Skip to contentSkip to conversation

fs: seq_file: optimize seq_pad()

PATCH In review Dmitry Voytik 2 participants 3 messages Sep 10, 2014 VFS

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 fs: seq_file: optimize seq_pad()

Use seq_putc() instead of seq_printf() in seq_pad() because the former is faster.

---
 fs/seq_file.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
fs/seq_file.c +4 -2
index 3857b72..fef338e 100644
@@ -787,8 +787,10 @@ EXPORT_SYMBOL(seq_write);
 void seq_pad(struct seq_file *m, char c)
 {
 	int size = m->pad_until - m->count;
-	if (size > 0)
-		seq_printf(m, "%*s", size, "");
+
+	while (size-- > 0)
+		if (unlikely(seq_putc(m, ' ')))
+			return;
 	if (c)
 		seq_putc(m, c);
 }
Signature · 1 line
1.9.1
maintainer
#1 lore raw
Quoted 2 lines · Dmitry Voytik

Use seq_putc() instead of seq_printf() in seq_pad() because the former is faster.

_Solitary_ seq_putc() is certainly going to be faster, but that loop... Do you have profiling results, or is it just an apriori "printf must be sloooowwww"?

#2 lore raw
Quoted 7 lines · Al Viro
Quoted 2 lines · Dmitry Voytik

Use seq_putc() instead of seq_printf() in seq_pad() because the former is faster.

_Solitary_ seq_putc() is certainly going to be faster, but that loop... Do you have profiling results, or is it just an apriori "printf must be sloooowwww"?

My fail, sorry. The commit message is little bit wrong. I meant that simple looping of seq_putc() is faster than seq_printf(). I haven't done profilings. I just realized that seq_printf() is more complex than simple loop with seq_putc()
(no need to decode format string as in vsnprintf(), etc).
If I resend the patch with the following commit message:

Use a simple loop with seq_putc() instead of seq_printf() in seq_pad() as this approach is faster due to less complexity in terms of machine cycles.

Would be it Okay?
Thank you for reviewing.

Signature · 3 lines
Best Regards,
Dmitry Voytik.
[email protected]