The patch titled
Subject: lib: vsprintf: optimize division by 10000
has been added to the -mm tree. Its filename is
lib-vsprintf-optimize-division-by-10000.patch
Before you just go and hit "reply", please:
a) Consider who else should be cc'ed
b) Prefer to cc a suitable mailing list as well
c) Ideally: find the original patch on the mailing list and do a
reply-to-all to that, adding suitable additional cc's
*** Remember to use Documentation/SubmitChecklist when testing your code ***
The -mm tree is included into linux-next and is updated
there every 3-4 working days
------------------------------------------------------
From: George Spelvin <linux@xxxxxxxxxxx>
Subject: lib: vsprintf: optimize division by 10000
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@xxxxxxxxxxx>
Cc: Denys Vlasenko <vda.linux@xxxxxxxxxxxxxx>
Cc: Michal Nazarewicz <mina86@xxxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
lib/vsprintf.c | 58 +++++++++++++++++++++++++----------------------
1 file changed, 32 insertions(+), 26 deletions(-)
diff -puN lib/vsprintf.c~lib-vsprintf-optimize-division-by-10000 lib/vsprintf.c
--- a/lib/vsprintf.c~lib-vsprintf-optimize-division-by-10000
+++ a/lib/vsprintf.c
@@ -245,17 +245,32 @@ char *put_dec(char *buf, unsigned long l
/* 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 l
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);
- buf = put_dec_full4(buf, q % 10000);
- q = q / 10000;
+ q += 4749 * d3 + 42 * d2;
+ q = put_dec_helper4(buf+8, q);
- 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;
_
Patches currently in -mm which might be from linux@xxxxxxxxxxx are
lib-vsprintf-optimize-division-by-10-for-small-integers.patch
lib-vsprintf-optimize-division-by-10000.patch
lib-vsprintf-optimize-put_dec_trunc8.patch
lib-vsprintf-fix-broken-comments.patch
--
To unsubscribe from this list: send the line "unsubscribe mm-commits" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
