[PATCH 2/2] fs: optimize mpage_bio_submit()

[PATCH 2/2] fs: optimize mpage_bio_submit()

[Changli Gao]

optimize mpage_bio_submit()

check rw, then initialize bio->bi_end_io.

Signed-off-by: Changli Gao <xiaosuo@gmail.com>
----
 fs/mpage.c |    4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)
diff --git a/fs/mpage.c b/fs/mpage.c
index 94ff0d1..632681c 100644
--- a/fs/mpage.c
+++ b/fs/mpage.c
@@ -85,9 +85,7 @@ static void mpage_end_io_write(struct bio *bio, int err)
 
 static struct bio *mpage_bio_submit(int rw, struct bio *bio)
 {
-	bio->bi_end_io = mpage_end_io_read;
-	if (rw == WRITE)
-		bio->bi_end_io = mpage_end_io_write;
+	bio->bi_end_io = rw != WRITE ? mpage_end_io_read : mpage_end_io_write;
 	submit_bio(rw, bio);
 	return NULL;
 }

Re: [PATCH 2/2] fs: optimize mpage_bio_submit()

[Borislav Petkov]

From: Changli Gao <xiaosuo@gmail.com>
Date: Sat, May 29, 2010 at 09:19:26AM +0800

> optimize mpage_bio_submit()
> 
> check rw, then initialize bio->bi_end_io.
> 
> Signed-off-by: Changli Gao <xiaosuo@gmail.com>
> ----
>  fs/mpage.c |    4 +---
>  1 file changed, 1 insertion(+), 3 deletions(-)
> diff --git a/fs/mpage.c b/fs/mpage.c
> index 94ff0d1..632681c 100644
> --- a/fs/mpage.c
> +++ b/fs/mpage.c
> @@ -85,9 +85,7 @@ static void mpage_end_io_write(struct bio *bio, int err)
>  
>  static struct bio *mpage_bio_submit(int rw, struct bio *bio)
>  {
> -	bio->bi_end_io = mpage_end_io_read;
> -	if (rw == WRITE)
> -		bio->bi_end_io = mpage_end_io_write;
> +	bio->bi_end_io = rw != WRITE ? mpage_end_io_read : mpage_end_io_write;
>  	submit_bio(rw, bio);
>  	return NULL;

ok, can you sincerely tell me that your change makes the code more
readable? And have you checked to see what the compiler actually
"optimizes"?

original code:

	movq	$mpage_end_io_read, %rdx	#, tmp65
	movq	$mpage_end_io_write, %rax	#, tmp64
	cmpl	$1, %edi	#, rw
	cmovne	%rdx, %rax	# tmp65,, tmp64
	movq	%rax, 80(%rsi)	# tmp64, <variable>.bi_end_io
	.loc 1 91 0
	call	submit_bio	#

your change:

	movq	$mpage_end_io_read, %rdx	#, tmp63
	movq	$mpage_end_io_write, %rax	#, iftmp.561
	cmpl	$1, %edi	#, rw
	cmovne	%rdx, %rax	# tmp63,, iftmp.561
	movq	%rax, 80(%rsi)	# iftmp.561, <variable>.bi_end_io
	.loc 1 90 0
	call	submit_bio	#

so that change does nothing except obfuscating the source a bit more so
that you get more headaches whenever you look at it.

-- 
Regards/Gruss,
    Boris.

Re: [PATCH 2/2] fs: optimize mpage_bio_submit()

[Changli Gao]

On Sat, May 29, 2010 at 8:47 PM, Borislav Petkov <bp@alien8.de> wrote:
>
> ok, can you sincerely tell me that your change makes the code more
> readable? And have you checked to see what the compiler actually
> "optimizes"?
>
> original code:
>
>        movq    $mpage_end_io_read, %rdx        #, tmp65
>        movq    $mpage_end_io_write, %rax       #, tmp64
>        cmpl    $1, %edi        #, rw
>        cmovne  %rdx, %rax      # tmp65,, tmp64
>        movq    %rax, 80(%rsi)  # tmp64, <variable>.bi_end_io
>        .loc 1 91 0
>        call    submit_bio      #
>
> your change:
>
>        movq    $mpage_end_io_read, %rdx        #, tmp63
>        movq    $mpage_end_io_write, %rax       #, iftmp.561
>        cmpl    $1, %edi        #, rw
>        cmovne  %rdx, %rax      # tmp63,, iftmp.561
>        movq    %rax, 80(%rsi)  # iftmp.561, <variable>.bi_end_io
>        .loc 1 90 0
>        call    submit_bio      #
>
> so that change does nothing except obfuscating the source a bit more so
> that you get more headaches whenever you look at it.
>

Oh, I didn't check the asm code. Compiler is smarter than I thought.
Thanks for your explaining.


-- 
Regards，
Changli Gao(xiaosuo@gmail.com)
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 2/2] fs: optimize mpage_bio_submit()

[Borislav Petkov]

From: Changli Gao <xiaosuo@gmail.com>
Date: Sat, May 29, 2010 at 09:33:02PM +0800

> Compiler is smarter than I thought.

Not as much as we would like to though :)

-- 
Regards/Gruss,
    Boris.