xdiff: add a preprocessing step that trims files

xdiff has a `xdl_trim_ends` step that removes common lines, unmatchable
lines. That is in theory good, but happens too late - after splitting,
hashing, and adjusting the hash values so they are unique. Those splitting,
hashing and adjusting hash values steps could have noticeable overhead.

Diffing two large files with minor (one-line-ish) changes are not uncommon.
In that case, the raw performance of those preparation steps seriously
matter. Even allocating an O(N) array and storing line offsets to it is
expensive.  Therefore my previous attempts [1] [2] cannot be good enough
since they do not remove the O(N) array assignment.

This patch adds a preprocessing step - `xdl_trim_files` that runs before
other preprocessing steps. It counts common prefix and suffix and lines in
them (needed for displaying line number), without doing anything else.

Testing with a crafted large (169MB) file, with minor change:

```
  open('a','w').write(''.join('%s\n' % (i % 100000) for i in xrange(30000000) if i != 6000000))
  open('b','w').write(''.join('%s\n' % (i % 100000) for i in xrange(30000000) if i != 6003000))
```

Running xdiff by a simple binary [3], this patch improves the xdiff perf by
more than 10x for the above case:

```
  # xdiff before this patch
  2.41s user 1.13s system 98% cpu 3.592 total
  # xdiff after this patch
  0.14s user 0.16s system 98% cpu 0.309 total
  # gnu diffutils
  0.12s user 0.15s system 98% cpu 0.272 total
  # (best of 20 runs)
```

It's still slightly slower than GNU diffutils. But it's pretty close now.

Testing with real repo data:

For the whole repo, this patch makes xdiff 25% faster:

```
  # hg perfbdiff --count 100 --alldata -c d334afc585e2 --blocks [--xdiff]
  # xdiff, after
  ! wall 0.058861 comb 0.050000 user 0.050000 sys 0.000000 (best of 100)
  # xdiff, before
  ! wall 0.077816 comb 0.080000 user 0.080000 sys 0.000000 (best of 91)
  # bdiff
  ! wall 0.117473 comb 0.120000 user 0.120000 sys 0.000000 (best of 67)
```

For files that are long (ex. commands.py), the speedup is more than 3x, very
significant:

```
  # hg perfbdiff --count 3000 --blocks commands.py.i 1 [--xdiff]
  # xdiff, after
  ! wall 0.690583 comb 0.690000 user 0.690000 sys 0.000000 (best of 12)
  # xdiff, before
  ! wall 2.240361 comb 2.210000 user 2.210000 sys 0.000000 (best of 4)
  # bdiff
  ! wall 2.469852 comb 2.440000 user 2.440000 sys 0.000000 (best of 4)
```

[1]: https://phab.mercurial-scm.org/D2631
[2]: https://phab.mercurial-scm.org/D2634
[3]:

```
// Code to run xdiff from command line. No proper error handling.
#include <stdlib.h>
#include <unistd.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include "mercurial/thirdparty/xdiff/xdiff.h"
#define ensure(x) if (!(x)) exit(255);
mmfile_t readfile(const char *path) {
  struct stat st; int fd = open(path, O_RDONLY);
  fstat(fd, &st); mmfile_t file = { malloc(st.st_size), st.st_size };
  ensure(read(fd, file.ptr, st.st_size) == st.st_size); close(fd);
  return file;
}
int main(int argc, char const *argv[]) {
  mmfile_t a = readfile(argv[1]), b = readfile(argv[2]);
  xpparam_t xpp = {0}; xdemitconf_t xecfg = {0}; xdemitcb_t ecb = {0};
  xdl_diff(&a, &b, &xpp, &xecfg, &ecb);
  return 0;
}
```

Differential Revision: https://phab.mercurial-scm.org/D2686

mercurial/thirdparty/xdiff/xdiffi.c

@@ -1062,6 +1062,7 b' xdchange_t *xdl_get_hunk(xdchange_t **xs'
 static int xdl_call_hunk_func(xdfenv_t *xe, xdchange_t *xscr, xdemitcb_t *ecb,
 			      xdemitconf_t const *xecfg)
 {
+	long p = xe->nprefix, s = xe->nsuffix;
 	xdchange_t *xch, *xche;
 
 	if (!xecfg->hunk_func)
@@ -1073,6 +1074,10 b' static int xdl_call_hunk_func(xdfenv_t *'
 			xche = xdl_get_hunk(&xch, xecfg);
 			if (!xch)
 				break;
+			if (xch != xche)
+				xdl_bug("xch != xche");
+			xch->i1 += p;
+			xch->i2 += p;
 			if (xch->i1 > i1 || xch->i2 > i2) {
 				if (xecfg->hunk_func(i1, xch->i1, i2, xch->i2, ecb->priv) < 0)
 					return -1;
@@ -1080,16 +1085,18 b' static int xdl_call_hunk_func(xdfenv_t *'
 			i1 = xche->i1 + xche->chg1;
 			i2 = xche->i2 + xche->chg2;
 		}
-		if (xecfg->hunk_func(i1, n1, i2, n2, ecb->priv) < 0)
+		if (xecfg->hunk_func(i1, n1 + p + s, i2, n2 + p + s,
+				     ecb->priv) < 0)
 			return -1;
 	} else {
 		for (xch = xscr; xch; xch = xche->next) {
 			xche = xdl_get_hunk(&xch, xecfg);
 			if (!xch)
 				break;
-			if (xecfg->hunk_func(
-					xch->i1, xche->i1 + xche->chg1 - xch->i1,
-					xch->i2, xche->i2 + xche->chg2 - xch->i2,
+			if (xecfg->hunk_func(xch->i1 + p,
+					xche->i1 + xche->chg1 - xch->i1,
+					xch->i2 + p,
+					xche->i2 + xche->chg2 - xch->i2,
 					ecb->priv) < 0)
 				return -1;
 		}

mercurial/thirdparty/xdiff/xprepare.c

@@ -156,6 +156,87 b' static int xdl_classify_record(unsigned '
 }
 
 
+/*
+ * Trim common prefix from files.
+ *
+ * Note: trimming could affect hunk shifting. But the performance benefit
+ * outweighs the shift change. A diff result with suboptimal shifting is still
+ * valid.
+ */
+static void xdl_trim_files(mmfile_t *mf1, mmfile_t *mf2, long reserved,
+		xdfenv_t *xe, mmfile_t *out_mf1, mmfile_t *out_mf2) {
+	mmfile_t msmall, mlarge;
+	/* prefix lines, prefix bytes, suffix lines, suffix bytes */
+	long plines = 0, pbytes = 0, slines = 0, sbytes = 0, i;
+	/* prefix char pointer for msmall and mlarge */
+	const char *pp1, *pp2;
+	/* suffix char pointer for msmall and mlarge */
+	const char *ps1, *ps2;
+
+	/* reserved must >= 0 for the line boundary adjustment to work */
+	if (reserved < 0)
+		reserved = 0;
+
+	if (mf1->size < mf2->size) {
+		memcpy(&msmall, mf1, sizeof(mmfile_t));
+		memcpy(&mlarge, mf2, sizeof(mmfile_t));
+	} else {
+		memcpy(&msmall, mf2, sizeof(mmfile_t));
+		memcpy(&mlarge, mf1, sizeof(mmfile_t));
+	}
+
+	pp1 = msmall.ptr, pp2 = mlarge.ptr;
+	for (i = 0; i < msmall.size && *pp1 == *pp2; ++i) {
+		plines += (*pp1 == '\n');
+		pp1++, pp2++;
+	}
+
+	ps1 = msmall.ptr + msmall.size - 1, ps2 = mlarge.ptr + mlarge.size - 1;
+	while (ps1 > pp1 && *ps1 == *ps2) {
+		slines += (*ps1 == '\n');
+		ps1--, ps2--;
+	}
+
+	/* Retract common prefix and suffix boundaries for reserved lines */
+	if (plines <= reserved + 1) {
+		plines = 0;
+	} else {
+		i = 0;
+		while (i <= reserved) {
+			pp1--;
+			i += (*pp1 == '\n');
+		}
+		/* The new mmfile starts at the next char just after '\n' */
+		pbytes = pp1 - msmall.ptr + 1;
+		plines -= reserved;
+	}
+
+	if (slines <= reserved + 1) {
+		slines = 0;
+	} else {
+		/* Note: with compiler SIMD support (ex. -O3 -mavx2), this
+		 * might perform better than memchr. */
+		i = 0;
+		while (i <= reserved) {
+			ps1++;
+			i += (*ps1 == '\n');
+		}
+		/* The new mmfile includes this '\n' */
+		sbytes = msmall.ptr + msmall.size - ps1 - 1;
+		slines -= reserved;
+		if (msmall.ptr[msmall.size - 1] == '\n')
+			slines -= 1;
+	}
+
+	xe->nprefix = plines;
+	xe->nsuffix = slines;
+	out_mf1->ptr = mf1->ptr + pbytes;
+	out_mf1->size = mf1->size - pbytes - sbytes;
+	out_mf2->ptr = mf2->ptr + pbytes;
+	out_mf2->size = mf2->size - pbytes - sbytes;
+}
+
+
 static int xdl_prepare_ctx(unsigned int pass, mmfile_t *mf, long narec, xpparam_t const *xpp,
 			   xdlclassifier_t *cf, xdfile_t *xdf) {
 	unsigned int hbits;
@@ -254,10 +335,13 b' static void xdl_free_ctx(xdfile_t *xdf) '
 	xdl_cha_free(&xdf->rcha);
 }
 
+/* Reserved lines for trimming, to leave room for shifting */
+#define TRIM_RESERVED_LINES 100
 
 int xdl_prepare_env(mmfile_t *mf1, mmfile_t *mf2, xpparam_t const *xpp,
 		    xdfenv_t *xe) {
 	long enl1, enl2, sample;
+	mmfile_t tmf1, tmf2;
 	xdlclassifier_t cf;
 
 	memset(&cf, 0, sizeof(cf));
@@ -270,12 +354,14 b' int xdl_prepare_env(mmfile_t *mf1, mmfil'
 	if (xdl_init_classifier(&cf, enl1 + enl2 + 1, xpp->flags) < 0)
 		return -1;
 
-	if (xdl_prepare_ctx(1, mf1, enl1, xpp, &cf, &xe->xdf1) < 0) {
+	xdl_trim_files(mf1, mf2, TRIM_RESERVED_LINES, xe, &tmf1, &tmf2);
+
+	if (xdl_prepare_ctx(1, &tmf1, enl1, xpp, &cf, &xe->xdf1) < 0) {
 
 		xdl_free_classifier(&cf);
 		return -1;
 	}
-	if (xdl_prepare_ctx(2, mf2, enl2, xpp, &cf, &xe->xdf2) < 0) {
+	if (xdl_prepare_ctx(2, &tmf2, enl2, xpp, &cf, &xe->xdf2) < 0) {
 
 		xdl_free_ctx(&xe->xdf1);
 		xdl_free_classifier(&cf);

mercurial/thirdparty/xdiff/xtypes.h

@@ -60,6 +60,10 b' typedef struct s_xdfile {'
 
 typedef struct s_xdfenv {
 	xdfile_t xdf1, xdf2;
+
+	/* number of lines for common prefix and suffix that are removed
+	 * from xdf1 and xdf2 as a preprocessing step */
+	long nprefix, nsuffix;
 } xdfenv_t;