Merge branch 'jk/xdiff-drop-xdl-fast-hash'
Retire the "fast hash" that had disastrous performance issues in some corner cases. * jk/xdiff-drop-xdl-fast-hash: xdiff: drop XDL_FAST_HASH
This commit is contained in:
9
Makefile
9
Makefile
@ -338,11 +338,6 @@ all::
|
||||
#
|
||||
# Define NATIVE_CRLF if your platform uses CRLF for line endings.
|
||||
#
|
||||
# Define XDL_FAST_HASH to use an alternative line-hashing method in
|
||||
# the diff algorithm. It gives a nice speedup if your processor has
|
||||
# fast unaligned word loads. Does NOT work on big-endian systems!
|
||||
# Enabled by default on x86_64.
|
||||
#
|
||||
# Define GIT_USER_AGENT if you want to change how git identifies itself during
|
||||
# network interactions. The default is "git/$(GIT_VERSION)".
|
||||
#
|
||||
@ -1485,10 +1480,6 @@ ifndef NO_MSGFMT_EXTENDED_OPTIONS
|
||||
MSGFMT += --check --statistics
|
||||
endif
|
||||
|
||||
ifneq (,$(XDL_FAST_HASH))
|
||||
BASIC_CFLAGS += -DXDL_FAST_HASH
|
||||
endif
|
||||
|
||||
ifdef GMTIME_UNRELIABLE_ERRORS
|
||||
COMPAT_OBJS += compat/gmtime.o
|
||||
BASIC_CFLAGS += -DGMTIME_UNRELIABLE_ERRORS
|
||||
|
@ -17,9 +17,6 @@ endif
|
||||
# because maintaining the nesting to match is a pain. If
|
||||
# we had "elif" things would have been much nicer...
|
||||
|
||||
ifeq ($(uname_M),x86_64)
|
||||
XDL_FAST_HASH = YesPlease
|
||||
endif
|
||||
ifeq ($(uname_S),OSF1)
|
||||
# Need this for u_short definitions et al
|
||||
BASIC_CFLAGS += -D_OSF_SOURCE
|
||||
|
106
xdiff/xutils.c
106
xdiff/xutils.c
@ -264,110 +264,6 @@ static unsigned long xdl_hash_record_with_whitespace(char const **data,
|
||||
return ha;
|
||||
}
|
||||
|
||||
#ifdef XDL_FAST_HASH
|
||||
|
||||
#define REPEAT_BYTE(x) ((~0ul / 0xff) * (x))
|
||||
|
||||
#define ONEBYTES REPEAT_BYTE(0x01)
|
||||
#define NEWLINEBYTES REPEAT_BYTE(0x0a)
|
||||
#define HIGHBITS REPEAT_BYTE(0x80)
|
||||
|
||||
/* Return the high bit set in the first byte that is a zero */
|
||||
static inline unsigned long has_zero(unsigned long a)
|
||||
{
|
||||
return ((a - ONEBYTES) & ~a) & HIGHBITS;
|
||||
}
|
||||
|
||||
static inline long count_masked_bytes(unsigned long mask)
|
||||
{
|
||||
if (sizeof(long) == 8) {
|
||||
/*
|
||||
* Jan Achrenius on G+: microoptimized version of
|
||||
* the simpler "(mask & ONEBYTES) * ONEBYTES >> 56"
|
||||
* that works for the bytemasks without having to
|
||||
* mask them first.
|
||||
*/
|
||||
/*
|
||||
* return mask * 0x0001020304050608 >> 56;
|
||||
*
|
||||
* Doing it like this avoids warnings on 32-bit machines.
|
||||
*/
|
||||
long a = (REPEAT_BYTE(0x01) / 0xff + 1);
|
||||
return mask * a >> (sizeof(long) * 7);
|
||||
} else {
|
||||
/* Carl Chatfield / Jan Achrenius G+ version for 32-bit */
|
||||
/* (000000 0000ff 00ffff ffffff) -> ( 1 1 2 3 ) */
|
||||
long a = (0x0ff0001 + mask) >> 23;
|
||||
/* Fix the 1 for 00 case */
|
||||
return a & mask;
|
||||
}
|
||||
}
|
||||
|
||||
unsigned long xdl_hash_record(char const **data, char const *top, long flags)
|
||||
{
|
||||
unsigned long hash = 5381;
|
||||
unsigned long a = 0, mask = 0;
|
||||
char const *ptr = *data;
|
||||
char const *end = top - sizeof(unsigned long) + 1;
|
||||
|
||||
if (flags & XDF_WHITESPACE_FLAGS)
|
||||
return xdl_hash_record_with_whitespace(data, top, flags);
|
||||
|
||||
ptr -= sizeof(unsigned long);
|
||||
do {
|
||||
hash += hash << 5;
|
||||
hash ^= a;
|
||||
ptr += sizeof(unsigned long);
|
||||
if (ptr >= end)
|
||||
break;
|
||||
a = *(unsigned long *)ptr;
|
||||
/* Do we have any '\n' bytes in this word? */
|
||||
mask = has_zero(a ^ NEWLINEBYTES);
|
||||
} while (!mask);
|
||||
|
||||
if (ptr >= end) {
|
||||
/*
|
||||
* There is only a partial word left at the end of the
|
||||
* buffer. Because we may work with a memory mapping,
|
||||
* we have to grab the rest byte by byte instead of
|
||||
* blindly reading it.
|
||||
*
|
||||
* To avoid problems with masking in a signed value,
|
||||
* we use an unsigned char here.
|
||||
*/
|
||||
const char *p;
|
||||
for (p = top - 1; p >= ptr; p--)
|
||||
a = (a << 8) + *((const unsigned char *)p);
|
||||
mask = has_zero(a ^ NEWLINEBYTES);
|
||||
if (!mask)
|
||||
/*
|
||||
* No '\n' found in the partial word. Make a
|
||||
* mask that matches what we read.
|
||||
*/
|
||||
mask = 1UL << (8 * (top - ptr) + 7);
|
||||
}
|
||||
|
||||
/* The mask *below* the first high bit set */
|
||||
mask = (mask - 1) & ~mask;
|
||||
mask >>= 7;
|
||||
hash += hash << 5;
|
||||
hash ^= a & mask;
|
||||
|
||||
/* Advance past the last (possibly partial) word */
|
||||
ptr += count_masked_bytes(mask);
|
||||
|
||||
if (ptr < top) {
|
||||
assert(*ptr == '\n');
|
||||
ptr++;
|
||||
}
|
||||
|
||||
*data = ptr;
|
||||
|
||||
return hash;
|
||||
}
|
||||
|
||||
#else /* XDL_FAST_HASH */
|
||||
|
||||
unsigned long xdl_hash_record(char const **data, char const *top, long flags) {
|
||||
unsigned long ha = 5381;
|
||||
char const *ptr = *data;
|
||||
@ -384,8 +280,6 @@ unsigned long xdl_hash_record(char const **data, char const *top, long flags) {
|
||||
return ha;
|
||||
}
|
||||
|
||||
#endif /* XDL_FAST_HASH */
|
||||
|
||||
unsigned int xdl_hashbits(unsigned int size) {
|
||||
unsigned int val = 1, bits = 0;
|
||||
|
||||
|
Reference in New Issue
Block a user