Machine learning and Optimization-guided Compilers for Heterogeneous Architectures (MOCHA)

DARPA - Information Innovation Office HR001124S0035

For over three decades, hardware and software developers could rely on improved performance

through increases in microprocessor clock speeds. This scaling broke down in the mid-2000s and

attention shifted to new architectural features, such as multithreading. That’s because today’s

technology, clock speed alone cannot lead to higher performance. Multithreading also has its

limits and modern systems are increasingly using a variety of specialized co-processors and

accelerators that are designed for high performance in specific domains and on specific tasks.

Traditional compilers are not designed to generate efficient machine code for such heterogenous

ensembles of Central Processing Units (CPUs), Graphics Processing Units (GPUs), and other

accelerators. Instead, software developers write unique code and libraries to take advantage of

specialized hardware, reducing productivity and losing much of the potential benefit of these

hardware components. Extending compilers to handle this heterogeneity is currently a manual

task that is performed by compiler experts. Adapting current compilers is time consuming and

error prone, does not address the challenge of taking advantage of hardware accelerators, and

does not improve our ability to upgrade mission-critical systems in a timely manner.

MOCHA seeks to build a new generation of compiler technology that can realize the full

potential performance of a system comprising multiple heterogenous computational elements. It

will accomplish this goal by 1) using data driven methods, machine learning (ML), and advanced

optimization techniques to rapidly adapt compilers to new hardware components with little

human effort and 2) developing new internal representations and programming languages that

enable compilers to determine how to make optimal use of available hardware, rather than

depending on humans to do so. Without this capability, the Department of Defense (DoD) and

the commercial world remain constrained by current compiler technologies and lack the ability

to fully and rapidly capitalize on emerging specialized hardware.

 

 

The MOCHA program will have a single technical area: Compiler Technology. As discussed

above and below, this technical area encompasses several distinct compiler components and

proposals may address one or more individual components or the entire compiler toolchain.

Multiple awards are anticipated.

Please contact the program director

Proposal Abstract Due Date: August 22, 2024, at 1:00 PM ET

Proposal Due Date / BAA Closing Date: September 26, 2024 at 1:00 PM

The BAA Coordinator for this effort may be reached at:

MOCHA@darpa.mil

DARPA/I2O

ATTN:HR001124S0035

675 North Randolph Street

Arlington, VA 22203-2114