|
|
IRIX 6.5 » Books » Developer »
SpeedShop User's Guide
(document number: 007-3311-011 / published: 2003-08-23)
table of contents | additional info | download find in page | jump to first hit | clear highlight
- basic block
A set of instructions with a single entry point, a single
exit point, and no branches into or out of the set.
- bead
A record in an experiment.
- caliper points
A caliper point is a point at which you wish to mark your
program so that later you may display performance taken between the marks
(caliper points) you have set. A caliper point may be set at a particular
location in the source, after a particular time interval, or when a particular
signal is received by your program. An implicit caliper point is always present
at the start of execution of the process. A final caliper point is set when
the process calls _exit. Caliper points are numbered so
you can select them with displaying performance data.
- call stack
A software stack of functions and routines that represent
the state of the program at any time. The functions and routines are listed
in the reverse order, from top to bottom, in which they were called. If function
a is immediately below function b in the stack,
then a was called by b. The function
at the bottom of the stack is the one currently executing.
- context switch
The act of saving the state of one process and replacing it
with that of another when both processes time-share a single processor.
- counts
The number of times an event takes place during data gathering.
For example, a count may be kept of the number of times a function executes.
- CPU time
Process virtual time plus time spent when the system is running
on behalf of the process, performing such tasks as executing a system call.
This is the time returned in pcsamp and usertime
experiments. It can be specified in an experiment by using the
ut,30000,2 marching orders.
- dynamic shared object (DSO)
An object file that is similar in structure to an executable
program, but it has no main program.
- exclusive time
The execution time of a given function but not of any functions
called by that function. See inclusive time.
- graduated instruction
As a performance enhancement, when an R10000 system comes
to a point in the execution of a program at which either of two paths might
be taken, it begins to execute both paths until it knows for sure which path
is correct. Graduated instructions are those on the path it will eventually
follow. Issued instructions are those on the path it does not follow.
- inclusive time
The execution time both of a given function and of any functions
called by that function. See exclusive time.
- overflow interval
As used by the hardware counter experiments, it is the number
at which a hardware counter exceeds a preset value. See the speedshop
man page, dsc_hwc experiment.
- PC
Program counter. A register that contains the address of the
instruction that is currently executing.
- process virtual time
Time spent when a program is actually running. This does not
include either 1) the time spent when the program is swapped out and waiting
for a CPU or 2) the time when the operating system is in control, such as
executing a system call for the program. The marching orders ut,30000,1
return process virtual time.
- rld
The runtime linker. This is invoked when a dynamic executable
is run. It maps in shared objects used by the executable, resolves relocations
as ld does at static link time, and allocates common, if
required.
- statistical data
Sampling. The results from this method of data gathering vary
from run to run.
- system time
The time the operating system spends performing services for
a program, such as executing system calls and I/O.
- TLB
Translation lookaside buffer. This is hardware used by the
CPU to quickly translate a virtual address (such as the name of a variable)
to a physical memory address.
- TDT model
Target Description Table model. A CPU model used to calculate
ideal time.
- user time
The same as CPU time.
- wall-clock time
Total time a program takes to execute, including the time
it takes waiting for a CPU. This is real time, not computer time. The marching
orders ut,30000,0 return wall-clock time.
SpeedShop User's Guide
(document number: 007-3311-011 / published: 2003-08-23)
table of contents | additional info | download
Front Matter
New Features in this Guide
About This Guide
Chapter 1. Introduction to Performance Analysis
Chapter 2. Tutorial for C Users
Chapter 3. Tutorial for Fortran Users
Chapter 4. Experiment Types
Chapter 5. Collecting Data on Machine Resource Usage
Chapter 6. Setting Up and Running Experiments: ssrun
Chapter 7. Analyzing Experiment Results
Chapter 8. Miscellaneous Commands
Glossary
Index
home/search |
what's new |
help
|
|
|