* [RFC] ramfs: rember last accessed chunk
@ 2012-05-15 9:15 Jan Weitzel
2012-05-15 19:29 ` Sascha Hauer
0 siblings, 1 reply; 5+ messages in thread
From: Jan Weitzel @ 2012-05-15 9:15 UTC (permalink / raw)
To: barebox
Writing big files takes longer and longer because of the chunk list
By storing a pointer of the recent used chunk in the inode, access times are
improved.
Testet on with tftp 10M:
OMAP4 chunk size 4096: 12244ms 8192: 4239ms
patched 2647ms 2785ms
i.MX35 chunk size 8192: 7225ms
patched 2691ms
No impact on much smaller files seen
Signed-off-by: Jan Weitzel <j.weitzel@phytec.de>
---
fs/ramfs.c | 46 ++++++++++++++++++++++++++++++++++++----------
1 files changed, 36 insertions(+), 10 deletions(-)
diff --git a/fs/ramfs.c b/fs/ramfs.c
index 83ab6df..5c7410b 100644
--- a/fs/ramfs.c
+++ b/fs/ramfs.c
@@ -48,6 +48,10 @@ struct ramfs_inode {
ulong size;
struct ramfs_chunk *data;
+
+ /* Points to recently used chunk */
+ int recent_chunk;
+ struct ramfs_chunk *recent_chunkp;
};
struct ramfs_priv {
@@ -297,6 +301,34 @@ static int ramfs_close(struct device_d *dev, FILE *f)
return 0;
}
+static struct ramfs_chunk *ramfs_find_chunk(struct ramfs_inode *node, int chunk)
+{
+ struct ramfs_chunk *data;
+ int left = chunk;
+
+ if (chunk == 0)
+ return node->data;
+
+ if (node->recent_chunk == chunk)
+ return node->recent_chunkp;
+
+ if (node->recent_chunk < chunk && node->recent_chunk != 0) {
+ /* Start at last known chunk */
+ data = node->recent_chunkp;
+ left -= node->recent_chunk;
+ } else
+ /* Start at first chunk */
+ data = node->data;
+
+ while (left--)
+ data = data->next;
+
+ node->recent_chunkp = data;
+ node->recent_chunk = chunk;
+
+ return data;
+}
+
static int ramfs_read(struct device_d *_dev, FILE *f, void *buf, size_t insize)
{
struct ramfs_inode *node = (struct ramfs_inode *)f->inode;
@@ -311,11 +343,7 @@ static int ramfs_read(struct device_d *_dev, FILE *f, void *buf, size_t insize)
debug("%s: reading from chunk %d\n", __FUNCTION__, chunk);
/* Position ourself in stream */
- data = node->data;
- while (chunk) {
- data = data->next;
- chunk--;
- }
+ data = ramfs_find_chunk(node, chunk);
ofs = f->pos % CHUNK_SIZE;
/* Read till end of current chunk */
@@ -364,11 +392,7 @@ static int ramfs_write(struct device_d *_dev, FILE *f, const void *buf, size_t i
debug("%s: writing to chunk %d\n", __FUNCTION__, chunk);
/* Position ourself in stream */
- data = node->data;
- while (chunk) {
- data = data->next;
- chunk--;
- }
+ data = ramfs_find_chunk(node, chunk);
ofs = f->pos % CHUNK_SIZE;
/* Write till end of current chunk */
@@ -429,6 +453,8 @@ static int ramfs_truncate(struct device_d *dev, FILE *f, ulong size)
ramfs_put_chunk(data);
data = tmp;
}
+ if (node->recent_chunk > newchunks)
+ node->recent_chunk = 0;
}
if (newchunks > oldchunks) {
--
1.7.0.4
_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: [RFC] ramfs: rember last accessed chunk
2012-05-15 9:15 [RFC] ramfs: rember last accessed chunk Jan Weitzel
@ 2012-05-15 19:29 ` Sascha Hauer
2012-05-16 6:10 ` [PATCH] " Jan Weitzel
0 siblings, 1 reply; 5+ messages in thread
From: Sascha Hauer @ 2012-05-15 19:29 UTC (permalink / raw)
To: Jan Weitzel; +Cc: barebox
Hi Jan,
On Tue, May 15, 2012 at 11:15:46AM +0200, Jan Weitzel wrote:
> Writing big files takes longer and longer because of the chunk list
> By storing a pointer of the recent used chunk in the inode, access times are
> improved.
> Testet on with tftp 10M:
> OMAP4 chunk size 4096: 12244ms 8192: 4239ms
> patched 2647ms 2785ms
> i.MX35 chunk size 8192: 7225ms
> patched 2691ms
>
The numbers look good and the code looks sane. We can give it a try. Two
nitpicks below.
> No impact on much smaller files seen
>
> Signed-off-by: Jan Weitzel <j.weitzel@phytec.de>
> ---
> fs/ramfs.c | 46 ++++++++++++++++++++++++++++++++++++----------
> 1 files changed, 36 insertions(+), 10 deletions(-)
>
> diff --git a/fs/ramfs.c b/fs/ramfs.c
> index 83ab6df..5c7410b 100644
> --- a/fs/ramfs.c
> +++ b/fs/ramfs.c
> @@ -48,6 +48,10 @@ struct ramfs_inode {
>
> ulong size;
> struct ramfs_chunk *data;
> +
Trailing whitespace here.
> + /* Points to recently used chunk */
> + int recent_chunk;
> + struct ramfs_chunk *recent_chunkp;
> };
>
> struct ramfs_priv {
> @@ -297,6 +301,34 @@ static int ramfs_close(struct device_d *dev, FILE *f)
> return 0;
> }
>
> +static struct ramfs_chunk *ramfs_find_chunk(struct ramfs_inode *node, int chunk)
> +{
> + struct ramfs_chunk *data;
> + int left = chunk;
> +
> + if (chunk == 0)
> + return node->data;
> +
> + if (node->recent_chunk == chunk)
> + return node->recent_chunkp;
> +
> + if (node->recent_chunk < chunk && node->recent_chunk != 0) {
> + /* Start at last known chunk */
> + data = node->recent_chunkp;
> + left -= node->recent_chunk;
> + } else
> + /* Start at first chunk */
> + data = node->data;
if you have brackets in the if path you should add them in the else path
aswell.
> }
> + if (node->recent_chunk > newchunks)
Also trailing whitespace
Sascha
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox
^ permalink raw reply [flat|nested] 5+ messages in thread* [PATCH] ramfs: rember last accessed chunk
2012-05-15 19:29 ` Sascha Hauer
@ 2012-05-16 6:10 ` Jan Weitzel
2012-05-16 11:20 ` Sascha Hauer
0 siblings, 1 reply; 5+ messages in thread
From: Jan Weitzel @ 2012-05-16 6:10 UTC (permalink / raw)
To: barebox
Writing big files takes longer and longer because of the chunk list
By storing a pointer of the recent used chunk in the inode, access times are
improved.
Testet on with tftp 10M:
OMAP4 chunk size 4096: 12244ms 8192: 4239ms
patched 2647ms 2785ms
i.MX35 chunk size 8192: 7225ms
patched 2691ms
No impact on much smaller files seen
Signed-off-by: Jan Weitzel <j.weitzel@phytec.de>
---
v2: I will use checkpatch
fs/ramfs.c | 47 +++++++++++++++++++++++++++++++++++++----------
1 files changed, 37 insertions(+), 10 deletions(-)
diff --git a/fs/ramfs.c b/fs/ramfs.c
index 83ab6df..cec5e76 100644
--- a/fs/ramfs.c
+++ b/fs/ramfs.c
@@ -48,6 +48,10 @@ struct ramfs_inode {
ulong size;
struct ramfs_chunk *data;
+
+ /* Points to recently used chunk */
+ int recent_chunk;
+ struct ramfs_chunk *recent_chunkp;
};
struct ramfs_priv {
@@ -297,6 +301,35 @@ static int ramfs_close(struct device_d *dev, FILE *f)
return 0;
}
+static struct ramfs_chunk *ramfs_find_chunk(struct ramfs_inode *node, int chunk)
+{
+ struct ramfs_chunk *data;
+ int left = chunk;
+
+ if (chunk == 0)
+ return node->data;
+
+ if (node->recent_chunk == chunk)
+ return node->recent_chunkp;
+
+ if (node->recent_chunk < chunk && node->recent_chunk != 0) {
+ /* Start at last known chunk */
+ data = node->recent_chunkp;
+ left -= node->recent_chunk;
+ } else {
+ /* Start at first chunk */
+ data = node->data;
+ }
+
+ while (left--)
+ data = data->next;
+
+ node->recent_chunkp = data;
+ node->recent_chunk = chunk;
+
+ return data;
+}
+
static int ramfs_read(struct device_d *_dev, FILE *f, void *buf, size_t insize)
{
struct ramfs_inode *node = (struct ramfs_inode *)f->inode;
@@ -311,11 +344,7 @@ static int ramfs_read(struct device_d *_dev, FILE *f, void *buf, size_t insize)
debug("%s: reading from chunk %d\n", __FUNCTION__, chunk);
/* Position ourself in stream */
- data = node->data;
- while (chunk) {
- data = data->next;
- chunk--;
- }
+ data = ramfs_find_chunk(node, chunk);
ofs = f->pos % CHUNK_SIZE;
/* Read till end of current chunk */
@@ -364,11 +393,7 @@ static int ramfs_write(struct device_d *_dev, FILE *f, const void *buf, size_t i
debug("%s: writing to chunk %d\n", __FUNCTION__, chunk);
/* Position ourself in stream */
- data = node->data;
- while (chunk) {
- data = data->next;
- chunk--;
- }
+ data = ramfs_find_chunk(node, chunk);
ofs = f->pos % CHUNK_SIZE;
/* Write till end of current chunk */
@@ -429,6 +454,8 @@ static int ramfs_truncate(struct device_d *dev, FILE *f, ulong size)
ramfs_put_chunk(data);
data = tmp;
}
+ if (node->recent_chunk > newchunks)
+ node->recent_chunk = 0;
}
if (newchunks > oldchunks) {
--
1.7.0.4
_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: [PATCH] ramfs: rember last accessed chunk
2012-05-16 6:10 ` [PATCH] " Jan Weitzel
@ 2012-05-16 11:20 ` Sascha Hauer
2012-05-16 18:42 ` Uwe Kleine-König
0 siblings, 1 reply; 5+ messages in thread
From: Sascha Hauer @ 2012-05-16 11:20 UTC (permalink / raw)
To: Jan Weitzel; +Cc: barebox
On Wed, May 16, 2012 at 08:10:16AM +0200, Jan Weitzel wrote:
> Writing big files takes longer and longer because of the chunk list
> By storing a pointer of the recent used chunk in the inode, access times are
> improved.
> Testet on with tftp 10M:
> OMAP4 chunk size 4096: 12244ms 8192: 4239ms
> patched 2647ms 2785ms
> i.MX35 chunk size 8192: 7225ms
> patched 2691ms
>
> No impact on much smaller files seen
>
> Signed-off-by: Jan Weitzel <j.weitzel@phytec.de>
Applied, thanks
Sascha
> ---
> v2: I will use checkpatch
>
> fs/ramfs.c | 47 +++++++++++++++++++++++++++++++++++++----------
> 1 files changed, 37 insertions(+), 10 deletions(-)
>
> diff --git a/fs/ramfs.c b/fs/ramfs.c
> index 83ab6df..cec5e76 100644
> --- a/fs/ramfs.c
> +++ b/fs/ramfs.c
> @@ -48,6 +48,10 @@ struct ramfs_inode {
>
> ulong size;
> struct ramfs_chunk *data;
> +
> + /* Points to recently used chunk */
> + int recent_chunk;
> + struct ramfs_chunk *recent_chunkp;
> };
>
> struct ramfs_priv {
> @@ -297,6 +301,35 @@ static int ramfs_close(struct device_d *dev, FILE *f)
> return 0;
> }
>
> +static struct ramfs_chunk *ramfs_find_chunk(struct ramfs_inode *node, int chunk)
> +{
> + struct ramfs_chunk *data;
> + int left = chunk;
> +
> + if (chunk == 0)
> + return node->data;
> +
> + if (node->recent_chunk == chunk)
> + return node->recent_chunkp;
> +
> + if (node->recent_chunk < chunk && node->recent_chunk != 0) {
> + /* Start at last known chunk */
> + data = node->recent_chunkp;
> + left -= node->recent_chunk;
> + } else {
> + /* Start at first chunk */
> + data = node->data;
> + }
> +
> + while (left--)
> + data = data->next;
> +
> + node->recent_chunkp = data;
> + node->recent_chunk = chunk;
> +
> + return data;
> +}
> +
> static int ramfs_read(struct device_d *_dev, FILE *f, void *buf, size_t insize)
> {
> struct ramfs_inode *node = (struct ramfs_inode *)f->inode;
> @@ -311,11 +344,7 @@ static int ramfs_read(struct device_d *_dev, FILE *f, void *buf, size_t insize)
> debug("%s: reading from chunk %d\n", __FUNCTION__, chunk);
>
> /* Position ourself in stream */
> - data = node->data;
> - while (chunk) {
> - data = data->next;
> - chunk--;
> - }
> + data = ramfs_find_chunk(node, chunk);
> ofs = f->pos % CHUNK_SIZE;
>
> /* Read till end of current chunk */
> @@ -364,11 +393,7 @@ static int ramfs_write(struct device_d *_dev, FILE *f, const void *buf, size_t i
> debug("%s: writing to chunk %d\n", __FUNCTION__, chunk);
>
> /* Position ourself in stream */
> - data = node->data;
> - while (chunk) {
> - data = data->next;
> - chunk--;
> - }
> + data = ramfs_find_chunk(node, chunk);
> ofs = f->pos % CHUNK_SIZE;
>
> /* Write till end of current chunk */
> @@ -429,6 +454,8 @@ static int ramfs_truncate(struct device_d *dev, FILE *f, ulong size)
> ramfs_put_chunk(data);
> data = tmp;
> }
> + if (node->recent_chunk > newchunks)
> + node->recent_chunk = 0;
> }
>
> if (newchunks > oldchunks) {
> --
> 1.7.0.4
>
>
> _______________________________________________
> barebox mailing list
> barebox@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/barebox
>
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: [PATCH] ramfs: rember last accessed chunk
2012-05-16 11:20 ` Sascha Hauer
@ 2012-05-16 18:42 ` Uwe Kleine-König
0 siblings, 0 replies; 5+ messages in thread
From: Uwe Kleine-König @ 2012-05-16 18:42 UTC (permalink / raw)
To: Sascha Hauer; +Cc: barebox
On Wed, May 16, 2012 at 01:20:25PM +0200, Sascha Hauer wrote:
> On Wed, May 16, 2012 at 08:10:16AM +0200, Jan Weitzel wrote:
> > Writing big files takes longer and longer because of the chunk list
> > By storing a pointer of the recent used chunk in the inode, access times are
> > improved.
> > Testet on with tftp 10M:
> > OMAP4 chunk size 4096: 12244ms 8192: 4239ms
> > patched 2647ms 2785ms
> > i.MX35 chunk size 8192: 7225ms
> > patched 2691ms
> >
> > No impact on much smaller files seen
> >
> > Signed-off-by: Jan Weitzel <j.weitzel@phytec.de>
>
> Applied, thanks
Don't know if it's already too late:
$Subject ~= s/rember/remember/
Best regards
Uwe
--
Pengutronix e.K. | Uwe Kleine-König |
Industrial Linux Solutions | http://www.pengutronix.de/ |
_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2012-05-16 18:42 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2012-05-15 9:15 [RFC] ramfs: rember last accessed chunk Jan Weitzel
2012-05-15 19:29 ` Sascha Hauer
2012-05-16 6:10 ` [PATCH] " Jan Weitzel
2012-05-16 11:20 ` Sascha Hauer
2012-05-16 18:42 ` Uwe Kleine-König
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox