6 releases
0.2.0 | Jun 23, 2023 |
---|---|
0.2.0-rc3 | Feb 24, 2023 |
0.1.2 | Apr 2, 2022 |
#723 in Hardware support
31 downloads per month
5.5MB
24K
SLoC
mcl-rs
This project hosts the high-level wrappers of the MCL rust bindings.
Summary
This crate provides high-level, rust-friendly bindings for MCL. The purpose of these bindings are to expose a user-friendlier (and safer) API to what the low-level libmcl-sys API offers. It provides wrappers for most mcl public functions and tries to provide safety at compilation type, however, because of the nature of the library counting on a C project there are cases that it's only possible to catch errors at runtime, as well as a few APIS that currently cannot be checked at all and are thus marked unsafe (but are protected by a feature flag).
Building mcl-rs
Required libraries/ crates
- libmcl-sys and its dependencies
- Clang
- OpenCL
- Autotools
- MCL (either manually installed or via
cargo install mcl_sched
)
- Other crates listed in Cargo.toml
Instructions
mcl-rs depends on the crate libmcl-sys which provides the low-level bindings between the C library of MCL and these higher bindings.
libmcl-sys
makes use of clang to generate the low-level rust binding from the MCL
header file, so if clang is not available it must be installed to the system.
- Install clang
Once all dependencies have been taken care of, we can build mcl-rs.
cargo build --release
Installing MCL Scheduler
The MCL scheduler can easily be installed via:
cargo install mcl_sched
Note, if you have manually built MCL from the C source code, you will already have the mcl_sched
binary in the MCL install directory.
You are free to use either your manually built mcl_sched or the one installed via cargo
FEATURE FLAGS
We re expose three feauture flags (from libmcl-rs), losely corresponding to configuration options of the underlying MCL c-library
- mcl_debug - enables debug logging output from the underlying MCL c-libary
- shared_mem - enables interprocess host shared memory buffers -- this enables a few unsafe APIs
- pocl_extensions - enables interprocess device based shared memory buffers, requires a patched version of POCL 1.8 to have been succesfully installed (please see https://github.com/pnnl/mcl/tree/dev#using-custom-pocl-extensions for more information) -- this enables a few unsafe APIs
Testing
mcl-rs comes with a set of unit tests that can be executed with:
cargo test <test_name>
Reminder: The MCL scheduler should be running when executing the tests. if you installed mcl_sched via cargo then you should be able to invoke directly:
mcl_sched
If you built mcl manually you may need to specify the path to the mcl_sched binary
Removing the test-name would cause cargo to run all available tests, however, this could create issues since tests would run in parallel causing multiple threads to try to acquire access to the mcl_scheduler shmem object at the same time which might lead to failure.
Documentation
Use cargo doc --open
to build and open the documentation of this crate.
STATUS
MCL, libmcl-sys, and mcl-rs are research prototypes and still under development, thus not all intended features are yet implemented.
CONTACTS
Please, contact Roberto Gioiosa at PNNL (roberto.gioiosa@pnnl.gov) if you have any MCL questions. For Rust related questions please contact Ryan Friese at PNNL (ryan.friese@pnnl.gov)
MCL-Rust Team
Roberto Gioiosa
Ryan Friese
Polykarpos Thomadakis
LICENCSE
This project is licensed under the BSD License - see the LICENSE file for details.
REFERENCES
IF you wish to cite MCL, please, use the following reference:
- Roberto Gioiosa, Burcu O. Mutlu, Seyong Lee, Jeffrey S. Vetter, Giulio Picierro, and Marco Cesati. 2020. The Minos Computing Library: efficient parallel programming for extremely heterogeneous systems. In Proceedings of the 13th Annual Workshop on General Purpose Processing using Graphics Processing Unit (GPGPU '20). Association for Computing Machinery, New York, NY, USA, 1–10. DOI:https://doi.org/10.1145/3366428.3380770
Other work that leverage or describe additional MCL features:
- A. V. Kamatar, R. D. Friese and R. Gioiosa, "Locality-Aware Scheduling for Scalable Heterogeneous Environments," 2020 IEEE/ACM International Workshop on Runtime and Operating Systems for Supercomputers (ROSS), 2020, pp. 50-58, doi:10.1109/ROSS51935.2020.00011.
- Rizwan Ashraf and Roberto Gioiosa, "Exploring the Use of Novel Spatial Accelerators in Scientific Applications" 2020 ACM/SPEC International Conference on Performance Engineering (ICPE), 2022.
Dependencies
~4.5–7.5MB
~143K SLoC