/* ChibiOS/RT - Copyright (C) 2006-2007 Giovanni Di Sirio. This file is part of ChibiOS/RT. ChibiOS/RT 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 3 of the License, or (at your option) any later version. ChibiOS/RT 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, see . */ #include #include "test.h" /** * @page test_benchmarks Kernel Benchmarks * *

Description

* This module implements a series of system benchmarks. The benchmarks are * useful as a stress test and as a reference when comparing ChibiOS/RT * with similar systems. * *

Objective

* Objective of the test module is to provide a performance index for the * most critical system subsystems. The performance numbers allow to * discover performance regressions between successive ChibiOS/RT releases. * *

Preconditions

* None. * *

Test Cases

* - @subpage test_benchmarks_001 * - @subpage test_benchmarks_002 * - @subpage test_benchmarks_003 * - @subpage test_benchmarks_004 * - @subpage test_benchmarks_005 * - @subpage test_benchmarks_006 * - @subpage test_benchmarks_007 * - @subpage test_benchmarks_008 * - @subpage test_benchmarks_009 * - @subpage test_benchmarks_010 * - @subpage test_benchmarks_011 * - @subpage test_benchmarks_012 * - @subpage test_benchmarks_013 * . * @file testbmk.c Kernel Benchmarks * @brief Kernel Benchmarks source file * @file testbmk.h * @brief Kernel Benchmarks header file */ static Semaphore sem1; #if CH_USE_MUTEXES static Mutex mtx1; #endif static msg_t thread1(void *p) { msg_t msg; (void)p; do { chMsgRelease(msg = chMsgWait()); } while (msg); return 0; } __attribute__((noinline)) static unsigned int msg_loop_test(Thread *tp) { uint32_t n = 0; test_wait_tick(); test_start_timer(1000); do { (void)chMsgSend(tp, 1); n++; #if defined(WIN32) ChkIntSources(); #endif } while (!test_timer_done); (void)chMsgSend(tp, 0); return n; } /** * @page test_benchmarks_001 Messages performance #1 * *

Description

* A message server thread is created with a lower priority than the client * thread, the messages throughput per second is measured and the result * printed in the output log. */ static char *bmk1_gettest(void) { return "Benchmark, messages #1"; } static void bmk1_execute(void) { uint32_t n; threads[0] = chThdCreateStatic(wa[0], WA_SIZE, chThdGetPriority()-1, thread1, NULL); n = msg_loop_test(threads[0]); chThdTerminate(threads[0]); test_wait_threads(); test_print("--- Score : "); test_printn(n); test_print(" msgs/S, "); test_printn(n << 1); test_println(" ctxswc/S"); } const struct testcase testbmk1 = { bmk1_gettest, NULL, NULL, bmk1_execute }; /** * @page test_benchmarks_002 Messages performance #2 * *

Description

* A message server thread is created with an higher priority than the client * thread, the messages throughput per second is measured and the result * printed in the output log. */ static char *bmk2_gettest(void) { return "Benchmark, messages #2"; } static void bmk2_execute(void) { uint32_t n; threads[0] = chThdCreateStatic(wa[0], WA_SIZE, chThdGetPriority()+1, thread1, NULL); n = msg_loop_test(threads[0]); chThdTerminate(threads[0]); test_wait_threads(); test_print("--- Score : "); test_printn(n); test_print(" msgs/S, "); test_printn(n << 1); test_println(" ctxswc/S"); } const struct testcase testbmk2 = { bmk2_gettest, NULL, NULL, bmk2_execute }; static msg_t thread2(void *p) { return (msg_t)p; } /** * @page test_benchmarks_003 Messages performance #3 * *

Description

* A message server thread is created with an higher priority than the client * thread, four lower priority threads crowd the ready list, the messages * throughput per second is measured while the ready list and the result * printed in the output log. */ static char *bmk3_gettest(void) { return "Benchmark, messages #3"; } static void bmk3_execute(void) { uint32_t n; threads[0] = chThdCreateStatic(wa[0], WA_SIZE, chThdGetPriority()+1, thread1, NULL); threads[1] = chThdCreateStatic(wa[1], WA_SIZE, chThdGetPriority()-2, thread2, NULL); threads[2] = chThdCreateStatic(wa[2], WA_SIZE, chThdGetPriority()-3, thread2, NULL); threads[3] = chThdCreateStatic(wa[3], WA_SIZE, chThdGetPriority()-4, thread2, NULL); threads[4] = chThdCreateStatic(wa[4], WA_SIZE, chThdGetPriority()-5, thread2, NULL); n = msg_loop_test(threads[0]); chThdTerminate(threads[0]); test_wait_threads(); test_print("--- Score : "); test_printn(n); test_print(" msgs/S, "); test_printn(n << 1); test_println(" ctxswc/S"); } const struct testcase testbmk3 = { bmk3_gettest, NULL, NULL, bmk3_execute }; /** * @page test_benchmarks_004 Context Switch performance * *

Description

* A thread is created that just performs a @p chSchGoSleepS() into a loop, * the thread is awakened as fast is possible by the tester thread.
* The Context Switch performance is calculated by measuring the number of * iterations after a second of continuous operations. */ static char *bmk4_gettest(void) { return "Benchmark, context switch"; } msg_t thread4(void *p) { msg_t msg; Thread *self = chThdSelf(); (void)p; chSysLock(); do { chSchGoSleepS(PRSUSPENDED); msg = self->p_rdymsg; } while (msg == RDY_OK); chSysUnlock(); return 0; } static void bmk4_execute(void) { Thread *tp; tp = threads[0] = chThdCreateStatic(wa[0], WA_SIZE, chThdGetPriority()+1, thread4, NULL); uint32_t n = 0; test_wait_tick(); test_start_timer(1000); do { chSysLock(); chSchWakeupS(tp, RDY_OK); chSchWakeupS(tp, RDY_OK); chSchWakeupS(tp, RDY_OK); chSchWakeupS(tp, RDY_OK); chSysUnlock(); n += 4; #if defined(WIN32) ChkIntSources(); #endif } while (!test_timer_done); chSysLock(); chSchWakeupS(tp, RDY_TIMEOUT); chSysUnlock(); test_wait_threads(); test_print("--- Score : "); test_printn(n * 2); test_println(" ctxswc/S"); } const struct testcase testbmk4 = { bmk4_gettest, NULL, NULL, bmk4_execute }; /** * @page test_benchmarks_005 Threads performance, full cycle * *

Description

* Threads are continuously created and terminated into a loop. A full * @p chThdCreateStatic() / @p chThdExit() / @p chThdWait() cycle is performed * in each iteration.
* The performance is calculated by measuring the number of iterations after * a second of continuous operations. */ static char *bmk5_gettest(void) { return "Benchmark, threads, full cycle"; } static void bmk5_execute(void) { uint32_t n = 0; void *wap = wa[0]; tprio_t prio = chThdGetPriority() - 1; test_wait_tick(); test_start_timer(1000); do { chThdWait(chThdCreateStatic(wap, WA_SIZE, prio, thread2, NULL)); n++; #if defined(WIN32) ChkIntSources(); #endif } while (!test_timer_done); test_print("--- Score : "); test_printn(n); test_println(" threads/S"); } const struct testcase testbmk5 = { bmk5_gettest, NULL, NULL, bmk5_execute }; /** * @page test_benchmarks_006 Threads performance, create/exit only * *

Description

* Threads are continuously created and terminated into a loop. A partial * @p chThdCreateStatic() / @p chThdExit() cycle is performed in each * iteration, the @p chThdWait() is not necessary because the thread is * created at an higher priority so there is no need to wait for it to * terminate.
* The performance is calculated by measuring the number of iterations after * a second of continuous operations. */ static char *bmk6_gettest(void) { return "Benchmark, threads, create only"; } static void bmk6_execute(void) { uint32_t n = 0; void *wap = wa[0]; tprio_t prio = chThdGetPriority() + 1; test_wait_tick(); test_start_timer(1000); do { chThdCreateStatic(wap, WA_SIZE, prio, thread2, NULL); n++; #if defined(WIN32) ChkIntSources(); #endif } while (!test_timer_done); test_print("--- Score : "); test_printn(n); test_println(" threads/S"); } const struct testcase testbmk6 = { bmk6_gettest, NULL, NULL, bmk6_execute }; /** * @page test_benchmarks_007 Mass reschedulation performance * *

Description

* Five threads are created and atomically reschedulated by resetting the * semaphore where they are waiting on. The operation is performed into a * continuous loop.
* The performance is calculated by measuring the number of iterations after * a second of continuous operations. */ static msg_t thread3(void *p) { (void)p; while (!chThdShouldTerminate()) chSemWait(&sem1); return 0; } static char *bmk7_gettest(void) { return "Benchmark, mass reschedulation, 5 threads"; } static void bmk7_setup(void) { chSemInit(&sem1, 0); } static void bmk7_execute(void) { uint32_t n; threads[0] = chThdCreateStatic(wa[0], WA_SIZE, chThdGetPriority()+5, thread3, NULL); threads[1] = chThdCreateStatic(wa[1], WA_SIZE, chThdGetPriority()+4, thread3, NULL); threads[2] = chThdCreateStatic(wa[2], WA_SIZE, chThdGetPriority()+3, thread3, NULL); threads[3] = chThdCreateStatic(wa[3], WA_SIZE, chThdGetPriority()+2, thread3, NULL); threads[4] = chThdCreateStatic(wa[4], WA_SIZE, chThdGetPriority()+1, thread3, NULL); n = 0; test_wait_tick(); test_start_timer(1000); do { chSemReset(&sem1, 0); n++; #if defined(WIN32) ChkIntSources(); #endif } while (!test_timer_done); test_terminate_threads(); chSemReset(&sem1, 0); test_wait_threads(); test_print("--- Score : "); test_printn(n); test_print(" reschedulations/S, "); test_printn(n * 6); test_println(" ctxswc/S"); } const struct testcase testbmk7 = { bmk7_gettest, bmk7_setup, NULL, bmk7_execute }; /** * @page test_benchmarks_008 I/O Round-Robin voluntary reschedulation. * *

Description

* Five threads are created at equal priority, each thread just increases a * variable and yields.
* The performance is calculated by measuring the number of iterations after * a second of continuous operations. */ static msg_t thread8(void *p) { do { chThdYield(); chThdYield(); chThdYield(); chThdYield(); (*(uint32_t *)p) += 4; #if defined(WIN32) ChkIntSources(); #endif } while(!chThdShouldTerminate()); return 0; } static char *bmk8_gettest(void) { return "Benchmark, round robin context switching"; } static void bmk8_execute(void) { uint32_t n; n = 0; test_wait_tick(); threads[0] = chThdCreateStatic(wa[0], WA_SIZE, chThdGetPriority()-1, thread8, (void *)&n); threads[1] = chThdCreateStatic(wa[1], WA_SIZE, chThdGetPriority()-1, thread8, (void *)&n); threads[2] = chThdCreateStatic(wa[2], WA_SIZE, chThdGetPriority()-1, thread8, (void *)&n); threads[3] = chThdCreateStatic(wa[3], WA_SIZE, chThdGetPriority()-1, thread8, (void *)&n); threads[4] = chThdCreateStatic(wa[4], WA_SIZE, chThdGetPriority()-1, thread8, (void *)&n); chThdSleepSeconds(1); test_terminate_threads(); test_wait_threads(); test_print("--- Score : "); test_printn(n); test_print(" reschedulations/S, "); test_printn(n); test_println(" ctxswc/S"); } const struct testcase testbmk8 = { bmk8_gettest, NULL, NULL, bmk8_execute }; /** * @page test_benchmarks_009 I/O Queues throughput * *

Description

* Four bytes are written and then read from an @p InputQueue into a continuous * loop.
* The performance is calculated by measuring the number of iterations after * a second of continuous operations. */ static char *bmk9_gettest(void) { return "Benchmark, I/O Queues throughput"; } static void bmk9_execute(void) { static uint8_t ib[16]; static InputQueue iq; chIQInit(&iq, ib, sizeof(ib), NULL); uint32_t n = 0; test_wait_tick(); test_start_timer(1000); do { chIQPutI(&iq, 0); chIQPutI(&iq, 1); chIQPutI(&iq, 2); chIQPutI(&iq, 3); (void)chIQGet(&iq); (void)chIQGet(&iq); (void)chIQGet(&iq); (void)chIQGet(&iq); n++; #if defined(WIN32) ChkIntSources(); #endif } while (!test_timer_done); test_print("--- Score : "); test_printn(n * 4); test_println(" bytes/S"); } const struct testcase testbmk9 = { bmk9_gettest, NULL, NULL, bmk9_execute }; /** * @page test_benchmarks_010 Virtual Timers set/reset performance * *

Description

* A virtual timer is set and immediately reset into a continuous loop.
* The performance is calculated by measuring the number of iterations after * a second of continuous operations. */ static char *bmk10_gettest(void) { return "Benchmark, virtual timers set/reset"; } static void tmo(void *param) {(void)param;} static void bmk10_execute(void) { static VirtualTimer vt1, vt2; uint32_t n = 0; test_wait_tick(); test_start_timer(1000); do { chSysLock(); chVTSetI(&vt1, 1, tmo, NULL); chVTSetI(&vt2, 10000, tmo, NULL); chVTResetI(&vt1); chVTResetI(&vt2); chSysUnlock(); n++; #if defined(WIN32) ChkIntSources(); #endif } while (!test_timer_done); test_print("--- Score : "); test_printn(n * 2); test_println(" timers/S"); } const struct testcase testbmk10 = { bmk10_gettest, NULL, NULL, bmk10_execute }; /** * @page test_benchmarks_011 Semaphores wait/signal performance * *

Description

* A counting semaphore is taken/released into a continuous loop, no Context * Switch happens because the counter is always non negative.
* The performance is calculated by measuring the number of iterations after * a second of continuous operations. */ static char *bmk11_gettest(void) { return "Benchmark, semaphores wait/signal"; } static void bmk11_setup(void) { chSemInit(&sem1, 1); } static void bmk11_execute(void) { uint32_t n = 0; test_wait_tick(); test_start_timer(1000); do { chSemWait(&sem1); chSemSignal(&sem1); chSemWait(&sem1); chSemSignal(&sem1); chSemWait(&sem1); chSemSignal(&sem1); chSemWait(&sem1); chSemSignal(&sem1); n++; #if defined(WIN32) ChkIntSources(); #endif } while (!test_timer_done); test_print("--- Score : "); test_printn(n * 4); test_println(" wait+signal/S"); } const struct testcase testbmk11 = { bmk11_gettest, bmk11_setup, NULL, bmk11_execute }; #if CH_USE_MUTEXES /** * @page test_benchmarks_012 Mutexes lock/unlock performance * *

Description

* A mutex is locked/unlocked into a continuous loop, no Context Switch happens * because there are no other threads asking for the mutex.
* The performance is calculated by measuring the number of iterations after * a second of continuous operations. */ static char *bmk12_gettest(void) { return "Benchmark, mutexes lock/unlock"; } static void bmk12_setup(void) { chMtxInit(&mtx1); } static void bmk12_execute(void) { uint32_t n = 0; test_wait_tick(); test_start_timer(1000); do { chMtxLock(&mtx1); chMtxUnlock(); chMtxLock(&mtx1); chMtxUnlock(); chMtxLock(&mtx1); chMtxUnlock(); chMtxLock(&mtx1); chMtxUnlock(); n++; #if defined(WIN32) ChkIntSources(); #endif } while (!test_timer_done); test_print("--- Score : "); test_printn(n * 4); test_println(" lock+unlock/S"); } const struct testcase testbmk12 = { bmk12_gettest, bmk12_setup, NULL, bmk12_execute }; #endif /** * @page test_benchmarks_013 RAM Footprint * *

Description

* The memory size of the various kernel objects is printed. */ static char *bmk13_gettest(void) { return "Benchmark, RAM footprint"; } static void bmk13_execute(void) { test_print("--- System: "); test_printn(sizeof(ReadyList) + sizeof(VTList) + IDLE_THREAD_STACK_SIZE + (sizeof(Thread) + sizeof(struct intctx) + sizeof(struct extctx) + INT_REQUIRED_STACK) * 2); test_println(" bytes"); test_print("--- Thread: "); test_printn(sizeof(Thread)); test_println(" bytes"); test_print("--- Timer : "); test_printn(sizeof(VirtualTimer)); test_println(" bytes"); test_print("--- Semaph: "); test_printn(sizeof(Semaphore)); test_println(" bytes"); #if CH_USE_EVENTS test_print("--- EventS: "); test_printn(sizeof(EventSource)); test_println(" bytes"); test_print("--- EventL: "); test_printn(sizeof(EventListener)); test_println(" bytes"); #endif #if CH_USE_MUTEXES test_print("--- Mutex : "); test_printn(sizeof(Mutex)); test_println(" bytes"); #endif #if CH_USE_CONDVARS test_print("--- CondV.: "); test_printn(sizeof(CondVar)); test_println(" bytes"); #endif #if CH_USE_QUEUES test_print("--- Queue : "); test_printn(sizeof(GenericQueue)); test_println(" bytes"); #endif #if CH_USE_MAILBOXES test_print("--- MailB.: "); test_printn(sizeof(Mailbox)); test_println(" bytes"); #endif } const struct testcase testbmk13 = { bmk13_gettest, NULL, NULL, bmk13_execute }; /* * Test sequence for benchmarks pattern. */ const struct testcase * const patternbmk[] = { #if !TEST_NO_BENCHMARKS &testbmk1, &testbmk2, &testbmk3, &testbmk4, &testbmk5, &testbmk6, &testbmk7, &testbmk8, &testbmk9, &testbmk10, &testbmk11, #if CH_USE_MUTEXES &testbmk12, #endif &testbmk13, #endif NULL };