All of lore.kernel.org
 help / color / mirror / Atom feed
From: Eric Dumazet <dada1@cosmosbay.com>
To: dipankar@in.ibm.com
Cc: "David S. Miller" <davem@davemloft.net>,
	linux-kernel@vger.kernel.org, torvalds@osdl.org, akpm@osdl.org
Subject: Re: [PATCH] reorder struct files_struct
Date: Thu, 15 Sep 2005 01:19:47 +0200	[thread overview]
Message-ID: <4328B013.1010400@cosmosbay.com> (raw)
In-Reply-To: <20050914225043.GD6237@in.ibm.com>

[-- Attachment #1: Type: text/plain, Size: 1619 bytes --]

Dipankar Sarma a écrit :
> On Thu, Sep 15, 2005 at 12:42:03AM +0200, Eric Dumazet wrote:
> 
>>Dipankar Sarma a écrit :

> Not just embedded fdtable, but also the embedded fdsets. I would expect
> count, fdt, fdtab and the fdsets to fit into one cache line in 
> some archs.
> 
> 
> 
>>But I wonder if 'next_fd' really has to be in 'struct fdtable', maybe it 
>>could be moved to 'struct files_struct' close to file_lock ?
> 
> 
> next_fd has to be in struct fdtable. It needs to be consistent
> with whichever fdtable a lock-free reader sees.
> 

I may be wrong, but I think a reader never look at next_fd.

next_fd is only used (read/written) by a 'writer', with file_lock hold,  in 
locate_fd(), copy_fdtable() and get_unused_fd(), __put_unused_fd()


> 
>>If yes, the whole embedded struct fdtable is readonly.
> 
> 
> But not close_on_exec_init or open_fds_init. We would update them
> on open/close.

Yes, sure, but those fields are not part of the embedded struct fdtable

> 
> Some benchmarking would be useful here.

This simple bench can be used, I got good results on a dual opteron machine. 
As Opterons have prety good NUMA links (Hypertransport), I suspect older 
hardware should obtain even better results.

$ gcc -O2 -o bench bench.c -lpthread
$ ./bench -t 2 -l 10   # run the bench with 2 threads for 10 seconds.
2 threads, 10 seconds, work_done=1721627

To run it with a small fdset (no more than 3 + (nbthreads) files opened)
$ ./bench -s -t 2 -l 10
2 threads, 10 seconds, work_done=1709716

Unfortunatly I cannot boot the dual opterons at the moment to try to move 
next_fd close to file_lock

Eric


[-- Attachment #2: bench.c --]
[-- Type: text/plain, Size: 1870 bytes --]

/*
 * Bench program to exercice multi threads using open()/close()/read()/lseek() calls.
 * Usage :
 *    bench [-t XX] [-l len]
 *   XX : number of threads
 *   len : bench time in seconds
 * -s : small fdset : try to use embedded sruct fdtable
 */
#include <pthread.h>
#include <stdlib.h>
#include <stdio.h>
#include <fcntl.h>
#include <unistd.h>

pthread_mutex_t  mut = PTHREAD_MUTEX_INITIALIZER;

char *file = "/etc/passwd";
int sflag; /* small fdset */
int end_prog;
unsigned long work_done;

void *perform_work(void *arg)
{
	int fd, i;
	unsigned long units = 0;
	char c;

	/* force this program to open more than 64 fds */
	if (!sflag)
		for (i = 0 ; i < 64 ; i++)
			open("/dev/null", O_RDONLY);

	while (!end_prog) {
		fd = open(file, O_RDONLY);
		read(fd, &c, 1);
		lseek(fd, 10, SEEK_SET);
		read(fd, &c, 1);
		lseek(fd, 20, SEEK_SET);
		read(fd, &c, 1);
		lseek(fd, 30, SEEK_SET);
		read(fd, &c, 1);
		lseek(fd, 40, SEEK_SET);
		read(fd, &c, 1);
		close(fd);
		units++;
	}
	pthread_mutex_lock(&mut);
	work_done += units;
	pthread_mutex_unlock(&mut);
	return 0;
}

void usage(int code)
{
	fprintf(stderr, "Usage : bench [-s] [-t threads] [-l duration]\n");
	exit(code);
}

int main(int argc, char *argv[])
{
	int i, c;
	int nbthreads = 2;
	unsigned int length = 10;
	pthread_t *tid;

	while ((c = getopt(argc, argv, "st:l:")) != -1) {
		if (c == 't')
			nbthreads = atoi(optarg);
		else if (c == 'l')
			length = atoi(optarg);
		else if (c == 's')
			sflag = 1;
		else usage(1);
	}


	tid = malloc(nbthreads*sizeof(pthread_t));
	for (i = 0 ; i < nbthreads; i++)
		pthread_create(tid + i, NULL, perform_work, NULL);

	sleep(length);
	end_prog = 1;
	for (i = 0 ; i < nbthreads; i++)
		pthread_join(tid[i], NULL); 

pthread_mutex_lock(&mut);
printf("%d threads, %u seconds, work_done=%lu\n", nbthreads, length, work_done);
pthread_mutex_unlock(&mut);
return 0;
}

  reply	other threads:[~2005-09-14 23:19 UTC|newest]

Thread overview: 19+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2005-09-14 18:31 [PATCH]: Brown paper bag in fs/file.c? David S. Miller
2005-09-14 18:48 ` David S. Miller
2005-09-14 19:05   ` David S. Miller
2005-09-14 19:18 ` Dipankar Sarma
2005-09-14 19:57   ` David S. Miller
2005-09-14 20:15     ` Dipankar Sarma
2005-09-14 20:29       ` David S. Miller
2005-09-14 21:17         ` [PATCH] reorder struct files_struct Eric Dumazet
2005-09-14 21:35           ` Peter Staubach
2005-09-14 22:02           ` Dipankar Sarma
2005-09-14 22:17             ` Andrew Morton
2005-09-14 22:42             ` Eric Dumazet
2005-09-14 22:50               ` Dipankar Sarma
2005-09-14 23:19                 ` Eric Dumazet [this message]
2005-09-15  4:54                   ` Dipankar Sarma
2005-09-15  6:17                     ` Eric Dumazet
2005-09-15  9:35                       ` Dipankar Sarma
2005-09-15 17:11                         ` Eric Dumazet
2005-09-15 21:06       ` [PATCH]: Brown paper bag in fs/file.c? David S. Miller

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=4328B013.1010400@cosmosbay.com \
    --to=dada1@cosmosbay.com \
    --cc=akpm@osdl.org \
    --cc=davem@davemloft.net \
    --cc=dipankar@in.ibm.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=torvalds@osdl.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.