From: Chris Wedgwood This is a rehash of a patch from Andreas Gruenbacher to add qsort as a kernel library function. Right now the only user of this is XFS others users are expected. The stack utilization is fairly modest and the function is not recursive. The symbol is exported GPL only recognized the GPL origin in the code used. --- 25-akpm/include/linux/kernel.h | 2 25-akpm/lib/Kconfig | 3 25-akpm/lib/Makefile | 1 25-akpm/lib/qsort.c | 239 +++++++++++++++++++++++++++++++++++++++++ 4 files changed, 245 insertions(+) diff -puN include/linux/kernel.h~add-qsort-library-function include/linux/kernel.h --- 25/include/linux/kernel.h~add-qsort-library-function 2004-05-10 01:20:02.514235448 -0700 +++ 25-akpm/include/linux/kernel.h 2004-05-10 01:20:02.519234688 -0700 @@ -80,6 +80,8 @@ extern int sscanf(const char *, const ch __attribute__ ((format (scanf,2,3))); extern int vsscanf(const char *, const char *, va_list); +extern void qsort(void *, size_t, size_t, int (*)(const void *,const void *)); + extern int get_option(char **str, int *pint); extern char *get_options(const char *str, int nints, int *ints); extern unsigned long long memparse(char *ptr, char **retptr); diff -puN lib/Kconfig~add-qsort-library-function lib/Kconfig --- 25/lib/Kconfig~add-qsort-library-function 2004-05-10 01:20:02.515235296 -0700 +++ 25-akpm/lib/Kconfig 2004-05-10 01:20:02.520234536 -0700 @@ -21,6 +21,9 @@ config LIBCRC32C require M here. See Castagnoli93. Module will be libcrc32c. +config QSORT + bool "Quick Sort" + # # compression support is select'ed if needed # diff -puN lib/Makefile~add-qsort-library-function lib/Makefile --- 25/lib/Makefile~add-qsort-library-function 2004-05-10 01:20:02.516235144 -0700 +++ 25-akpm/lib/Makefile 2004-05-10 01:20:02.520234536 -0700 @@ -20,6 +20,7 @@ endif obj-$(CONFIG_CRC32) += crc32.o obj-$(CONFIG_LIBCRC32C) += libcrc32c.o +obj-$(CONFIG_QSORT) += qsort.o obj-$(CONFIG_ZLIB_INFLATE) += zlib_inflate/ obj-$(CONFIG_ZLIB_DEFLATE) += zlib_deflate/ diff -puN /dev/null lib/qsort.c --- /dev/null 2003-09-15 06:40:47.000000000 -0700 +++ 25-akpm/lib/qsort.c 2004-05-10 01:25:36.879404216 -0700 @@ -0,0 +1,239 @@ +/* + * qsort implementation for the Linux kernel. + * + * Original implementation taken form glibc and credited to Douglas + * C. Schmidt (schmidt@ics.uci.edu). + * + * This source code is licensed under the GNU General Public License, + * Version 2. See the file COPYING for more details. + */ + +/* + * If you consider tuning this algorithm, you should consult first: + * Engineering a sort function; Jon Bentley and M. Douglas McIlroy; + * Software - Practice and Experience; Vol. 23 (11), 1249-1265, 1993. + */ + +# include +# include +# include + +MODULE_LICENSE("GPL"); + +/* Byte-wise swap two items of size SIZE. */ +#define SWAP(a, b, size) \ + do { \ + size_t __size = (size); \ + char *__a = (a), *__b = (b); \ + do { \ + char __tmp = *__a; \ + *__a++ = *__b; \ + *__b++ = __tmp; \ + } while (--__size > 0); \ + } while (0) + +/* Discontinue quicksort algorithm when partition gets below this + size. This particular magic number was chosen to work best on a + Sun 4/260. */ +#define MAX_THRESH 4 + +/* Stack node declarations used to store unfulfilled partition + * obligations. */ +typedef struct { + char *lo; + char *hi; +} stack_node; + +/* The next 5 #defines implement a very fast in-line stack + * abstraction. The stack needs log (total_elements) entries (we + * could even subtract log(MAX_THRESH)). Since total_elements has + * type size_t, we get as upper bound for log (total_elements): bits + * per byte (CHAR_BIT) * sizeof(size_t). */ + +#define CHAR_BIT 8 +#define STACK_SIZE (CHAR_BIT * sizeof(size_t)) +#define PUSH(low, high) ((top->lo = (low)), (top->hi = (high)), ++top) +#define POP(low, high) (--top, (low = top->lo), (high = top->hi)) +#define STACK_NOT_EMPTY (stack < top) + +/* Order size using quicksort. This implementation incorporates four + optimizations discussed in Sedgewick: + + 1. Non-recursive, using an explicit stack of pointer that store the + next array partition to sort. To save time, this maximum amount + of space required to store an array of SIZE_MAX is allocated on + the stack. Assuming a 32-bit (64 bit) integer for size_t, this + needs only 32 * sizeof(stack_node) == 256 bytes (for 64 bit: + 1024 bytes). Pretty cheap, actually. + + 2. Chose the pivot element using a median-of-three decision tree. + This reduces the probability of selecting a bad pivot value and + eliminates certain extraneous comparisons. + + 3. Only quicksorts TOTAL_ELEMS / MAX_THRESH partitions, leaving + insertion sort to order the MAX_THRESH items within each + partition. This is a big win, since insertion sort is faster + for small, mostly sorted array segments. + + 4. The larger of the two sub-partitions is always pushed onto the + stack first, with the algorithm then concentrating on the + smaller partition. This *guarantees* no more than log + (total_elems) stack size is needed (actually O(1) in this case)! +*/ + +void +qsort(void *const pbase, size_t total_elems, size_t size, + int (*cmp)(const void*, const void*)) +{ + char *base_ptr = (char *)pbase; + + const size_t max_thresh = MAX_THRESH * size; + + /* Avoid lossage with unsigned arithmetic below. */ + if (total_elems == 0) { + return; + } + + if (total_elems > MAX_THRESH) { + char *lo = base_ptr; + char *hi = &lo[size * (total_elems - 1)]; + stack_node stack[STACK_SIZE]; + stack_node *top = stack + 1; + + while (STACK_NOT_EMPTY) { + char *left_ptr; + char *right_ptr; + + /* Select median value from among LO, MID, and + HI. Rearrange LO and HI so the three values + are sorted. This lowers the probability of + picking a pathological pivot value and + skips a comparison for both the LEFT_PTR + and RIGHT_PTR in the while loops. */ + + char *mid = lo + size * ((hi - lo) / size >> 1); + + if ((*cmp)((void*)mid, (void*)lo) < 0) + SWAP(mid, lo, size); + if ((*cmp)((void*)hi, (void*)mid) < 0) + SWAP(mid, hi, size); + else + goto jump_over; + if ((*cmp)((void*)mid, (void*)lo) < 0) + SWAP(mid, lo, size); + jump_over: + + left_ptr = lo + size; + right_ptr = hi - size; + + /* Here's the famous ``collapse the walls'' + section of quicksort. Gotta like those + tight inner loops! They are the main + reason that this algorithm runs much faster + than others. */ + do { + while ((*cmp)((void*)left_ptr, (void*)mid) < 0) + left_ptr += size; + + while ((*cmp)((void*)mid, (void*)right_ptr) < 0) + right_ptr -= size; + + if (left_ptr < right_ptr) { + SWAP(left_ptr, right_ptr, size); + if (mid == left_ptr) + mid = right_ptr; + else if (mid == right_ptr) + mid = left_ptr; + left_ptr += size; + right_ptr -= size; + } else if (left_ptr == right_ptr) { + left_ptr += size; + right_ptr -= size; + break; + } + } + while (left_ptr <= right_ptr); + + /* Set up pointers for next iteration. First + determine whether left and right partitions + are below the threshold size. If so, + ignore one or both. Otherwise, push the + larger partition's bounds on the stack and + continue sorting the smaller one. */ + + if ((size_t) (right_ptr - lo) <= max_thresh) { + if ((size_t) (hi - left_ptr) <= max_thresh) + /* Ignore both small partitions. */ + POP(lo, hi); + else + /* Ignore small left partition. */ + lo = left_ptr; + } else if ((size_t) (hi - left_ptr) <= max_thresh) + /* Ignore small right partition. */ + hi = right_ptr; + else if ((right_ptr - lo) > (hi - left_ptr)) { + /* Push larger left partition indices. */ + PUSH(lo, right_ptr); + lo = left_ptr; + } else { + /* Push larger right partition indices. */ + PUSH(left_ptr, hi); + hi = right_ptr; + } + } + } + + /* Once the BASE_PTR array is partially sorted by quicksort + the rest is completely sorted using insertion sort, since + this is efficient for partitions below MAX_THRESH + size. BASE_PTR points to the beginning of the array to + sort, and END_PTR points at the very last element in the + array (*not* one beyond it!). */ + + { + char *end_ptr = &base_ptr[size * (total_elems - 1)]; + char *tmp_ptr = base_ptr; + char *thresh = min(end_ptr, base_ptr + max_thresh); + char *run_ptr; + + /* Find smallest element in first threshold and place + it at the array's beginning. This is the smallest + array element, and the operation speeds up + insertion sort's inner loop. */ + + for (run_ptr = tmp_ptr + size; run_ptr <= thresh; + run_ptr += size) + if ((*cmp)((void*)run_ptr, (void*)tmp_ptr) < 0) + tmp_ptr = run_ptr; + + if (tmp_ptr != base_ptr) + SWAP(tmp_ptr, base_ptr, size); + + /* Insertion sort, running from left-hand-side up to + * right-hand-side. */ + + run_ptr = base_ptr + size; + while ((run_ptr += size) <= end_ptr) { + tmp_ptr = run_ptr - size; + while ((*cmp)((void*)run_ptr, (void*)tmp_ptr) < 0) + tmp_ptr -= size; + + tmp_ptr += size; + if (tmp_ptr != run_ptr) { + char *trav; + + trav = run_ptr + size; + while (--trav >= run_ptr) { + char c = *trav; + char *hi, *lo; + + for (hi = lo = trav; + (lo -= size) >= tmp_ptr; hi = lo) + *hi = *lo; + *hi = c; + } + } + } + } +} +EXPORT_SYMBOL_GPL(qsort); _