CMU-SAFARI / Pythia

A customizable hardware prefetching framework using online reinforcement learning as described in the MICRO 2021 paper by Bera et al. (https://arxiv.org/pdf/2109.12021.pdf).

Home Page:https://arxiv.org/pdf/2109.12021.pdf

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Logo

A Customizable Hardware Prefetching Framework Using Online Reinforcement Learning

GitHub GitHub release DOI

Table of Contents
  1. What is Pythia?
  2. About the Framework
  3. Prerequisites
  4. Installation
  5. Preparing Traces
  6. Experimental Workflow
  7. HDL Implementation
  8. Code Walkthrough
  9. Citation
  10. License
  11. Contact
  12. Acknowledgements

What is Pythia?

Pythia is a hardware-realizable, light-weight data prefetcher that uses reinforcement learning to generate accurate, timely, and system-aware prefetch requests.

Pythia formulates hardware prefetching as a reinforcement learning task. For every demand request, Pythia observes multiple different types of program context information to take a prefetch decision. For every prefetch decision, Pythia receives a numerical reward that evaluates prefetch quality under the current memory bandwidth utilization. Pythia uses this reward to reinforce the correlation between program context information and prefetch decision to generate highly accurate, timely, and system-aware prefetch requests in the future.

Pythia is presetend at MICRO 2021.

Rahul Bera, Konstantinos Kanellopoulos, Anant V. Nori, Taha Shahroodi, Sreenivas Subramoney, Onur Mutlu, "Pythia: A Customizable Hardware Prefetching Framework Using Online Reinforcement Learning", In Proceedings of the 54th Annual IEEE/ACM International Symposium on Microarchitecture (MICRO), 2021

About The Framework

Pythia is implemented in ChampSim simulator. We have significantly modified the prefetcher integration pipeline in ChampSim to add support to a wide range of prior prefetching proposals mentioned below:

  • Stride [Fu+, MICRO'92]
  • Streamer [Chen and Baer, IEEE TC'95]
  • SMS [Somogyi+, ISCA'06]
  • AMPM [Ishii+, ICS'09]
  • Sandbox [Pugsley+, HPCA'14]
  • BOP [Michaud, HPCA'16]
  • SPP [Kim+, MICRO'16]
  • Bingo [Bakshalipour+, HPCA'19]
  • SPP+PPF [Bhatia+, ISCA'19]
  • DSPatch [Bera+, MICRO'19]
  • MLOP [Shakerinava+, DPC-3'19]
  • IPCP [Pakalapati+, ISCA'20]

Most of the prefetchers (e.g., SPP [1], Bingo [2], IPCP [3]) reuse codes from 2nd and 3rd data prefetching championships (DPC). Others (e.g., AMPM [4], SMS [5]) are implemented from scratch and shows similar relative performance reported by previous works.

Prerequisites

The infrastructure has been tested with the following system configuration:

  • G++ v6.3.0 20170516
  • CMake v3.20.2
  • md5sum v8.26
  • Perl v5.24.1
  • Megatools 1.11.0 (Note that v1.9.98 does NOT work)

Installation

  1. Install necessary prequisites

    sudo apt install perl
  2. Clone the GitHub repo

    git clone https://github.com/CMU-SAFARI/Pythia.git
  3. Clone the bloomfilter library inside Pythia home directory

    cd Pythia
    git clone https://github.com/mavam/libbf.git libbf
  4. Build bloomfilter library. This should create the static libbf.a library inside build directory

    cd libbf
    mkdir build && cd build
    cmake ../
    make clean && make
  5. Build Pythia for single/multi core using build script. This should create the executable inside bin directory.

    cd $PYTHIA_HOME
    # ./build_champsim.sh <l1_pref> <l2_pref> <llc_pref> <ncores>
    ./build_champsim.sh multi multi no 1

    Please use build_champsim_highcore.sh to build ChampSim for more than four cores.

  6. Set appropriate environment variables as follows:

    source setvars.sh

Preparing Traces

  1. Install the megatools executable

    cd $PYTHIA_HOME/scripts
    wget https://megatools.megous.com/builds/builds/megatools-1.11.1.20230212-linux-x86_64.tar.gz
    tar -xvf megatools-1.11.1.20230212-linux-x86_64.tar.gz 

Note: The megatools link might change in the future depending on latest release. Please recheck the link if the download fails.

  1. Use the download_traces.pl perl script to download necessary ChampSim traces used in our paper.

    mkdir $PYTHIA_HOME/traces/
    cd $PYTHIA_HOME/scripts/
    perl download_traces.pl --csv artifact_traces.csv --dir ../traces/

Note: The script should download 233 traces. Please check the final log for any incomplete downloads. The total size of all traces would be ~52 GB.

  1. Once the trace download completes, please verify the checksum as follows. Please make sure all traces pass the checksum test.

    cd $PYTHIA_HOME/traces
    md5sum -c ../scripts/artifact_traces.md5
  2. If the traces are downloaded in some other path, please change the full path in experiments/MICRO21_1C.tlist and experiments/MICRO21_4C.tlist accordingly.

More Traces

  1. We are also releasing a new set of ChampSim traces from PARSEC 2.1 and Ligra. The trace drop-points are measured using Intel Pinplay and the traces are captured by the ChampSim PIN tool. The traces can be found in the following links. To download these traces in bulk, please use the "Download as ZIP" option from mega.io web-interface.

  2. Our simulation infrastructure is completely compatible with all prior ChampSim traces used in CRC-2 and DPC-3. One can also convert the CVP-2 traces (courtesy of Qualcomm Datacenter Technologies) to ChampSim format using the following converter. The traces can be found in the follwing websites:

Experimental Workflow

Our experimental workflow consists of two stages: (1) launching experiments, and (2) rolling up statistics from experiment outputs.

Launching Experiments

  1. To create necessary experiment commands in bulk, we will use scripts/create_jobfile.pl

  2. create_jobfile.pl requires three necessary arguments:

    • exe: the full path of the executable to run
    • tlist: contains trace definitions
    • exp: contains knobs of the experiements to run
  3. Create experiments as follows. Please make sure the paths used in tlist and exp files are appropriate.

    cd $PYTHIA_HOME/experiments/
    perl ../scripts/create_jobfile.pl --exe $PYTHIA_HOME/bin/perceptron-multi-multi-no-ship-1core --tlist MICRO21_1C.tlist --exp MICRO21_1C.exp --local 1 > jobfile.sh
  4. Go to a run directory (or create one) inside experiements to launch runs in the following way:

    cd experiments_1C
    source ../jobfile.sh
  5. If you have slurm support to launch multiple jobs in a compute cluster, please provide --local 0 to create_jobfile.pl

Rolling-up Statistics

  1. To rollup stats in bulk, we will use scripts/rollup.pl

  2. rollup.pl requires three necessary arguments:

    • tlist
    • exp
    • mfile: specifies stat names and reduction method to rollup
  3. Rollup statistics as follows. Please make sure the paths used in tlist and exp files are appropriate.

    cd experiements_1C/
    perl ../../scripts/rollup.pl --tlist ../MICRO21_1C.tlist --exp ../MICRO21_1C.exp --mfile ../rollup_1C_base_config.mfile > rollup.csv
  4. Export the rollup.csv file in you favourite data processor (Python Pandas, Excel, Numbers, etc.) to gain insights.

HDL Implementation

We also implement Pythia in Chisel HDL to faithfully measure the area and power cost. The implementation, along with the reports from umcL65 library, can be found the following GitHub repo. Please note that the area and power projections in the sample report is different than what is reported in the paper due to different technology.

Pythia-HDL Build

Code Walkthrough

Pythia was code-named Scooby (the mistery-solving dog) during the developement. So any mention of Scooby anywhere in the code inadvertently means Pythia.

  • The top-level files for Pythia are prefetchers/scooby.cc and inc/scooby.h. These two files declare and define the high-level functions for Pythia (e.g., invoke_prefetcher, register_fill, etc.).
  • The released version of Pythia has two types of RL engine defined: basic and featurewise. They differ only in terms of the QVStore organization (please refer to our paper to know more about QVStore). The QVStore for basic version is simply defined as a two-dimensional table, whereas the featurewise version defines it as a hierarchichal organization of multiple small tables. The implementation of respective engines can be found in src/ and inc/ directories.
  • inc/feature_knowledge.h and src/feature_knowldege.cc define how to compute each program feature from the raw attributes of a deamand request. If you want to define your own feature, extend the enum FeatureType in inc/feature_knowledge.h and define its corresponding process function.
  • inc/util.h and src/util.cc contain all hashing functions used in our evaluation. Play around with them, as a better hash function can also provide performance benefits.

Citation

If you use this framework, please cite the following paper:

@inproceedings{bera2021,
  author = {Bera, Rahul and Kanellopoulos, Konstantinos and Nori, Anant V. and Shahroodi, Taha and Subramoney, Sreenivas and Mutlu, Onur},
  title = {{Pythia: A Customizable Hardware Prefetching Framework Using Online Reinforcement Learning}},
  booktitle = {Proceedings of the 54th Annual IEEE/ACM International Symposium on Microarchitecture},
  year = {2021}
}

License

Distributed under the MIT License. See LICENSE for more information.

Contact

Rahul Bera - write2bera@gmail.com

Acknowledgements

We acklowledge support from SAFARI Research Group's industrial partners.

About

A customizable hardware prefetching framework using online reinforcement learning as described in the MICRO 2021 paper by Bera et al. (https://arxiv.org/pdf/2109.12021.pdf).

https://arxiv.org/pdf/2109.12021.pdf

License:MIT License


Languages

Language:C++ 56.3%Language:Perl 31.4%Language:XS 7.8%Language:C 2.5%Language:Roff 1.1%Language:Shell 0.8%Language:Makefile 0.2%