I have spotted the biggest bottleneck in "bdiff.c". Actually it was

pretty easy to find after I recompiled the python interpreter and
mercurial for profiling.

In "bdiff.c" function "equatelines" allocates the minimum hash table
size, which can lead to tons of collisions. I introduced an
"overcommit" factor of 16, this is, I allocate 16 times more memory
than the minimum value. Overcommiting 128 times does not improve the
performance over the 16-times case.

mercurial/bdiff.c

@@ -1,373 +1,380
 /*
  bdiff.c - efficient binary diff extension for Mercurial
 
  Copyright 2005, 2006 Matt Mackall <[e-mail]>
 
  This software may be used and distributed according to the terms of
  the GNU General Public License, incorporated herein by reference.
 
  Based roughly on Python difflib
 */
 
 #include <Python.h>
 #include <stdlib.h>
 #include <string.h>
 
 #if defined __hpux || defined __SUNPRO_C || defined _AIX
 # define inline
 #endif
 
 #ifdef _WIN32
 #ifdef _MSC_VER
 #define inline __inline
 typedef unsigned long uint32_t;
 #else
 #include <stdint.h>
 #endif
 static uint32_t htonl(uint32_t x)
 {
 	return ((x & 0x000000ffUL) << 24) |
 		((x & 0x0000ff00UL) <<  8) |
 		((x & 0x00ff0000UL) >>  8) |
 		((x & 0xff000000UL) >> 24);
 }
 #else
 #include <sys/types.h>
 #ifdef __BEOS__
 #include <ByteOrder.h>
 #else
 #include <arpa/inet.h>
 #endif
 #include <inttypes.h>
 #endif
 
 struct line {
 	int h, len, n, e;
 	const char *l;
 };
 
 struct pos {
 	int pos, len;
 };
 
 struct hunk {
 	int a1, a2, b1, b2;
 };
 
 struct hunklist {
 	struct hunk *base, *head;
 };
 
 static inline uint32_t rol32(uint32_t word, unsigned int shift)
 {
         return (word << shift) | (word >> (32 - shift));
 }
 
 int splitlines(const char *a, int len, struct line **lr)
 {
 	int g, h, i;
 	const char *p, *b = a;
 	struct line *l;
 
 	/* count the lines */
 	i = 1; /* extra line for sentinel */
 	for (p = a; p < a + len; p++)
 		if (*p == '\n' || p == a + len - 1)
 			i++;
 
 	*lr = l = (struct line *)malloc(sizeof(struct line) * i);
 	if (!l)
 		return -1;
 
 	/* build the line array and calculate hashes */
 	h = 0;
 	for (p = a; p < a + len; p++) {
 		/*
 		 * a simple hash from GNU diff, with better collision
 		 * resistance from hashpjw. this slows down common
 		 * case by 10%, but speeds up worst case by 100x.
 		 */
 		h = *p + rol32(h, 7);
 		if ((g = h & 0xf0000000)) {
 			h ^= g >> 24;
 			h ^= g;
 		}
 		if (*p == '\n' || p == a + len - 1) {
 			l->len = p - b + 1;
 			l->h = h * l->len;
 			l->l = b;
 			l->n = -1;
 			l++;
 			b = p + 1;
 			h = 0;
 		}
 	}
 
 	/* set up a sentinel */
 	l->h = l->len = 0;
 	l->l = a + len;
 	return i - 1;
 }
 
 int inline cmp(struct line *a, struct line *b)
 {
 	return a->h != b->h || a->len != b->len || memcmp(a->l, b->l, a->len);
 }
 
 static int equatelines(struct line *a, int an, struct line *b, int bn)
 {
 	int i, j, buckets = 1, t;
+	int scale = 32;
 	struct pos *h;
 
 	/* build a hash table of the next highest power of 2 */
 	while (buckets < bn + 1)
 		buckets *= 2;
 
-	h = (struct pos *)malloc(buckets * sizeof(struct pos));
-	buckets = buckets - 1;
+	/* try to allocate a large hash table to avoid collisions */
+	do {
+		scale /= 2;
+		h = (struct pos *)malloc(scale * buckets * sizeof(struct pos));
+	} while (!h && scale != 1);
+
 	if (!h)
 		return 0;
 
+	buckets = buckets * scale - 1;
+
 	/* clear the hash table */
 	for (i = 0; i <= buckets; i++) {
 		h[i].pos = -1;
 		h[i].len = 0;
 	}
 
 	/* add lines to the hash table chains */
 	for (i = bn - 1; i >= 0; i--) {
 		/* find the equivalence class */
 		for (j = b[i].h & buckets; h[j].pos != -1;
 		     j = (j + 1) & buckets)
 			if (!cmp(b + i, b + h[j].pos))
 				break;
 
 		/* add to the head of the equivalence class */
 		b[i].n = h[j].pos;
 		b[i].e = j;
 		h[j].pos = i;
 		h[j].len++; /* keep track of popularity */
 	}
 
 	/* compute popularity threshold */
 	t = (bn >= 200) ? bn / 100 : bn + 1;
 
 	/* match items in a to their equivalence class in b */
 	for (i = 0; i < an; i++) {
 		/* find the equivalence class */
 		for (j = a[i].h & buckets; h[j].pos != -1;
 		     j = (j + 1) & buckets)
 			if (!cmp(a + i, b + h[j].pos))
 				break;
 
 		a[i].e = j; /* use equivalence class for quick compare */
 		if (h[j].len <= t)
 			a[i].n = h[j].pos; /* point to head of match list */
 		else
 			a[i].n = -1; /* too popular */
 	}
 
 	/* discard hash tables */
 	free(h);
 	return 1;
 }
 
 static int longest_match(struct line *a, struct line *b, struct pos *pos,
 			 int a1, int a2, int b1, int b2, int *omi, int *omj)
 {
 	int mi = a1, mj = b1, mk = 0, mb = 0, i, j, k;
 
 	for (i = a1; i < a2; i++) {
 		/* skip things before the current block */
 		for (j = a[i].n; j != -1 && j < b1; j = b[j].n)
 			;
 
 		/* loop through all lines match a[i] in b */
 		for (; j != -1 && j < b2; j = b[j].n) {
 			/* does this extend an earlier match? */
 			if (i > a1 && j > b1 && pos[j - 1].pos == i - 1)
 				k = pos[j - 1].len + 1;
 			else
 				k = 1;
 			pos[j].pos = i;
 			pos[j].len = k;
 
 			/* best match so far? */
 			if (k > mk) {
 				mi = i;
 				mj = j;
 				mk = k;
 			}
 		}
 	}
 
 	if (mk) {
 		mi = mi - mk + 1;
 		mj = mj - mk + 1;
 	}
 
 	/* expand match to include neighboring popular lines */
 	while (mi - mb > a1 && mj - mb > b1 &&
 	       a[mi - mb - 1].e == b[mj - mb - 1].e)
 		mb++;
 	while (mi + mk < a2 && mj + mk < b2 &&
 	       a[mi + mk].e == b[mj + mk].e)
 		mk++;
 
 	*omi = mi - mb;
 	*omj = mj - mb;
 	return mk + mb;
 }
 
 static void recurse(struct line *a, struct line *b, struct pos *pos,
 		    int a1, int a2, int b1, int b2, struct hunklist *l)
 {
 	int i, j, k;
 
 	/* find the longest match in this chunk */
 	k = longest_match(a, b, pos, a1, a2, b1, b2, &i, &j);
 	if (!k)
 		return;
 
 	/* and recurse on the remaining chunks on either side */
 	recurse(a, b, pos, a1, i, b1, j, l);
 	l->head->a1 = i;
 	l->head->a2 = i + k;
 	l->head->b1 = j;
 	l->head->b2 = j + k;
 	l->head++;
 	recurse(a, b, pos, i + k, a2, j + k, b2, l);
 }
 
 static struct hunklist diff(struct line *a, int an, struct line *b, int bn)
 {
 	struct hunklist l;
 	struct pos *pos;
 	int t;
 
 	/* allocate and fill arrays */
 	t = equatelines(a, an, b, bn);
 	pos = (struct pos *)calloc(bn, sizeof(struct pos));
 	/* we can't have more matches than lines in the shorter file */
 	l.head = l.base = (struct hunk *)malloc(sizeof(struct hunk) *
 	                                        ((an<bn ? an:bn) + 1));
 
 	if (pos && l.base && t) {
 		/* generate the matching block list */
 		recurse(a, b, pos, 0, an, 0, bn, &l);
 		l.head->a1 = l.head->a2 = an;
 		l.head->b1 = l.head->b2 = bn;
 		l.head++;
 	}
 
 	free(pos);
 	return l;
 }
 
 static PyObject *blocks(PyObject *self, PyObject *args)
 {
 	PyObject *sa, *sb, *rl = NULL, *m;
 	struct line *a, *b;
 	struct hunklist l = {NULL, NULL};
 	struct hunk *h;
 	int an, bn, pos = 0;
 
 	if (!PyArg_ParseTuple(args, "SS:bdiff", &sa, &sb))
 		return NULL;
 
 	an = splitlines(PyString_AsString(sa), PyString_Size(sa), &a);
 	bn = splitlines(PyString_AsString(sb), PyString_Size(sb), &b);
 	if (!a || !b)
 		goto nomem;
 
 	l = diff(a, an, b, bn);
 	rl = PyList_New(l.head - l.base);
 	if (!l.head || !rl)
 		goto nomem;
 
 	for (h = l.base; h != l.head; h++) {
 		m = Py_BuildValue("iiii", h->a1, h->a2, h->b1, h->b2);
 		PyList_SetItem(rl, pos, m);
 		pos++;
 	}
 
 nomem:
 	free(a);
 	free(b);
 	free(l.base);
 	return rl ? rl : PyErr_NoMemory();
 }
 
 static PyObject *bdiff(PyObject *self, PyObject *args)
 {
 	char *sa, *sb;
 	PyObject *result = NULL;
 	struct line *al, *bl;
 	struct hunklist l = {NULL, NULL};
 	struct hunk *h;
 	char encode[12], *rb;
 	int an, bn, len = 0, la, lb;
 
 	if (!PyArg_ParseTuple(args, "s#s#:bdiff", &sa, &la, &sb, &lb))
 		return NULL;
 
 	an = splitlines(sa, la, &al);
 	bn = splitlines(sb, lb, &bl);
 	if (!al || !bl)
 		goto nomem;
 
 	l = diff(al, an, bl, bn);
 	if (!l.head)
 		goto nomem;
 
 	/* calculate length of output */
 	la = lb = 0;
 	for (h = l.base; h != l.head; h++) {
 		if (h->a1 != la || h->b1 != lb)
 			len += 12 + bl[h->b1].l - bl[lb].l;
 		la = h->a2;
 		lb = h->b2;
 	}
 
 	result = PyString_FromStringAndSize(NULL, len);
 	if (!result)
 		goto nomem;
 
 	/* build binary patch */
 	rb = PyString_AsString(result);
 	la = lb = 0;
 
 	for (h = l.base; h != l.head; h++) {
 		if (h->a1 != la || h->b1 != lb) {
 			len = bl[h->b1].l - bl[lb].l;
 			*(uint32_t *)(encode)     = htonl(al[la].l - al->l);
 			*(uint32_t *)(encode + 4) = htonl(al[h->a1].l - al->l);
 			*(uint32_t *)(encode + 8) = htonl(len);
 			memcpy(rb, encode, 12);
 			memcpy(rb + 12, bl[lb].l, len);
 			rb += 12 + len;
 		}
 		la = h->a2;
 		lb = h->b2;
 	}
 
 nomem:
 	free(al);
 	free(bl);
 	free(l.base);
 	return result ? result : PyErr_NoMemory();
 }
 
 static char mdiff_doc[] = "Efficient binary diff.";
 
 static PyMethodDef methods[] = {
 	{"bdiff", bdiff, METH_VARARGS, "calculate a binary diff\n"},
 	{"blocks", blocks, METH_VARARGS, "find a list of matching lines\n"},
 	{NULL, NULL}
 };
 
 PyMODINIT_FUNC initbdiff(void)
 {
 	Py_InitModule3("bdiff", methods, mdiff_doc);
 }