linux-ext4.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH] ext4: re-inline ext4_rec_len_(to|from)_disk functions
@ 2010-08-04 15:52 Eric Sandeen
  2010-08-05  0:11 ` Ted Ts'o
  0 siblings, 1 reply; 4+ messages in thread
From: Eric Sandeen @ 2010-08-04 15:52 UTC (permalink / raw)
  To: ext4 development

commit 3d0518f4, "ext4: New rec_len encoding for very
large blocksizes" made several changes to this path, but from
a perf perspective, un-inlining ext4_rec_len_from_disk() seems
most significant.  This function is called from ext4_check_dir_entry(),
which on a file-creation workload is called extremely often.

I tested this with bonnie:

# bonnie++ -u root -s 0 -f -x 200 -d /mnt/test -n 32

(this does 200 iterations) and got this for the file creations:

ext4 stock:   Average =  21206.8 files/s
ext4 inlined: Average =  22346.7 files/s  (+5%)

We may get a little more by optimizing around the extra complexity
for > 64K blocks, but for now simply inlining is a nice improvement,
without a huge change in code size:

   text	   data	    bss	    dec	    hex	filename
 314580	  11888	    176	 326644	  4fbf4	ext4-stock.ko
 315246	  11888	    176	 327310	  4fe8e	ext4-inlined.ko

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
---

diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index 19a4de5..8aba5b8 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -1378,6 +1378,31 @@ struct ext4_dir_entry_2 {
 					 ~EXT4_DIR_ROUND)
 #define EXT4_MAX_REC_LEN		((1<<16)-1)
 
+static inline unsigned int
+ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
+{
+	unsigned len = le16_to_cpu(dlen);
+
+	if (len == EXT4_MAX_REC_LEN || len == 0)
+		return blocksize;
+	return (len & 65532) | ((len & 3) << 16);
+}
+
+static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
+{
+	if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
+		BUG();
+	if (len < 65536)
+		return cpu_to_le16(len);
+	if (len == blocksize) {
+		if (blocksize == 65536)
+			return cpu_to_le16(EXT4_MAX_REC_LEN);
+		else
+			return cpu_to_le16(0);
+	}
+	return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
+}
+
 /*
  * Hash Tree Directory indexing
  * (c) Daniel Phillips, 2001
@@ -1601,8 +1626,6 @@ extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
 extern int ext4_ext_migrate(struct inode *);
 
 /* namei.c */
-extern unsigned int ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize);
-extern __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize);
 extern int ext4_orphan_add(handle_t *, struct inode *);
 extern int ext4_orphan_del(handle_t *, struct inode *);
 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
diff --git a/fs/ext4/namei.c b/fs/ext4/namei.c
index a43e661..77a7468 100644
--- a/fs/ext4/namei.c
+++ b/fs/ext4/namei.c
@@ -179,30 +179,6 @@ static struct buffer_head * ext4_dx_find_entry(struct inode *dir,
 static int ext4_dx_add_entry(handle_t *handle, struct dentry *dentry,
 			     struct inode *inode);
 
-unsigned int ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
-{
-	unsigned len = le16_to_cpu(dlen);
-
-	if (len == EXT4_MAX_REC_LEN || len == 0)
-		return blocksize;
-	return (len & 65532) | ((len & 3) << 16);
-}
-
-__le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
-{
-	if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
-		BUG();
-	if (len < 65536)
-		return cpu_to_le16(len);
-	if (len == blocksize) {
-		if (blocksize == 65536)
-			return cpu_to_le16(EXT4_MAX_REC_LEN);
-		else
-			return cpu_to_le16(0);
-	}
-	return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
-}

^ permalink raw reply related	[flat|nested] 4+ messages in thread

* Re: [PATCH] ext4: re-inline ext4_rec_len_(to|from)_disk functions
  2010-08-04 15:52 [PATCH] ext4: re-inline ext4_rec_len_(to|from)_disk functions Eric Sandeen
@ 2010-08-05  0:11 ` Ted Ts'o
  2010-08-05  0:28   ` Eric Sandeen
  0 siblings, 1 reply; 4+ messages in thread
From: Ted Ts'o @ 2010-08-05  0:11 UTC (permalink / raw)
  To: Eric Sandeen; +Cc: ext4 development

Thanks for the commit.  Since we don't have support for fs block size
> page size (and having done a private investigation about what would
be required, I'm not sure it's going to happen any time soon --- and
if it does, yanking the #if statements is going to be a tiny part of
the patch :-).

So I slightly modified your patch to only do the extra complexity if
the page size is >= 65536.

						- Ted

commit 097639d271f40376f3b3a94c7ab242c22ddc8bc1
Author: Eric Sandeen <sandeen@redhat.com>
Date:   Wed Aug 4 20:09:07 2010 -0400

    ext4: re-inline ext4_rec_len_(to|from)_disk functions
    
    commit 3d0518f4, "ext4: New rec_len encoding for very
    large blocksizes" made several changes to this path, but from
    a perf perspective, un-inlining ext4_rec_len_from_disk() seems
    most significant.  This function is called from ext4_check_dir_entry(),
    which on a file-creation workload is called extremely often.
    
    I tested this with bonnie:
    
    # bonnie++ -u root -s 0 -f -x 200 -d /mnt/test -n 32
    
    (this does 200 iterations) and got this for the file creations:
    
    ext4 stock:   Average =  21206.8 files/s
    ext4 inlined: Average =  22346.7 files/s  (+5%)
    
    Signed-off-by: Eric Sandeen <sandeen@redhat.com>
    Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index ed14e1d..0f340bf 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -1412,6 +1412,43 @@ struct ext4_dir_entry_2 {
 #define EXT4_MAX_REC_LEN		((1<<16)-1)
 
 /*
+ * If we ever get support for fs block sizes > page_size, we'll need
+ * to remove the #if statements in the next two functions...
+ */
+static inline unsigned int
+ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
+{
+	unsigned len = le16_to_cpu(dlen);
+
+#if (PAGE_SIZE >= 65536)
+	if (len == EXT4_MAX_REC_LEN || len == 0)
+		return blocksize;
+	return (len & 65532) | ((len & 3) << 16);
+#else
+	return len;
+#endif
+}
+
+static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
+{
+	if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
+		BUG();
+#if (PAGE_SIZE >= 65536)
+	if (len < 65536)
+		return cpu_to_le16(len);
+	if (len == blocksize) {
+		if (blocksize == 65536)
+			return cpu_to_le16(EXT4_MAX_REC_LEN);
+		else
+			return cpu_to_le16(0);
+	}
+	return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
+#else
+	return cpu_to_le16(len);
+#endif
+}
+
+/*
  * Hash Tree Directory indexing
  * (c) Daniel Phillips, 2001
  */
@@ -1636,8 +1673,6 @@ extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
 extern int ext4_ext_migrate(struct inode *);
 
 /* namei.c */
-extern unsigned int ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize);
-extern __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize);
 extern int ext4_orphan_add(handle_t *, struct inode *);
 extern int ext4_orphan_del(handle_t *, struct inode *);
 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
diff --git a/fs/ext4/namei.c b/fs/ext4/namei.c
index ea8b59d..314c0d3 100644
--- a/fs/ext4/namei.c
+++ b/fs/ext4/namei.c
@@ -179,30 +179,6 @@ static struct buffer_head * ext4_dx_find_entry(struct inode *dir,
 static int ext4_dx_add_entry(handle_t *handle, struct dentry *dentry,
 			     struct inode *inode);
 
-unsigned int ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
-{
-	unsigned len = le16_to_cpu(dlen);
-
-	if (len == EXT4_MAX_REC_LEN || len == 0)
-		return blocksize;
-	return (len & 65532) | ((len & 3) << 16);
-}
-
-__le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
-{
-	if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
-		BUG();
-	if (len < 65536)
-		return cpu_to_le16(len);
-	if (len == blocksize) {
-		if (blocksize == 65536)
-			return cpu_to_le16(EXT4_MAX_REC_LEN);
-		else
-			return cpu_to_le16(0);
-	}
-	return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
-}

^ permalink raw reply related	[flat|nested] 4+ messages in thread

* Re: [PATCH] ext4: re-inline ext4_rec_len_(to|from)_disk functions
  2010-08-05  0:11 ` Ted Ts'o
@ 2010-08-05  0:28   ` Eric Sandeen
  2010-08-05  5:37     ` Ted Ts'o
  0 siblings, 1 reply; 4+ messages in thread
From: Eric Sandeen @ 2010-08-05  0:28 UTC (permalink / raw)
  To: Ted Ts'o; +Cc: ext4 development

Ted Ts'o wrote:
> Thanks for the commit.  Since we don't have support for fs block size
>> page size (and having done a private investigation about what would
> be required, I'm not sure it's going to happen any time soon --- and
> if it does, yanking the #if statements is going to be a tiny part of
> the patch :-).
> 
> So I slightly modified your patch to only do the extra complexity if
> the page size is >= 65536.

I had resisted doing that because at least the original rationale in
the patch was for when the VM allows block size > page size ...

Side note, if you do this is PAGE_SIZE or PAGE_CACHE_SIZE the right macro?

-Eric

> 						- Ted
> 
> commit 097639d271f40376f3b3a94c7ab242c22ddc8bc1
> Author: Eric Sandeen <sandeen@redhat.com>
> Date:   Wed Aug 4 20:09:07 2010 -0400
> 
>     ext4: re-inline ext4_rec_len_(to|from)_disk functions
>     
>     commit 3d0518f4, "ext4: New rec_len encoding for very
>     large blocksizes" made several changes to this path, but from
>     a perf perspective, un-inlining ext4_rec_len_from_disk() seems
>     most significant.  This function is called from ext4_check_dir_entry(),
>     which on a file-creation workload is called extremely often.
>     
>     I tested this with bonnie:
>     
>     # bonnie++ -u root -s 0 -f -x 200 -d /mnt/test -n 32
>     
>     (this does 200 iterations) and got this for the file creations:
>     
>     ext4 stock:   Average =  21206.8 files/s
>     ext4 inlined: Average =  22346.7 files/s  (+5%)
>     
>     Signed-off-by: Eric Sandeen <sandeen@redhat.com>
>     Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
> 
> diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
> index ed14e1d..0f340bf 100644
> --- a/fs/ext4/ext4.h
> +++ b/fs/ext4/ext4.h
> @@ -1412,6 +1412,43 @@ struct ext4_dir_entry_2 {
>  #define EXT4_MAX_REC_LEN		((1<<16)-1)
>  
>  /*
> + * If we ever get support for fs block sizes > page_size, we'll need
> + * to remove the #if statements in the next two functions...
> + */
> +static inline unsigned int
> +ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
> +{
> +	unsigned len = le16_to_cpu(dlen);
> +
> +#if (PAGE_SIZE >= 65536)
> +	if (len == EXT4_MAX_REC_LEN || len == 0)
> +		return blocksize;
> +	return (len & 65532) | ((len & 3) << 16);
> +#else
> +	return len;
> +#endif
> +}
> +
> +static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
> +{
> +	if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
> +		BUG();
> +#if (PAGE_SIZE >= 65536)
> +	if (len < 65536)
> +		return cpu_to_le16(len);
> +	if (len == blocksize) {
> +		if (blocksize == 65536)
> +			return cpu_to_le16(EXT4_MAX_REC_LEN);
> +		else
> +			return cpu_to_le16(0);
> +	}
> +	return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
> +#else
> +	return cpu_to_le16(len);
> +#endif
> +}
> +
> +/*
>   * Hash Tree Directory indexing
>   * (c) Daniel Phillips, 2001
>   */
> @@ -1636,8 +1673,6 @@ extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
>  extern int ext4_ext_migrate(struct inode *);
>  
>  /* namei.c */
> -extern unsigned int ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize);
> -extern __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize);
>  extern int ext4_orphan_add(handle_t *, struct inode *);
>  extern int ext4_orphan_del(handle_t *, struct inode *);
>  extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
> diff --git a/fs/ext4/namei.c b/fs/ext4/namei.c
> index ea8b59d..314c0d3 100644
> --- a/fs/ext4/namei.c
> +++ b/fs/ext4/namei.c
> @@ -179,30 +179,6 @@ static struct buffer_head * ext4_dx_find_entry(struct inode *dir,
>  static int ext4_dx_add_entry(handle_t *handle, struct dentry *dentry,
>  			     struct inode *inode);
>  
> -unsigned int ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
> -{
> -	unsigned len = le16_to_cpu(dlen);
> -
> -	if (len == EXT4_MAX_REC_LEN || len == 0)
> -		return blocksize;
> -	return (len & 65532) | ((len & 3) << 16);
> -}
> -
> -__le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
> -{
> -	if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
> -		BUG();
> -	if (len < 65536)
> -		return cpu_to_le16(len);
> -	if (len == blocksize) {
> -		if (blocksize == 65536)
> -			return cpu_to_le16(EXT4_MAX_REC_LEN);
> -		else
> -			return cpu_to_le16(0);
> -	}
> -	return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
> -}
> -
>  /*
>   * p is at least 6 bytes before the end of page
>   */


^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [PATCH] ext4: re-inline ext4_rec_len_(to|from)_disk functions
  2010-08-05  0:28   ` Eric Sandeen
@ 2010-08-05  5:37     ` Ted Ts'o
  0 siblings, 0 replies; 4+ messages in thread
From: Ted Ts'o @ 2010-08-05  5:37 UTC (permalink / raw)
  To: Eric Sandeen; +Cc: ext4 development

On Wed, Aug 04, 2010 at 07:28:20PM -0500, Eric Sandeen wrote:
> I had resisted doing that because at least the original rationale in
> the patch was for when the VM allows block size > page size ...

Well, that and apparently you can do completely insane/ludicrous page
sizes on the Itanic.  "The architecture supports 11 different page
sizes from 4k to 4GB."

> Side note, if you do this is PAGE_SIZE or PAGE_CACHE_SIZE the right macro?

Quite right, it should be PAGE_CACHE_SIZE (although I very much doubt
the two will be different any time in the near future).

    	     		      	      	       - Ted

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2010-08-05  5:37 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2010-08-04 15:52 [PATCH] ext4: re-inline ext4_rec_len_(to|from)_disk functions Eric Sandeen
2010-08-05  0:11 ` Ted Ts'o
2010-08-05  0:28   ` Eric Sandeen
2010-08-05  5:37     ` Ted Ts'o

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).