
  pages={212--227},
  year={2020},
  publisher={IEEE}
}

Sparse Stream Semantic Registers: A Lightweight ISA Extension Accelerating General Sparse Linear Algebra

@article{scheffler2023sparsessr,
  author={Scheffler, Paul and Zaruba, Florian and Schuiki, Fabian and Hoefler, Torsten and Benini, Luca},
  journal={IEEE Transactions on Parallel and Distributed Systems},
  title={Sparse Stream Semantic Registers: A Lightweight ISA Extension Accelerating General Sparse Linear Algebra},
  year={2023},
  volume={34},
  number={12},
  pages={3147-3161},
  doi={10.1109/TPDS.2023.3322029}
}

A High-performance, Energy-efficient Modular DMA Engine Architecture

@ARTICLE{benz2023idma,
  author={Benz, Thomas and Rogenmoser, Michael and Scheffler, Paul and Riedel, Samuel and Ottaviano, Alessandro and Kurth, Andreas and Hoefler, Torsten and Benini, Luca},
  journal={IEEE Transactions on Computers},
  title={A High-performance, Energy-efficient Modular DMA Engine Architecture},
  year={2023},
  volume={},
  number={},
  pages={1-14},
  doi={10.1109/TC.2023.3329930}}

MiniFloat-NN and ExSdotp: An ISA Extension and a Modular Open Hardware Unit for Low-Precision Training on RISC-V Cores

@inproceedings{bertaccini2022minifloat,
  author={Bertaccini, Luca and Paulin, Gianna and Fischer, Tim and Mach, Stefan and Benini, Luca},
  booktitle={2022 IEEE 29th Symposium on Computer Arithmetic (ARITH)},
  title={MiniFloat-NN and ExSdotp: An ISA Extension and a Modular Open Hardware Unit for Low-Precision Training on RISC-V Cores},
  year={2022},
  volume={},
  number={},
  pages={1-8}
}

Soft Tiles: Capturing Physical Implementation Flexibility for Tightly-Coupled Parallel Processing Clusters

@inproceedings{paulin2022softtiles,
  author={Paulin, Gianna and Cavalcante, Matheus and Scheffler, Paul and Bertaccini, Luca and Zhang, Yichao and Gürkaynak, Frank and Benini, Luca},
  booktitle={2022 IEEE Computer Society Annual Symposium on VLSI (ISVLSI)},
  title={Soft Tiles: Capturing Physical Implementation Flexibility for Tightly-Coupled Parallel Processing Clusters},
  year={2022},
  volume={},
  number={},
  pages={44-49},
  doi={10.1109/ISVLSI54635.2022.00021}
}

SARIS: Accelerating Stencil Computations on Energy-Efficient RISC-V Compute Clusters with Indirect Stream Registers

@misc{scheffler2024saris,
      title={SARIS: Accelerating Stencil Computations on Energy-Efficient
             RISC-V Compute Clusters with Indirect Stream Registers},
      author={Paul Scheffler and Luca Colagrande and Luca Benini},
      year={2024},
      eprint={},
      archivePrefix={arXiv},
      primaryClass={cs.MS}
}

	'fseq_fpu_yield':
	safe_div(
	safe_div(seg['fpss_fpu_issues'], seg['snitch_fseq_offloads']),
	fpss_fpu_rel_issues),

	def launch(self, run_dir=None, dry_run=False):
	# Default to current working directory as simulation directory
	if not run_dir:
	run_dir = Path.cwd()

	# Print launch message and simulation command
	cprint(f'Run test {colored(self.elf, "cyan")}', attrs=["bold"])
	cmd_string = ' '.join(self.cmd)
	print(f'$ {cmd_string}', flush=True)

	# Launch simulation if not doing a dry run
	if not dry_run:
	# Create run directory and log file
	os.makedirs(run_dir, exist_ok=True)
	self.log = run_dir / self.LOG_FILE
	# Launch simulation subprocess
	with open(self.log, 'w') as f:
	self.process = subprocess.Popen(self.cmd, stdout=f, stderr=subprocess.STDOUT,
	cwd=run_dir, universal_newlines=True)

	# Launch simulation in current working directory, by default
	if run_dir is None:
	run_dir = Path.cwd()
	# Create unique subdirectory for each test under run directory, if multiple tests
	if uniquify_run_dir:
	unique_run_dir = run_dir / running_sims[-1].testname
	else:
	unique_run_dir = run_dir
	running_sims[-1].launch(run_dir=unique_run_dir, dry_run=dry_run)

	/// Synchronize cores in a cluster with a hardware barrier
	inline void snrt_cluster_hw_barrier() {
	asm volatile("csrr x0, 0x7C2" ::: "memory");
	}

	fpss_fpu_rel_issues = safe_div(seg['fpss_fpu_issues'],
	seg['fpss_issues'])

	'fseq_yield':
	safe_div(seg['fpss_issues'], seg['snitch_fseq_offloads']),

	def successful(self):
	return self.process.returncode == 0

pulp-platform / snitch_cluster Goto Github PK

snitch_cluster's Introduction

Snitch Cluster

Getting Started

Content

Tool Requirements

License

Publications

snitch_cluster's People

Contributors

Stargazers

Watchers

Forkers

snitch_cluster's Issues

Bug description

Solution

1. Fix dry run bug

2. Override run directory per test

Recommend Projects

Recommend Topics

Recommend Org