lkml.org 
[lkml]   [2012]   [Aug]   [24]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH] fs/proc: Move kfree outside pde_unload_lock
On Fri, Aug 24, 2012 at 11:45:45AM -0500, Nathan Zimmer wrote:
> On 08/24/2012 09:58 AM, Eric Dumazet wrote:
>> Le vendredi 24 août 2012 à 09:48 -0500, Nathan Zimmer a écrit :
>>> On Wed, Aug 22, 2012 at 11:42:58PM +0200, Eric Dumazet wrote:
>>>> On Wed, 2012-08-22 at 20:28 +0200, Eric Dumazet wrote:
>>>>
>>>>> Thats interesting, but if you really want this to fly, one RCU
>>>>> conversion would be much better ;)
>>>>>
>>>>> pde_users would be an atomic_t and you would avoid the spinlock
>>>>> contention.
>>>> Here is what I had in mind, I would be interested to know how it helps a 512 core machine ;)
>>>>
>>> Here are the results and they look great.
>>>
>>> cpuinfo baseline moved kfree Rcu
>>> tasks read-sec read-sec read-sec
>>> 1 0.0141 0.0141 0.0141
>>> 2 0.0140 0.0140 0.0142
>>> 4 0.0140 0.0141 0.0141
>>> 8 0.0145 0.0145 0.0140
>>> 16 0.0553 0.0548 0.0168
>>> 32 0.1688 0.1622 0.0549
>>> 64 0.5017 0.3856 0.1690
>>> 128 1.7005 0.9710 0.5038
>>> 256 5.2513 2.6519 2.0804
>>> 512 8.0529 6.2976 3.0162
>>>
>>>
>>>
>> Indeed...
>>
>> Could you explicit the test you are actually doing ?
>>
>> Thanks
>>
>>
>
>
> It is a dead simple test.
> The test starts by forking off X number of tasks
> assigning each their own cpu.
> Each task then allocs a bit of memory.
> All tasks wait on a memory cell for the go order.
> We measure the read time starting here.
> Once the go order is given they all read a chunk of the selected proc file.
> I was using /proc/cpuinfo to test.
> Once everyone has finished we take the end read time.
>

Here is the text for those who are curious.


/*------------------------------------------------------------------------------------*/
char *helpstr[] = {
"This test program is a generic template.",
0
};

#include <stdio.h>
#include <sys/types.h>
#include <unistd.h>
#include <stdlib.h>
#include <sys/wait.h>
#include <sys/mman.h>

#include <sched.h>
#include <sys/time.h>
#include <string.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <linux/unistd.h>

//#include "setup.h"

#define MAXCPUS 4096
#define perrorx(s) do { perror(s); exit(1);} while(0)
#define mb() asm volatile("mfence":::"memory")
#define barrier() asm volatile("": : :"memory")
#define cpu_relax() asm volatile ("rep;nop":::"memory");


extern int optind, opterr;
extern char *optarg;

static int verbose = 0;
static int header = 0;
static char *file = "/proc/stat";
static int numtasks = 1;
static int repeat = 1;
static int bufsize = 1024;

struct control_s {
int ready;
int done;
int go;
int exit;
} *cntl;


static cpu_set_t *defmask;
static int cpu_set_size;

static void runon_init(void)
{
if (!defmask) {
cpu_set_size = CPU_ALLOC_SIZE(MAXCPUS);
defmask = CPU_ALLOC(MAXCPUS);
if (sched_getaffinity(0, cpu_set_size, defmask) < 0)
perrorx("unexpected failure in runon_init");
}
}


static double timeInSeconds(long time_in_microseconds)
{
double temp;

temp = time_in_microseconds;
temp /= 1000000;

return temp;
}

static int runon(int cpu)
{
cpu_set_t *mask;

runon_init();
mask = CPU_ALLOC(MAXCPUS);
if (cpu < 0 || cpu >= MAXCPUS)
return -1;
CPU_ZERO_S(cpu_set_size, mask);
CPU_SET_S(cpu, cpu_set_size, mask);
if (sched_setaffinity(0, cpu_set_size, mask) < 0)
return -1;
CPU_FREE(mask);
return 0;
}

static long getCurrentTime()
{
struct timeval tp;
long usec;

mb();
gettimeofday(&tp, 0);
usec = tp.tv_sec * 1000000 + tp.tv_usec;
mb();
return usec;
}


static void do_help(void)
{
char **p;

for (p = helpstr; *p; p++)
printf("%s\n", *p);
exit(0);
}

static void slave(int id)
{
FILE *f;
int i;
char *buf;

runon(id);
buf = malloc(bufsize);
memset(buf, 0, bufsize);

if ((f = fopen(file, "r")) < 0)
perrorx("open failed");
while (fgets(buf, bufsize, f) != NULL) {
}
fclose(f);

(void)__sync_fetch_and_add(&cntl->ready, 1);
while (!cntl->go)
cpu_relax();

for (i = 0; i < repeat; i++) {
if ((f = fopen(file, "r")) < 0)
perrorx("open failed");
while (fgets(buf, bufsize, f) != NULL) {
}
fclose(f);
barrier();
}

(void)__sync_fetch_and_add(&cntl->done, 1);
while (!cntl->exit)
cpu_relax();
exit(0);
}

int main(int argc, char **argv)
{
int i, c, stat, er = 0;
static char optstr[] = "b:f:hn:r:v";
unsigned long t, tfork, tready, tread, texit;

opterr = 1;
while ((c = getopt(argc, argv, optstr)) != EOF)
switch (c) {
case 'b':
bufsize = atoi(optarg);
break;
case 'f':
file = optarg;
break;
case 'h':
header++;
break;
case 'n':
numtasks = atoi(optarg);
break;
case 'r':
repeat = atoi(optarg);
break;
case 'v':
verbose++;
break;
case '?':
er = 1;
break;
}

if (er)
do_help();

runon(0);
cntl = mmap(NULL, getpagesize(), PROT_WRITE | PROT_READ, MAP_SHARED | MAP_ANONYMOUS, -1, 0);

tfork = getCurrentTime();
for (i = 0; i < numtasks; i++)
if (fork() == 0)
slave(i + 1);
t = getCurrentTime();
tfork = t - tfork;

tready = t;
while (cntl->ready != numtasks)
usleep(1000);
t = getCurrentTime();
tready = t - tready;

tread = t;
cntl->go = 1;

while (cntl->done != numtasks)
cpu_relax();
t = getCurrentTime();
tread = t - tread;

texit = t;
cntl->exit = 1;
while (wait(&stat) > 0)
usleep(1000);
texit = getCurrentTime() - texit;

if (header) {
printf("File: %s\n", file);
printf("Bufsize: %d\n", bufsize);
printf("Repeats: %d\n", repeat);
printf("%6s%18s%18s%18s%18s%18s\n", "tasks", "fork-sec", "ready-sec", "read-sec", "read/repeat sec",
"texit");
}
printf("%6d%18.6f%18.6f%18.6f%18.6f%18.6f\n", numtasks, timeInSeconds(tfork), timeInSeconds(tready),
timeInSeconds(tread), timeInSeconds(tread) / repeat, timeInSeconds(texit));
return 0;
}
\
 
 \ /
  Last update: 2012-08-25 00:21    [W:0.083 / U:0.244 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site