ABOUT MIAMI:

MIAMI (Machine Independent Application Models for performance Insight) is a
collection of tools for automatic construction of application centric,
thread-level, performance models, focused on facilitating performance
prediction and micro-architecture level performance diagnosis. At its lowest
level, MIAMI uses static and dynamic analysis of fully optimized x86-64 
application binaries.


MOTIVATION:

Performance diagnosis is by and large a manual process. While performance
measurement is well supported by existing tools, analysis of performance
results to understand performance inefficiencies and to identify
opportunities for tuning is left to the users. The difficulty of this task
is compounded by the fact that most performance tools measure performance
effects, the result of interactions between code and a target architecture.
Such measurements are useful to identify hotspots in applications, places
where execution time is spent or where cache misses are incurred. However,
a process of deconvolution through which we can attribute parts of the
observed effects to specific application and architectural factors is needed
to perform root cause analysis from hardware counter measurements. While
certain correlations between application or architectural factors and the
observed performance effects can be established, this process is based on
human intuition, and many times results in a trial and error tuning approach.

In contrast, MIAMI uses a bottom-up approach. Instead of measuring
performance effects, MIAMI collects profiles of application characteristics
that are largely architecture independent. MIAMI uses detailed static
analysis and performance modeling based on first order principles to reason
about the performance of a code on a target architecture, to understand
performance inefficiencies, and to identify opportunities for tuning. MIAMI 
computes a large number of performance metrics that correlate well with 
several types of common performance inefficiencies. Experienced users that
can interpret this data would get the most benefit from the provided
insight. While manual interpretation of performance results is still needed 
for now, our experience shows that MIAMI drastically reduces the amount of
guesswork traditionally involved in uncovering tuning opportunities.


INSTALLATION:

For building instructions, see the INSTALL file in this folder.


USE INSTRUCTIONS:

For information on using the tools, see the README file and associated
documentation in the doc/ subfolder.


CONTACT INFORMATION:

For other questions, contact Gabriel Marin <mgabi99 .at. gmail>.