* [PATCH 0/4] Some minor printf improvements
@ 2015-02-20 23:21 Rasmus Villemoes
2015-02-20 23:21 ` [PATCH 1/4] lib/vsprintf.c: Eliminate some branches Rasmus Villemoes
` (3 more replies)
0 siblings, 4 replies; 5+ messages in thread
From: Rasmus Villemoes @ 2015-02-20 23:21 UTC (permalink / raw)
To: Andrew Morton; +Cc: Rasmus Villemoes, linux-kernel
While studying the kernel's printf internals in order to write the
kernel-specific format checking for smatch, I stumbled on a few things
one could do to shave an instruction here and there. Nothing
earth-shattering.
Rasmus Villemoes (4):
lib/vsprintf.c: Eliminate some branches
lib/vsprintf.c: Reduce stack use in number()
lib/vsprintf.c: Eliminate duplicate hex string array
lib/vsprintf.c: Another small hack
lib/vsprintf.c | 49 +++++++++++++++++--------------------------------
1 file changed, 17 insertions(+), 32 deletions(-)
--
2.1.3
^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH 1/4] lib/vsprintf.c: Eliminate some branches
2015-02-20 23:21 [PATCH 0/4] Some minor printf improvements Rasmus Villemoes
@ 2015-02-20 23:21 ` Rasmus Villemoes
2015-02-20 23:21 ` [PATCH 2/4] lib/vsprintf.c: Reduce stack use in number() Rasmus Villemoes
` (2 subsequent siblings)
3 siblings, 0 replies; 5+ messages in thread
From: Rasmus Villemoes @ 2015-02-20 23:21 UTC (permalink / raw)
To: Andrew Morton, Rasmus Villemoes, Peter Zijlstra (Intel), Tejun Heo
Cc: linux-kernel
Since FORMAT_TYPE_INT is simply 1 more than FORMAT_TYPE_UINT, and
similarly for BYTE/UBYTE, SHORT/USHORT, LONG/ULONG, we can eliminate a
few instructions by making SIGN have the value 1 instead of 2, and
then use arithmetic instead of branches for computing the right
spec->type. It's a little hacky, but certainly in the same spirit as
SMALL needing to have the value 0x20. For example for the
spec->qualifier == 'l' case, gcc now generates
75e: 0f b6 53 01 movzbl 0x1(%rbx),%edx
762: 83 e2 01 and $0x1,%edx
765: 83 c2 09 add $0x9,%edx
768: 88 13 mov %dl,(%rbx)
instead of
763: 0f b6 53 01 movzbl 0x1(%rbx),%edx
767: 83 e2 02 and $0x2,%edx
76a: 80 fa 01 cmp $0x1,%dl
76d: 19 d2 sbb %edx,%edx
76f: 83 c2 0a add $0xa,%edx
772: 88 13 mov %dl,(%rbx)
Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
---
lib/vsprintf.c | 30 +++++++++++-------------------
1 file changed, 11 insertions(+), 19 deletions(-)
diff --git a/lib/vsprintf.c b/lib/vsprintf.c
index b235c96167d3..318d583fe862 100644
--- a/lib/vsprintf.c
+++ b/lib/vsprintf.c
@@ -340,8 +340,8 @@ int num_to_str(char *buf, int size, unsigned long long num)
return len;
}
-#define ZEROPAD 1 /* pad with zero */
-#define SIGN 2 /* unsigned/signed long */
+#define SIGN 1 /* unsigned/signed, must be 1 */
+#define ZEROPAD 2 /* pad with zero */
#define PLUS 4 /* show plus */
#define SPACE 8 /* space if plus */
#define LEFT 16 /* left justified */
@@ -447,7 +447,7 @@ char *number(char *buf, char *end, unsigned long long num,
spec.precision = i;
/* leading space padding */
spec.field_width -= spec.precision;
- if (!(spec.flags & (ZEROPAD+LEFT))) {
+ if (!(spec.flags & (ZEROPAD | LEFT))) {
while (--spec.field_width >= 0) {
if (buf < end)
*buf = ' ';
@@ -1738,29 +1738,21 @@ qualifier:
if (spec->qualifier == 'L')
spec->type = FORMAT_TYPE_LONG_LONG;
else if (spec->qualifier == 'l') {
- if (spec->flags & SIGN)
- spec->type = FORMAT_TYPE_LONG;
- else
- spec->type = FORMAT_TYPE_ULONG;
+ BUILD_BUG_ON(FORMAT_TYPE_ULONG + SIGN != FORMAT_TYPE_LONG);
+ spec->type = FORMAT_TYPE_ULONG + (spec->flags & SIGN);
} else if (_tolower(spec->qualifier) == 'z') {
spec->type = FORMAT_TYPE_SIZE_T;
} else if (spec->qualifier == 't') {
spec->type = FORMAT_TYPE_PTRDIFF;
} else if (spec->qualifier == 'H') {
- if (spec->flags & SIGN)
- spec->type = FORMAT_TYPE_BYTE;
- else
- spec->type = FORMAT_TYPE_UBYTE;
+ BUILD_BUG_ON(FORMAT_TYPE_UBYTE + SIGN != FORMAT_TYPE_BYTE);
+ spec->type = FORMAT_TYPE_UBYTE + (spec->flags & SIGN);
} else if (spec->qualifier == 'h') {
- if (spec->flags & SIGN)
- spec->type = FORMAT_TYPE_SHORT;
- else
- spec->type = FORMAT_TYPE_USHORT;
+ BUILD_BUG_ON(FORMAT_TYPE_USHORT + SIGN != FORMAT_TYPE_SHORT);
+ spec->type = FORMAT_TYPE_USHORT + (spec->flags & SIGN);
} else {
- if (spec->flags & SIGN)
- spec->type = FORMAT_TYPE_INT;
- else
- spec->type = FORMAT_TYPE_UINT;
+ BUILD_BUG_ON(FORMAT_TYPE_UINT + SIGN != FORMAT_TYPE_INT);
+ spec->type = FORMAT_TYPE_UINT + (spec->flags & SIGN);
}
return ++fmt - start;
--
2.1.3
^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH 2/4] lib/vsprintf.c: Reduce stack use in number()
2015-02-20 23:21 [PATCH 0/4] Some minor printf improvements Rasmus Villemoes
2015-02-20 23:21 ` [PATCH 1/4] lib/vsprintf.c: Eliminate some branches Rasmus Villemoes
@ 2015-02-20 23:21 ` Rasmus Villemoes
2015-02-20 23:21 ` [PATCH 3/4] lib/vsprintf.c: Eliminate duplicate hex string array Rasmus Villemoes
2015-02-20 23:21 ` [PATCH 4/4] lib/vsprintf.c: Another small hack Rasmus Villemoes
3 siblings, 0 replies; 5+ messages in thread
From: Rasmus Villemoes @ 2015-02-20 23:21 UTC (permalink / raw)
To: Andrew Morton, Rasmus Villemoes, Peter Zijlstra (Intel), Tejun Heo
Cc: linux-kernel
At least since the initial git commit, when base was passed as a
separate parameter, number() has only been called with bases 8, 10 and
16. I'm guessing that 66 was to accommodate 64 0/1, a sign and a '\0',
but the buffer is only used for the actual digits. Octal digits carry
3 bits of information, so 24 is enough. Spell that 3*sizeof(num) so
one less place needs to be changed should long long ever be 128
bits. Also remove the commented-out code that would handle an
arbitrary base.
Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
---
lib/vsprintf.c | 7 +------
1 file changed, 1 insertion(+), 6 deletions(-)
diff --git a/lib/vsprintf.c b/lib/vsprintf.c
index 318d583fe862..98f1ce907d3d 100644
--- a/lib/vsprintf.c
+++ b/lib/vsprintf.c
@@ -386,7 +386,7 @@ char *number(char *buf, char *end, unsigned long long num,
/* we are called with base 8, 10 or 16, only, thus don't need "G..." */
static const char digits[16] = "0123456789ABCDEF"; /* "GHIJKLMNOPQRSTUVWXYZ"; */
- char tmp[66];
+ char tmp[3 * sizeof(num)];
char sign;
char locase;
int need_pfx = ((spec.flags & SPECIAL) && spec.base != 10);
@@ -423,11 +423,6 @@ char *number(char *buf, char *end, unsigned long long num,
i = 0;
if (num < spec.base)
tmp[i++] = digits[num] | locase;
- /* Generic code, for any base:
- else do {
- tmp[i++] = (digits[do_div(num,base)] | locase);
- } while (num != 0);
- */
else if (spec.base != 10) { /* 8 or 16 */
int mask = spec.base - 1;
int shift = 3;
--
2.1.3
^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH 3/4] lib/vsprintf.c: Eliminate duplicate hex string array
2015-02-20 23:21 [PATCH 0/4] Some minor printf improvements Rasmus Villemoes
2015-02-20 23:21 ` [PATCH 1/4] lib/vsprintf.c: Eliminate some branches Rasmus Villemoes
2015-02-20 23:21 ` [PATCH 2/4] lib/vsprintf.c: Reduce stack use in number() Rasmus Villemoes
@ 2015-02-20 23:21 ` Rasmus Villemoes
2015-02-20 23:21 ` [PATCH 4/4] lib/vsprintf.c: Another small hack Rasmus Villemoes
3 siblings, 0 replies; 5+ messages in thread
From: Rasmus Villemoes @ 2015-02-20 23:21 UTC (permalink / raw)
To: Andrew Morton, Rasmus Villemoes, Peter Zijlstra (Intel), Tejun Heo
Cc: linux-kernel
gcc doesn't merge or overlap const char[] objects with identical
contents (probably language lawyers would also insist that these
things have different addresses), but there's no reason to have the
string "0123456789ABCDEF" occur in multiple places. hex_asc_upper is
declared in kernel.h and defined in lib/hexdump.c, which is
unconditionally compiled in.
Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
---
lib/vsprintf.c | 7 ++-----
1 file changed, 2 insertions(+), 5 deletions(-)
diff --git a/lib/vsprintf.c b/lib/vsprintf.c
index 98f1ce907d3d..7a299d43987a 100644
--- a/lib/vsprintf.c
+++ b/lib/vsprintf.c
@@ -383,9 +383,6 @@ static noinline_for_stack
char *number(char *buf, char *end, unsigned long long num,
struct printf_spec spec)
{
- /* we are called with base 8, 10 or 16, only, thus don't need "G..." */
- static const char digits[16] = "0123456789ABCDEF"; /* "GHIJKLMNOPQRSTUVWXYZ"; */
-
char tmp[3 * sizeof(num)];
char sign;
char locase;
@@ -422,7 +419,7 @@ char *number(char *buf, char *end, unsigned long long num,
/* generate full string in tmp[], in reverse order */
i = 0;
if (num < spec.base)
- tmp[i++] = digits[num] | locase;
+ tmp[i++] = hex_asc_upper[num] | locase;
else if (spec.base != 10) { /* 8 or 16 */
int mask = spec.base - 1;
int shift = 3;
@@ -430,7 +427,7 @@ char *number(char *buf, char *end, unsigned long long num,
if (spec.base == 16)
shift = 4;
do {
- tmp[i++] = (digits[((unsigned char)num) & mask] | locase);
+ tmp[i++] = (hex_asc_upper[((unsigned char)num) & mask] | locase);
num >>= shift;
} while (num);
} else { /* base 10 */
--
2.1.3
^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH 4/4] lib/vsprintf.c: Another small hack
2015-02-20 23:21 [PATCH 0/4] Some minor printf improvements Rasmus Villemoes
` (2 preceding siblings ...)
2015-02-20 23:21 ` [PATCH 3/4] lib/vsprintf.c: Eliminate duplicate hex string array Rasmus Villemoes
@ 2015-02-20 23:21 ` Rasmus Villemoes
3 siblings, 0 replies; 5+ messages in thread
From: Rasmus Villemoes @ 2015-02-20 23:21 UTC (permalink / raw)
To: Andrew Morton, Rasmus Villemoes, Peter Zijlstra (Intel), Tejun Heo
Cc: linux-kernel
Making ZEROPAD == '0'-' ', we can eliminate a few more instructions.
Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
---
lib/vsprintf.c | 7 ++++---
1 file changed, 4 insertions(+), 3 deletions(-)
diff --git a/lib/vsprintf.c b/lib/vsprintf.c
index 7a299d43987a..2753f9261115 100644
--- a/lib/vsprintf.c
+++ b/lib/vsprintf.c
@@ -341,10 +341,10 @@ int num_to_str(char *buf, int size, unsigned long long num)
}
#define SIGN 1 /* unsigned/signed, must be 1 */
-#define ZEROPAD 2 /* pad with zero */
+#define LEFT 2 /* left justified */
#define PLUS 4 /* show plus */
#define SPACE 8 /* space if plus */
-#define LEFT 16 /* left justified */
+#define ZEROPAD 16 /* pad with zero, must be 16 == '0' - ' ' */
#define SMALL 32 /* use lowercase in hex (must be 32 == 0x20) */
#define SPECIAL 64 /* prefix hex with "0x", octal with "0" */
@@ -467,7 +467,8 @@ char *number(char *buf, char *end, unsigned long long num,
}
/* zero or space padding */
if (!(spec.flags & LEFT)) {
- char c = (spec.flags & ZEROPAD) ? '0' : ' ';
+ char c = ' ' + (spec.flags & ZEROPAD);
+ BUILD_BUG_ON(' ' + ZEROPAD != '0');
while (--spec.field_width >= 0) {
if (buf < end)
*buf = c;
--
2.1.3
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2015-02-20 23:22 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2015-02-20 23:21 [PATCH 0/4] Some minor printf improvements Rasmus Villemoes
2015-02-20 23:21 ` [PATCH 1/4] lib/vsprintf.c: Eliminate some branches Rasmus Villemoes
2015-02-20 23:21 ` [PATCH 2/4] lib/vsprintf.c: Reduce stack use in number() Rasmus Villemoes
2015-02-20 23:21 ` [PATCH 3/4] lib/vsprintf.c: Eliminate duplicate hex string array Rasmus Villemoes
2015-02-20 23:21 ` [PATCH 4/4] lib/vsprintf.c: Another small hack Rasmus Villemoes
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®