Grant: Engage Grants Program
University of Toronto
$20,350.00
Auto-Tuning GPU Performance using Machine Learning
FY 2012-13
Summary
There has been a phenomenal growth in software applications for mobile devices such as smart phones and
tablets. This growth has increased the pressure on mobile processor manufacturers such as Intel, AMD and
Qualcomm to make more powerful processors that consume less power. Today, these manufacturers are
making processors with multiple cores (CPUs) combined with a general-purpose graphics-processing unit
(GPU) on the same chip. The use of a GPU is attractive because, for some applications, the GPU can deliver a
ten-fold increase in speed at a fraction of the power, compared to a CPU. However, this performance and
power-saving potential can only be achieved if programs are re-structured and optimized to effectively utilize
the GPU-a process that is tedious and error prone.
Qualcomm is a world leader in next generation mobile technologies. They produce processors that integrate
several cores with a GPU and are used in mobile phones, tablets and other wireless devices. We will
collaborate with Qualcomm scientists and engineers located in Toronto to explore the use of machine learning
to automatically deal with the complexities of applying GPU code optimizations and auto-tuning performance.
Our long-term goal is to build a machine learning framework around Support Vector Machines (SVMs) for
performance auto-tuning on GPUs. Our goal for this Engage grant application is to perform an initial
exploration of the use of SVMs in this regard for visual media processing applications.
We will work with Qualcomm's scientists and engineers to identify a small set of relevant training kernels,
to characterize the properties of these kernels, to port and optimize these kernels on the GPU, to assess the
suitability of our hiCUDA compiler infrastructure to express GPU optimizations in a parameterized fashion
and to explore training and classification of a simple SVM using our small set kernels. The outcome of this
initial research will inform both Qualcomm and us on the feasibility of machine learning in GPU performance
auto-tuning, potentially leading to longer-term collaboration.
Program
Engage Grants Program
Recipient
Location
Ontario, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data
As published
Application id
432940-2012
Name nom
Abdelrahman, Tarek
Department département
Electrical & Computer Engineering
Organization id
31
Institution établissement
University of Toronto
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2012
Competition year année de concours
2012
Award amount
20350
Program id
EGP
Program name en
Engage Grants Program
Program name fr
Programme de subventions d'engagement partenariat
Group en
Research partnerships programs
Group fr
Programmes de partenariat
Committee code
1553
Committee name en
Ontario Internal Decision Committee
Committee name fr
Comité de décision interne pour l'Ontario
Area of application code
801
Area of application group en
Information and communication services
Area of application group fr
Services d'information et de télécommunications
Area of application en
Computer software
Area of application fr
Logiciel
Research subject code
2704
Research subject group en
Information technology
Research subject group fr
Technologie de l'information
Research subject en
Distributed and parallel processing
Research subject fr
Traitement réparti et simultané
Partie
1
Nb partie
1
Application title
Auto-Tuning GPU Performance using Machine Learning
Keyword
NSERC
Application summary
There has been a phenomenal growth in software applications for mobile devices such as smart phones and
tablets. This growth has increased the pressure on mobile processor manufacturers such as Intel, AMD and
Qualcomm to make more powerful processors that consume less power. Today, these manufacturers are
making processors with multiple cores (CPUs) combined with a general-purpose graphics-processing unit
(GPU) on the same chip. The use of a GPU is attractive because, for some applications, the GPU can deliver a
ten-fold increase in speed at a fraction of the power, compared to a CPU. However, this performance and
power-saving potential can only be achieved if programs are re-structured and optimized to effectively utilize
the GPU-a process that is tedious and error prone.
Qualcomm is a world leader in next generation mobile technologies. They produce processors that integrate
several cores with a GPU and are used in mobile phones, tablets and other wireless devices. We will
collaborate with Qualcomm scientists and engineers located in Toronto to explore the use of machine learning
to automatically deal with the complexities of applying GPU code optimizations and auto-tuning performance.
Our long-term goal is to build a machine learning framework around Support Vector Machines (SVMs) for
performance auto-tuning on GPUs. Our goal for this Engage grant application is to perform an initial
exploration of the use of SVMs in this regard for visual media processing applications.
We will work with Qualcomm's scientists and engineers to identify a small set of relevant training kernels,
to characterize the properties of these kernels, to port and optimize these kernels on the GPU, to assess the
suitability of our hiCUDA compiler infrastructure to express GPU optimizations in a parameterized fashion
and to explore training and classification of a simple SVM using our small set kernels. The outcome of this
initial research will inform both Qualcomm and us on the feasibility of machine learning in GPU performance
auto-tuning, potentially leading to longer-term collaboration.