From: "Al Boldi" <a1426z@gawab.com>
To: "'Jens Axboe'" <axboe@suse.de>,
"'Steven Pratt'" <slpratt@austin.ibm.com>
Cc: "'Andrew Morton'" <akpm@osdl.org>, <torvalds@osdl.org>,
<grant_lkml@dodo.com.au>, <linux@rainbow-software.org>,
<andre@tomt.net>, <linux-kernel@vger.kernel.org>
Subject: RE: [git patches] IDE update
Date: Sun, 10 Jul 2005 00:39:17 +0300 [thread overview]
Message-ID: <200507092139.AAA03256@raad.intranet> (raw)
In-Reply-To: <20050709054053.GX7050@suse.de>
[-- Attachment #1: Type: text/plain, Size: 6142 bytes --]
Jens Axboe <axboe@suse.de> wrote:{
> >>
> >>>Some more investigation - it appears to be broken read-ahead, actually.
> >>>
> >>>--- mm/readahead.c~ 2005-07-08 11:16:14.000000000 +0200
> >>>+++ mm/readahead.c 2005-07-08 11:17:49.000000000 +0200
> >>>@@ -351,7 +351,9 @@
> >>> ra->cache_hit += nr_to_read;
> >>> if (ra->cache_hit >= VM_MAX_CACHE_HIT) {
> >>> ra_off(ra);
> >>>+#if 0
> >>> ra->flags |= RA_FLAG_INCACHE;
> >>>+#endif
> >>> return 0;
> >>> }
> >>> } else {
> >>
> >> >Just use the test app I posted, it shows the problem just fine. If I
> >use a regular file, behaviour is identical as expected (ie equally
> >broken :-).
}
Modified test app shows comparison w/ and w/o O_DIRECT.
Your comments please!
2.4.31# (hdparm -f /dev/hda;free;time /tmp/readisk /dev/hda 1 )
/dev/hda:
total used free shared buffers cached
Mem: 256388 23728 232660 0 1944 4324
-/+ buffers/cache: 17460 238928
Swap: 0 0 0
Mem Throughput: 107 MiB/sec - 8388608-16 bs-blks
Mem Throughput: 121 MiB/sec - 4194304-32 bs-blks
Mem Throughput: 122 MiB/sec - 2097152-64 bs-blks
Mem Throughput: 122 MiB/sec - 1048576-128 bs-blks
Mem Throughput: 120 MiB/sec - 524288-256 bs-blks
Mem Throughput: 120 MiB/sec - 262144-512 bs-blks
Mem Throughput: 131 MiB/sec - 131072-1024 bs-blks
Mem Throughput: 336 MiB/sec - 65536-2048 bs-blks
Mem Throughput: 476 MiB/sec - 32768-4096 bs-blks
Mem Throughput: 432 MiB/sec - 16384-8192 bs-blks
Mem Throughput: 824 MiB/sec - 8192-16384 bs-blks <--- 20% faster
Mem Throughput: 799 MiB/sec - 4096-32768 bs-blks
Mem Throughput: 494 MiB/sec - 2048-65536 bs-blks
Mem Throughput: 270 MiB/sec - 1024-131072 bs-blks
Mem Throughput: 146 MiB/sec - 512-262144 bs-blks
Mem Throughput: 76 MiB/sec - 256-524288 bs-blks
Disk Throughput: 39 MiB/sec
1.13user 12.35system 0:15.65elapsed 86%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (87major+2061minor)pagefaults 0swaps
=====================================================================
2.4.31# (hdparm -f /dev/hda;free;time /tmp/readisk /dev/hda 1 1 )
O_DIRECT
/dev/hda:
total used free shared buffers cached
Mem: 256388 23824 232564 0 2024 4324
-/+ buffers/cache: 17476 238912
Swap: 0 0 0
Mem Throughput: 36 MiB/sec - 8388608-16 bs-blks
Mem Throughput: 34 MiB/sec - 4194304-32 bs-blks
Mem Throughput: 34 MiB/sec - 2097152-64 bs-blks
Mem Throughput: 37 MiB/sec - 1048576-128 bs-blks
Mem Throughput: 29 MiB/sec - 524288-256 bs-blks <--- small dip
Mem Throughput: 15 MiB/sec - 262144-512 bs-blks <--- 333% slower
Mem Throughput: 42 MiB/sec - 131072-1024 bs-blks
Mem Throughput: 45 MiB/sec - 65536-2048 bs-blks
Mem Throughput: 40 MiB/sec - 32768-4096 bs-blks
Mem Throughput: 36 MiB/sec - 16384-8192 bs-blks
Mem Throughput: 33 MiB/sec - 8192-16384 bs-blks
Mem Throughput: 25 MiB/sec - 4096-32768 bs-blks
Mem Throughput: 16 MiB/sec - 2048-65536 bs-blks
Mem Throughput: 10 MiB/sec - 1024-131072 bs-blks
read infile: Invalid argument 512-262144 bs-blks <--- too small?
read infile: Invalid argument 256-524288 bs-blks
Disk Throughput: 38 MiB/sec
0.27user 10.88system 1:12.95elapsed 15%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (99major+2062minor)pagefaults 0swaps
=====================================================================
2.6.12# (hdparm -f /dev/hda;free;time /tmp/readisk /dev/hda 1 )
/dev/hda:
total used free shared buffers cached
Mem: 255228 15668 239560 0 5060 4252
-/+ buffers/cache: 6356 248872
Swap: 0 0 0
Mem Throughput: 97 MiB/sec - 8388608-16 bs-blks
Mem Throughput: 120 MiB/sec - 4194304-32 bs-blks
Mem Throughput: 120 MiB/sec - 2097152-64 bs-blks
Mem Throughput: 120 MiB/sec - 1048576-128 bs-blks
Mem Throughput: 121 MiB/sec - 524288-256 bs-blks
Mem Throughput: 120 MiB/sec - 262144-512 bs-blks
Mem Throughput: 129 MiB/sec - 131072-1024 bs-blks
Mem Throughput: 237 MiB/sec - 65536-2048 bs-blks
Mem Throughput: 451 MiB/sec - 32768-4096 bs-blks
Mem Throughput: 418 MiB/sec - 16384-8192 bs-blks
Mem Throughput: 636 MiB/sec - 8192-16384 bs-blks
Mem Throughput: 672 MiB/sec - 4096-32768 bs-blks <--- 20% slower
Mem Throughput: 399 MiB/sec - 2048-65536 bs-blks
Mem Throughput: 217 MiB/sec - 1024-131072 bs-blks
Mem Throughput: 115 MiB/sec - 512-262144 bs-blks
Mem Throughput: 59 MiB/sec - 256-524288 bs-blks
Disk Throughput: 26 MiB/sec
0.93user 13.79system 0:18.59elapsed 79%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (0major+2150minor)pagefaults 0swaps
=====================================================================
2.6.12# (hdparm -f /dev/hda;free;time /tmp/readisk /dev/hda 1 1)
O_DIRECT
/dev/hda:
total used free shared buffers cached
Mem: 255228 15668 239560 0 5060 4252
-/+ buffers/cache: 6356 248872
Swap: 0 0 0
Mem Throughput: 33 MiB/sec - 8388608-16 bs-blks
Mem Throughput: 30 MiB/sec - 4194304-32 bs-blks
Mem Throughput: 34 MiB/sec - 2097152-64 bs-blks
Mem Throughput: 39 MiB/sec - 1048576-128 bs-blks
Mem Throughput: 30 MiB/sec - 524288-256 bs-blks <--- small dip
Mem Throughput: 49 MiB/sec - 262144-512 bs-blks <--- 333% faster
Mem Throughput: 49 MiB/sec - 131072-1024 bs-blks
Mem Throughput: 44 MiB/sec - 65536-2048 bs-blks
Mem Throughput: 42 MiB/sec - 32768-4096 bs-blks
Mem Throughput: 40 MiB/sec - 16384-8192 bs-blks
Mem Throughput: 29 MiB/sec - 8192-16384 bs-blks
Mem Throughput: 22 MiB/sec - 4096-32768 bs-blks
Mem Throughput: 14 MiB/sec - 2048-65536 bs-blks
Mem Throughput: 8 MiB/sec - 1024-131072 bs-blks
Mem Throughput: 4 MiB/sec - 512-262144 bs-blks
read infile: Invalid argument 256-524288 bs-blks <--- too small?
Disk Throughput: 34 MiB/sec
0.69user 32.09system 1:39.15elapsed 33%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (0major+2164minor)pagefaults 0swaps
=====================================================================
[-- Attachment #2: rd128.c --]
[-- Type: application/octet-stream, Size: 1702 bytes --]
//#include <stdio.h>
#include <unistd.h>
#define __USE_GNU
#include <fcntl.h>
#include <stdlib.h>
#include <sys/time.h>
#include <sys/ioctl.h>
#include <linux/fs.h>
#define BS (8192*1024)
#define BLOCKS (128*1024/8192)
#define ALIGN(buf) (char *) (((unsigned long) (buf) + 4095) & ~(4095))
void print_time(struct timeval *s, int memory)
{
unsigned long ms, mb;
struct timeval e;
mb = BS * BLOCKS / 1024;
gettimeofday(&e, NULL);
ms = (e.tv_sec - s->tv_sec) * 1000 + (e.tv_usec - s->tv_usec) / 1000;
if (memory)
printf("Mem Throughput: %lu MiB/sec - %lu-%lu bs-blks\n", mb / ms, BS/memory, BLOCKS*memory);
else
printf("Disk Throughput: %lu MiB/sec\n", mb / ms);
}
void read_stuff(int fd, char *buffer, int memory)
{
struct timeval s;
int i, ret;
int j=1; if (memory)\r j=memory;
gettimeofday(&s, NULL);
for (i = 0; i < (BLOCKS*j); i++) {
if (memory)
lseek(fd, 0, SEEK_SET);
ret = read(fd, buffer, BS/j);
if (!ret)
break;
else if (ret < 0) {
perror("read infile");
// break;
return;
}
}
print_time(&s, memory);
}
int main(int argc, char *argv[])
{
char *buffer;
int i, fd, seek;
if (argc < 2) {
printf("%s: <device>\n", argv[0]);
return 1;
}
if (argc > 2)
seek = 1;
else
seek = 0;
if (argc > 3)
fd = open(argv[1], O_RDONLY | O_DIRECT);
else
fd = open(argv[1], O_RDONLY );
if (argc > 4)
seek = 0;
if (fd == -1) {
perror("open");
return 2;
}
ioctl(fd, BLKFLSBUF, 0);
buffer = ALIGN(malloc(BS + 4095));
if (seek) {
for (i = 0; i < 16 ; i++) {
read_stuff(fd, buffer, (2<<i)/2);
}}
read_stuff(fd, buffer, 0);
return 0;
}
next prev parent reply other threads:[~2005-07-09 21:41 UTC|newest]
Thread overview: 74+ messages / expand[flat|nested] mbox.gz Atom feed top
2005-07-03 16:52 [git patches] IDE update Bartlomiej Zolnierkiewicz
2005-07-04 12:01 ` Al Boldi
2005-07-04 12:30 ` Bartlomiej Zolnierkiewicz
2005-07-04 15:30 ` Al Boldi
2005-07-04 15:41 ` Bartlomiej Zolnierkiewicz
2005-07-04 17:06 ` Al Boldi
2005-07-04 17:38 ` Ondrej Zary
2005-07-04 19:51 ` Bartlomiej Zolnierkiewicz
2005-07-04 20:32 ` Al Boldi
2005-07-04 20:47 ` Bartlomiej Zolnierkiewicz
2005-07-04 23:25 ` André Tomt
2005-07-05 3:43 ` IOWAIT block layer problem Al Boldi
2005-07-05 10:01 ` [git patches] IDE update Ondrej Zary
2005-07-05 10:14 ` Jens Axboe
2005-07-05 10:19 ` Ondrej Zary
2005-07-05 10:42 ` Jens Axboe
2005-07-05 12:35 ` Ondrej Zary
2005-07-05 12:51 ` Jens Axboe
2005-07-05 13:02 ` Ondrej Zary
2005-07-05 13:11 ` Jens Axboe
2005-07-05 15:51 ` Ondrej Zary
2005-07-05 14:21 ` Jens Axboe
2005-07-05 15:00 ` Ondrej Zary
2005-07-05 19:18 ` Jens Axboe
2005-07-05 19:25 ` Jens Axboe
2005-07-05 21:36 ` Ondrej Zary
2005-07-05 17:27 ` Linus Torvalds
2005-07-05 19:14 ` Jens Axboe
2005-07-05 21:39 ` Ondrej Zary
2005-07-11 14:21 ` Alan Cox
2005-07-06 0:35 ` Grant Coady
2005-07-06 0:51 ` Linus Torvalds
2005-07-06 3:26 ` Al Boldi
2005-07-06 4:56 ` Grant Coady
2005-07-06 5:22 ` Linus Torvalds
2005-07-08 8:48 ` Jens Axboe
2005-07-08 9:24 ` Jens Axboe
2005-07-08 9:34 ` Andrew Morton
2005-07-08 9:53 ` Jens Axboe
2005-07-08 22:25 ` Steven Pratt
2005-07-09 5:40 ` Jens Axboe
2005-07-09 21:39 ` Al Boldi [this message]
2005-07-08 10:20 ` Ingo Molnar
2005-07-08 11:45 ` Jens Axboe
2005-07-07 22:32 ` Mark Lord
2005-07-08 0:06 ` Grant Coady
2005-07-08 11:37 ` Erik Slagter
2005-07-06 20:56 ` Bill Davidsen
2005-07-07 13:47 ` Ondrej Zary
2005-07-07 13:48 ` Bartlomiej Zolnierkiewicz
2005-07-07 19:34 ` Bill Davidsen
2005-07-05 2:47 ` Jeff Garzik
-- strict thread matches above, loose matches on Subject: below --
2005-08-18 21:37 [git patches] ide update Bartlomiej Zolnierkiewicz
2005-08-18 22:15 ` Linus Torvalds
2005-08-18 22:19 ` Nish Aravamudan
2005-08-19 0:44 ` Mark Lord
2005-08-18 23:08 ` Alan Cox
2005-08-19 9:02 ` Bartlomiej Zolnierkiewicz
2005-08-19 18:06 ` Alan Cox
2005-08-19 23:51 ` Bartlomiej Zolnierkiewicz
2005-08-19 23:52 ` Bartlomiej Zolnierkiewicz
2005-11-10 1:00 Bartlomiej Zolnierkiewicz
2005-11-18 23:21 Bartlomiej Zolnierkiewicz
2005-11-19 23:46 Bartlomiej Zolnierkiewicz
2005-12-15 2:03 Bartlomiej Zolnierkiewicz
2007-05-09 22:46 [git patches] IDE update Bartlomiej Zolnierkiewicz
2007-05-09 22:46 ` Jeff Garzik
2007-05-09 23:20 ` David Miller
2007-05-09 23:23 ` Bartlomiej Zolnierkiewicz
2007-05-09 23:18 ` Jeff Garzik
2007-05-09 22:47 ` Jeff Garzik
2007-05-09 22:59 ` Andrew Morton
2007-05-09 23:15 ` Jeff Garzik
2007-07-09 21:46 Bartlomiej Zolnierkiewicz
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=200507092139.AAA03256@raad.intranet \
--to=a1426z@gawab.com \
--cc=akpm@osdl.org \
--cc=andre@tomt.net \
--cc=axboe@suse.de \
--cc=grant_lkml@dodo.com.au \
--cc=linux-kernel@vger.kernel.org \
--cc=linux@rainbow-software.org \
--cc=slpratt@austin.ibm.com \
--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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox