Created
July 4, 2019 06:02
-
-
Save s417-lama/92842ef5a3bc3fb23d7353ae949c9c00 to your computer and use it in GitHub Desktop.
experiments on memory consistency (x86)
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
// compile: g++ dekker.cpp -lpthread | |
#include <stdio.h> | |
#include <thread> | |
#include <unistd.h> | |
#include <pthread.h> | |
#include <sys/syscall.h> | |
#define N 100000 | |
#define USE_MEMFENCE 1 | |
#define CACHE_LINE_SIZE 64 | |
struct Data { | |
char pad0[CACHE_LINE_SIZE]; | |
volatile int A = 0; | |
char pad1[CACHE_LINE_SIZE]; | |
volatile int B = 0; | |
char pad2[CACHE_LINE_SIZE]; | |
}; | |
volatile int ready = 0; | |
pthread_barrier_t ready_barrier, finish_barrier; | |
volatile int A_result, B_result; | |
Data data; | |
pid_t gettid(void) { | |
return syscall(SYS_gettid); | |
} | |
int set_affinity(int cpu) { | |
pid_t pid = gettid(); | |
cpu_set_t cpuset; | |
CPU_ZERO(&cpuset); | |
CPU_SET(cpu, &cpuset); | |
int result = sched_setaffinity(pid, sizeof(cpu_set_t), &cpuset); | |
if (result != 0 || sched_getcpu() != cpu) { | |
fprintf(stderr, "failed to set affinity.\n"); | |
exit(1); | |
} | |
} | |
int main() { | |
int result[2][2] = {{0, 0}, {0, 0}}; | |
pthread_barrier_init(&ready_barrier , NULL, 3); | |
pthread_barrier_init(&finish_barrier, NULL, 3); | |
set_affinity(0); | |
auto thread1 = std::thread([] { | |
set_affinity(1); | |
for (int i = 0; i < N; i++) { | |
pthread_barrier_wait(&ready_barrier); | |
while (!ready); | |
data.A = 1; | |
#if USE_MEMFENCE | |
asm volatile("mfence":::"memory"); | |
#endif | |
B_result = data.B; | |
pthread_barrier_wait(&finish_barrier); | |
} | |
}); | |
auto thread2 = std::thread([] { | |
set_affinity(2); | |
for (int i = 0; i < N; i++) { | |
pthread_barrier_wait(&ready_barrier); | |
while (!ready); | |
data.B = 1; | |
#if USE_MEMFENCE | |
asm volatile("mfence":::"memory"); | |
#endif | |
A_result = data.A; | |
pthread_barrier_wait(&finish_barrier); | |
} | |
}); | |
for (int i = 0; i < N; i++) { | |
pthread_barrier_wait(&ready_barrier); | |
ready = 1; | |
pthread_barrier_wait(&finish_barrier); | |
result[A_result][B_result]++; | |
data.A = 0; data.B = 0; | |
ready = 0; | |
} | |
thread1.join(); | |
thread2.join(); | |
for (int a = 0; a < 2; a++) | |
for (int b = 0; b < 2; b++) | |
printf("(A == %d, B == %d) : %d\n", a, b, result[a][b]); | |
return 0; | |
} |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment