mingw-compat: Add simplified merge sort implementation from glibc

qsort in Windows 2000 (and possibly other older Windows' C libraries)
is a Quicksort with the usual O(n^2) worst case.  Unfortunately, sorting
Git trees seems to get very close to that worst case quite often:

    $ /git/gitbad runstatus
    # On branch master
    qsort, nmemb = 30842
    done, 237838087 comparisons.

This patch adds a simplified version of the merge sort that is glibc's
qsort(3).  As a merge sort, this needs a temporary array equal in size
to the array that is to be sorted.

The complexity that was removed is:

* Doing direct stores for word-size and -aligned data.
* Falling back to quicksort if the allocation required to perform the
  merge sort would likely push the machine into swap.

Even with these simplifications, this seems to outperform the Windows
qsort(3) implementation, even in Windows XP (where it is "fixed" and
doesn't trigger O(n^2) complexity on trees).

[jes: moved into compat/qsort.c, as per Johannes Sixt's suggestion]

Signed-off-by: Brian Downing <bdowning@lavos.net>
Signed-off-by: Steffen Prohaska <prohaska@zib.de>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>

Makefile

@@ -136,6 +136,9 @@ all::
 # Define THREADED_DELTA_SEARCH if you have pthreads and wish to exploit
 # parallel delta searching when packing objects.
 #
+# Define NEEDS_QUICK_QSORT if your qsort() implementation has O(n^2)
+# worst case complexity.
+#
 
 GIT-VERSION-FILE: .FORCE-GIT-VERSION-FILE
 	@"$(SHELL_PATH)" ./GIT-VERSION-GEN
@@ -523,6 +526,7 @@ ifneq (,$(findstring MINGW,$(uname_S)))
 	NO_SVN_TESTS = YesPlease
 	NO_PERL_MAKEMAKER = YesPlease
 	NO_R_TO_GCC_LINKER = YesPlease
+	NEEDS_QUICK_QSORT = YesPlease
 	NO_EXTRA_PROGRAMS = YesPlease
 	COMPAT_CFLAGS += -D__USE_MINGW_ACCESS -DNOGDI -Icompat
 	COMPAT_CFLAGS += -DSTRIP_EXTENSION=\".exe\"
@@ -741,6 +745,10 @@ ifdef NO_MEMMEM
 	COMPAT_CFLAGS += -DNO_MEMMEM
 	COMPAT_OBJS += compat/memmem.o
 endif
+ifdef NEEDS_QUICK_QSORT
+	COMPAT_CFLAGS += -DNEEDS_QUICK_QSORT
+	COMPAT_OBJS += compat/qsort.o
+endif
 
 ifdef THREADED_DELTA_SEARCH
 	BASIC_CFLAGS += -DTHREADED_DELTA_SEARCH

compat/qsort.c

@@ -0,0 +1,60 @@
+#include "../git-compat-util.h"
+
+/* This merge sort implementation is simplified from glibc's. */
+static void msort_with_tmp(void *b, size_t n, size_t s,
+			   int (*cmp)(const void *, const void *),
+			   char *t)
+{
+	char *tmp;
+	char *b1, *b2;
+	size_t n1, n2;
+
+	if (n <= 1)
+		return;
+
+	n1 = n / 2;
+	n2 = n - n1;
+	b1 = b;
+	b2 = (char *)b + (n1 * s);
+
+	msort_with_tmp(b1, n1, s, cmp, t);
+	msort_with_tmp(b2, n2, s, cmp, t);
+
+	tmp = t;
+
+	while (n1 > 0 && n2 > 0) {
+		if (cmp(b1, b2) <= 0) {
+			memcpy(tmp, b1, s);
+			tmp += s;
+			b1 += s;
+			--n1;
+		} else {
+			memcpy(tmp, b2, s);
+			tmp += s;
+			b2 += s;
+			--n2;
+		}
+	}
+	if (n1 > 0)
+		memcpy(tmp, b1, n1 * s);
+	memcpy(b, t, (n - n2) * s);
+}
+
+void git_qsort(void *b, size_t n, size_t s,
+	       int (*cmp)(const void *, const void *))
+{
+	const size_t size = n * s;
+
+	if (size < 1024) {
+		char buf[size]; /* gcc-ism */
+
+		/* The temporary array is small, so put it on
+		   the stack.  */
+		msort_with_tmp(b, n, s, cmp, buf);
+	} else {
+		/* It's somewhat large, so malloc it.  */
+		char *tmp = malloc(size);
+		msort_with_tmp(b, n, s, cmp, tmp);
+		free(tmp);
+	}
+}

git-compat-util.h

@@ -645,4 +645,10 @@ extern __attribute__((noreturn)) int git_exit(int code);
 
 #endif /* __MINGW32__ */
 
+#ifdef NEEDS_QUICK_QSORT
+void git_qsort(void *base, size_t nmemb, size_t size,
+	       int(*compar)(const void *, const void *));
+#define qsort git_qsort
+#endif
+
 #endif