Performance Analysis Measurements; Introduction; State Label Overview Measurement - HP 1630A Operating And Programming Manual

Model 1630A/D/G Performance Analysis Measurements
INTRODUCTION
Chapter 5
Performance Analysis Measurements
As the software begins to push the limits of the hardware of a system, the need for more efficient code becomes
much more apparent . The software must be able to respond to all of its inputs in a timely fashion, with little or
no delays . Performance Analysis allows the programmer to analyze code-to find bottlenecks and
inefficiencies . The Overview Measurements of the 1630 series logic analyzers (available in the state-only
configuration) provide the programmer with tools with which to perform this analysis . These tools allow the
programmer to analyze module processor usage, intermodule linkage, module duration, and time-based
interaction .
The output of these tools appear as bar graphs (called histograms) relating the relative actions between
measured items . These histograms are presented in the chart menu after the trace specification has been set
up for an overview measurement. With the exception of the Time Positional measurement, these
measurements execute in a continuous random sample mode . A sample consists of 1024 qualified states . The
time between states varies randomly to insure random sampling . To stop a measurement immediately, press
the STOP key twice . The first press will force the analyzer to take one more sample and then quit . Two
consecutive STOPs will force the analyzer to abort the current sample .
The following chapter describes the overview measurements, and provides example usages of these
measurements .
The 1630A/D analyzers can show two types of histograms : state label histograms and time interval histograms .
The 1630G analyzer can show two types of state label histograms (counting states or program activity), time
interval histograms, time position histograms and linkage histograms .
STATE LABEL OVERVIEW MEASUREMENT
State Label Overview Measurement allows the programmer to define module code ranges and then compute
the amount of time spent within these modules, relative to each other, or relative to the entire system software .
The ranges are defined for one of the eight possible labels defined in the format menu . (For a typical inverse
assembler application, these would be ADDR, DATA, and STAT .) The ranges may be defined as any locations
between the maximum and minimum values possible for that label .
State Label Measurement is very beneficial in improving overall system performance . It allows the
programmer to define sections of code critical to the overall system or module performance, and thereby avoid
wasting time attempting to optimize code that does not noticeably affect overall performance .
For example, a module that performs floating point processing is found to be noticeably slow . The floating
point module consists of seven routines : F_ADD, F_SUB, F_MUL, F_DIV, F_CMP, F_MOVE, and F_NORM .
These routines provide floating point addition, subtraction, multiplication, division, comparison, moves, and
normalization . To improve the module performance, the individual routines are defined in the State Label
Trace menu, and then analyzed, as in the following steps .