public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
From: George Spelvin <linux@horizon.com>
To: vda.linux@googlemail.com, mina86@mina86.com
Cc: hughd@google.com, linux-kernel@vger.kernel.org, linux@horizon.com
Subject: [PATCH 2/4] lib: vsprintf: Optimize division by 10000
Date: Fri,  3 Aug 2012 01:21:09 -0400	[thread overview]
Message-ID: <1343971271-13355-2-git-send-email-linux@horizon.com> (raw)
In-Reply-To: <1343971271-13355-1-git-send-email-linux@horizon.com>

The same multiply-by-inverse technique can be used to
convert division by 10000 to a 32x32->64-bit multiply.

Signed-off-by: George Spelvin <linux@horizon.com>
---
 lib/vsprintf.c |   60 +++++++++++++++++++++++++++++++-------------------------
 1 file changed, 33 insertions(+), 27 deletions(-)

This is something of an RFC, I haven't benchmarked the helper
function.  But it sure cleans up the code!

diff --git a/lib/vsprintf.c b/lib/vsprintf.c
index 2f32fe8..a8e7392 100644
--- a/lib/vsprintf.c
+++ b/lib/vsprintf.c
@@ -245,17 +245,32 @@ char *put_dec(char *buf, unsigned long long n)
 
 /* See comment in put_dec_full9 for choice of constants */
 static noinline_for_stack
-char *put_dec_full4(char *buf, unsigned q)
+void put_dec_full4(char *buf, unsigned q)
 {
 	unsigned r;
 	r      = (q * 0xccd) >> 15;
-	*buf++ = (q - 10 * r) + '0';
+	buf[0] = (q - 10 * r) + '0';
 	q      = (r * 0xcd) >> 11;
-	*buf++ = (r - 10 * q)  + '0';
+	buf[1] = (r - 10 * q)  + '0';
 	r      = (q * 0xcd) >> 11;
-	*buf++ = (q - 10 * r)  + '0';
-	*buf++ = r + '0';
-	return buf;
+	buf[2] = (q - 10 * r)  + '0';
+	buf[3] = r + '0';
+}
+
+/*
+ * Call put_dec_full4 on x % 10000, return x / 10000.
+ * The approximation x/10000 == (x * 0x346DC5D7) >> 43
+ * holds for all x < 1,128,869,999.  The largest value this
+ * helper will ever be asked to convert is 1,125,520,955.
+ * (d1 in the put_dec code, assuming n is all-ones).
+ */
+static
+unsigned put_dec_helper4(char *buf, unsigned x)
+{
+        uint32_t q = (x * (uint64_t)0x346DC5D7) >> 43;
+
+        put_dec_full4(buf, x - q * 10000);
+        return q;
 }
 
 /* Based on code by Douglas W. Jones found at
@@ -277,28 +292,19 @@ char *put_dec(char *buf, unsigned long long n)
 	d3  = (h >> 16); /* implicit "& 0xffff" */
 
 	q   = 656 * d3 + 7296 * d2 + 5536 * d1 + ((uint32_t)n & 0xffff);
+	q = put_dec_helper4(buf, q);
+
+	q += 7671 * d3 + 9496 * d2 + 6 * d1;
+	q = put_dec_helper4(buf+4, q);
+
+	q += 4749 * d3 + 42 * d2;
+	q = put_dec_helper4(buf+8, q);
 
-	buf = put_dec_full4(buf, q % 10000);
-	q   = q / 10000;
-
-	d1  = q + 7671 * d3 + 9496 * d2 + 6 * d1;
-	buf = put_dec_full4(buf, d1 % 10000);
-	q   = d1 / 10000;
-
-	d2  = q + 4749 * d3 + 42 * d2;
-	buf = put_dec_full4(buf, d2 % 10000);
-	q   = d2 / 10000;
-
-	d3  = q + 281 * d3;
-	if (!d3)
-		goto done;
-	buf = put_dec_full4(buf, d3 % 10000);
-	q   = d3 / 10000;
-	if (!q)
-		goto done;
-	buf = put_dec_full4(buf, q);
- done:
-	while (buf[-1] == '0')
+	q += 281 * d3;
+	buf += 12;
+	if (q)
+		buf = put_dec_trunc8(buf, q);
+	else while (buf[-1] == '0')
 		--buf;
 
 	return buf;
-- 
1.7.10.4


  reply	other threads:[~2012-08-03  5:21 UTC|newest]

Thread overview: 29+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2012-08-03  5:21 [PATCH 1/4] lib: vsprintf: Optimize division by 10 for small integers George Spelvin
2012-08-03  5:21 ` George Spelvin [this message]
2012-09-23 17:30   ` [PATCH 2/4] lib: vsprintf: Optimize division by 10000 Michal Nazarewicz
2012-09-24 12:16     ` George Spelvin
2012-09-24 12:41       ` Michal Nazarewicz
2012-09-24 13:56         ` George Spelvin
2012-09-24 15:14           ` Geert Uytterhoeven
2012-09-24 15:48             ` George Spelvin
2012-09-24  9:03   ` Denys Vlasenko
2012-09-24 12:35     ` George Spelvin
2012-09-24 15:02       ` Denys Vlasenko
2012-08-03  5:21 ` [PATCH 3/4] lib: vsprintf: Optimize put_dec_trunc8 George Spelvin
2012-09-23 14:18   ` Rabin Vincent
2012-09-24 11:13     ` George Spelvin
2012-09-24 14:33     ` George Spelvin
2012-09-24 14:53       ` Michal Nazarewicz
2012-09-24 14:57         ` Michal Nazarewicz
2012-09-23 18:22   ` Michal Nazarewicz
2012-09-24 11:46     ` George Spelvin
2012-09-24 12:29       ` Michal Nazarewicz
2012-09-24 13:49         ` George Spelvin
2012-09-24 15:06           ` Michal Nazarewicz
2012-09-25 11:44           ` George Spelvin
2012-09-25 13:00             ` Denys Vlasenko
2012-08-03  5:21 ` [PATCH 4/4] lib: vsprintf: Fix broken comments George Spelvin
2012-09-23 17:22 ` [PATCH 1/4] lib: vsprintf: Optimize division by 10 for small integers Michal Nazarewicz
2012-09-24 14:18   ` George Spelvin
2012-09-24  9:06 ` Denys Vlasenko
2012-09-24 11:27   ` George Spelvin

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=1343971271-13355-2-git-send-email-linux@horizon.com \
    --to=linux@horizon.com \
    --cc=hughd@google.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mina86@mina86.com \
    --cc=vda.linux@googlemail.com \
    /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