You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

237 lines
9.6 KiB

/******************************************************************************/
/* */
/* Copyright (c) International Business Machines Corp., 2007 */
/* */
/* This program is free software; you can redistribute it and/or modify */
/* it under the terms of the GNU General Public License as published by */
/* the Free Software Foundation; either version 2 of the License, or */
/* (at your option) any later version. */
/* */
/* This program is distributed in the hope that it will be useful, */
/* but WITHOUT ANY WARRANTY; without even the implied warranty of */
/* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See */
/* the GNU General Public License for more details. */
/* */
/* You should have received a copy of the GNU General Public License */
/* along with this program; if not, write to the Free Software */
/* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */
/* */
/******************************************************************************/
OO_DESCRIPTION.txt
==================
The contents of the LTP/testcases/realtime/ directory:
doc - Documentation
include - Header files which are to be included in the tests
lib - c library files for the real-time tests
scripts - Set up and parsing scripts needed to run the tests
logs - All test logs are stored in this directory
config - autotools configure support
confugure.ac - Top level autotools configure script
COPYING - GPL license
GNUmakefile.am - Top level automake makefile
README - How to build and run the tests
Below is the decription of what each testcase tests.
func/async_handler testcases :
=============================
async_handler.c:
- Measures latencies involved in asynchronous event handlers. Specifically it
measures the latency of the pthread_cond_signal call until signalled thread
is scheduled.
async_handler_jk.c:
- Mimics an async event handler in a real-time JVM. An async event server
thread is created that goes to sleep waiting to be woken up to do some
work. A user thread is created that simulates the firing of an event by
signalling the async handler thread to do some work.
async_handler_tsc.c:
- Is similar to the above two i.e. it measures latencies involved in
asynchronous event handlers. This test measures time in terms of CPU
timestamp clock(TSC), for pthread_cond_signal latency.
func/gtod_latency testcases :
=============================
gtod_infinite.c:
- Designed to run forever. It must manually be killed so it is
not suited to be part of functional validation suite of tests. It is
designed to look for 'delays' between two calls to clock_gettime()
gtod-latency.c:
- Simple program to measure the time between several pairs of calls to
gettimeofday(). It provides the additional capability to produce graphical
output as a histogram or a scatter graph.
func/matrix_mult testcases :
============================
matrix_mult.c:
- Compares running sequential matrix multiplication routines to running them
in parallel in order to judge multiprocessor performance.
Test runs for 100 iterations and calculates the average time.
func/measurement testcases :
============================
preempt_timing.c:
- Measures the preemption delays that may be encountered by realtime apps.
The program runs with the scheduling policy of SCHED_FIFO at a maximum
SCHED_FIFO priority. It is bound to a single processor and its address space
is locked as well. It makes successive calls to the gettimeofday() function
(via inlined assembly to read the TSC). The value returned between two such
consecutive calls is reported as the latency. The maximum, minimum and
average delays are reported for x pairs of such calls.
rdtsc-latency.c:
- Measures the time between several pairs of calls to rdtsc(); rdtsc counts
the timestamp clock
func/periodic_cpu_load testcases :
==================================
periodic_cpu_load.c:
- Measures variation in computational execution time at various periods and
priorities. This provides the timing information at different CPU loads.
periodic_cpu_load_single.c:
- Measures variation in computational execution time at specified period
priority and loop.
func/pi-tests testcases :
=========================
testpi-0.c:
- Tests whether the priority inheritance feature is present in kernel
testpi-1.c:
- Priority inheritance under two different scenarios. It checks whether the
presence of priority inheritance allows higher priority threads to make more
progress than in absence of the same.
testpi-2.c:
- Introduces a noise thread in above test and checks if the high priority
thread preempts low prio thread multiple times
testpi-4.c:
- The scheduling policies of threads are different from previous testcase
testpi-5.c:
- Uses priority inheritance protocol (PTHREAD_PRIO_INHERIT) and uses
test-skeleton. Test creates a child thread which tries to acquire lock
twice.
testpi-6.c:
- Uses robust mutex lock (PTHREAD_MUTEX_ROBUST) and uses test-skeleton for
other things.
testpi-7.c:
- Measures latencies involved in priority boosting through priority inheritance
test-skeleton.c:
- is a skeleton test which creates two processes through fork(). Also it
handles signals by noting down timing information.
sbrk_mutex.c:
- Uses NUM_THREADS to walk through an array of malloc'd pthread mutexes.
Each thread holds up to NUM_CONCURRENT locks at a time.
func/prio-preempt testcases :
=============================
prio-preempt.c:
- Tests priority preemption. Main thread creates multiple number of threads
with different priorities, all fight for holding mutexes. Threads sleep and
wake-up with condvars. Testcase exhibit scheduling of threads in accordance
with priority-preemption.
func/prio-wake testcases :
==========================
prio-wake.c:
- Tests priority ordered wakeup with pthread_cond_*. It creates number of
worker threads with increasing FIFO priorities. By default, the number of
worker threads is equal to number of cpus. The time when worker thread
starts running is noted. Each of the worker thread then waits on same
_condvar_. The time it wakes up is also noted. Once all the threads finish
execution, the start and wakeup times of all the threads are displayed. The
output must indicate that the thread wakeup happened in a priority order.
func/pthread_kill_latency testcases :
====================================
pthread_kill_latency.c:
- Measures the latency involved in sending a signal to a thread using
pthread_kill. Two threads are created - the one that receives the signal
(thread1) and other that sends signal (thread2). Before sending the signal,
the thread2 waits for thread1 to initialize, notes the time and sends
pthread_kill signal to thread1. Thread2, which has defined a handler for the
signal, notes the time it receives the signal. The maximum and minimum
latency is reported.
func/sched_football testcases :
===============================
sched_football.c:
- A scheduler test that uses a football analogy. The premise is that we want
to make sure that lower priority threads (the offensive team) do not preempt
higher priority threads (the defensive team).
func/sched_jitter testcases :
=============================
sched_jitter.c:
- Measures scheduling jitter between realtime processes.
func/sched_latency testcases :
==============================
sched_latency.c:
- Measures the latency involved with periodic scheduling. A thread is created
at priority 89. It periodically sleeps for a specified duration (PERIOD).
The delay is measured as delay = (now - start - i*PERIOD) converted to
microseconds where now = CLOCK_MONOTONIC gettime in ns, start =
CLOCK_MONOTONIC gettime at the start of the test, i = iteration number,
PERIOD = the period chosen.
func/thread_clock testcases :
=============================
tc-2.c:
- Check if clock_gettime is working properly. This test creates NUMSLEEP
threads that just sleep and NUMWORK threads that spend time on the CPU. It
then reads the thread cpu clocks of all these threads and compares the sum
of thread cpu clocks with the process that spend time on the CPU. It then
reads the cpu clock of all these threads and compares the sum of thread cpu
clocks with the process cpu clock value. The test expects that: the cpu
clock of every sleeping thread shows close to zero value. Sum of cpu clocks
of all threads is comparable with the process cpu clock.
perf/latency testcases :
========================
pthread_cond_latency.c:
- Measures latencies involved in pthread_cond_t
pthread_cond_many.c:
- Measures latencies involved in pthread_cond_t. This test executes in
many processes running together and contesting to being scheduled.
stress/pi-tests testcases :
===========================
testpi-3.c:
- Tests priority inheritance under stress conditions. Test runs in two
scenarios: non-pi and pi.
lookup_pi_state.c:
- Tests lookup_pi_state under stress conditions. Test creates one master and
several slave threads, all fighting for mutexes.