^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1) // SPDX-License-Identifier: GPL-2.0
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3) * A fast, small, non-recursive O(n log n) sort for the Linux kernel
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5) * This performs n*log2(n) + 0.37*n + o(n) comparisons on average,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6) * and 1.5*n*log2(n) + O(n) in the (very contrived) worst case.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8) * Glibc qsort() manages n*log2(n) - 1.26*n for random inputs (1.63*n
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9) * better) at the expense of stack usage and much larger code to avoid
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10) * quicksort's O(n^2) worst case.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 12)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 13) #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 14)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 15) #include <linux/types.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 16) #include <linux/export.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 17) #include <linux/sort.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 18)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 19) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 20) * is_aligned - is this pointer & size okay for word-wide copying?
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 21) * @base: pointer to data
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 22) * @size: size of each element
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 23) * @align: required alignment (typically 4 or 8)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 24) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 25) * Returns true if elements can be copied using word loads and stores.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 26) * The size must be a multiple of the alignment, and the base address must
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 27) * be if we do not have CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 28) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 29) * For some reason, gcc doesn't know to optimize "if (a & mask || b & mask)"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 30) * to "if ((a | b) & mask)", so we do that by hand.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 31) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 32) __attribute_const__ __always_inline
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 33) static bool is_aligned(const void *base, size_t size, unsigned char align)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 34) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 35) unsigned char lsbits = (unsigned char)size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 36)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 37) (void)base;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 38) #ifndef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 39) lsbits |= (unsigned char)(uintptr_t)base;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 40) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 41) return (lsbits & (align - 1)) == 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 42) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 43)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 44) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 45) * swap_words_32 - swap two elements in 32-bit chunks
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 46) * @a: pointer to the first element to swap
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 47) * @b: pointer to the second element to swap
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 48) * @n: element size (must be a multiple of 4)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 49) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 50) * Exchange the two objects in memory. This exploits base+index addressing,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 51) * which basically all CPUs have, to minimize loop overhead computations.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 52) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 53) * For some reason, on x86 gcc 7.3.0 adds a redundant test of n at the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 54) * bottom of the loop, even though the zero flag is stil valid from the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 55) * subtract (since the intervening mov instructions don't alter the flags).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 56) * Gcc 8.1.0 doesn't have that problem.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 57) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 58) static void swap_words_32(void *a, void *b, size_t n)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 59) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 60) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 61) u32 t = *(u32 *)(a + (n -= 4));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 62) *(u32 *)(a + n) = *(u32 *)(b + n);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 63) *(u32 *)(b + n) = t;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 64) } while (n);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 65) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 66)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 67) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 68) * swap_words_64 - swap two elements in 64-bit chunks
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 69) * @a: pointer to the first element to swap
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 70) * @b: pointer to the second element to swap
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 71) * @n: element size (must be a multiple of 8)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 72) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 73) * Exchange the two objects in memory. This exploits base+index
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 74) * addressing, which basically all CPUs have, to minimize loop overhead
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 75) * computations.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 76) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 77) * We'd like to use 64-bit loads if possible. If they're not, emulating
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 78) * one requires base+index+4 addressing which x86 has but most other
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 79) * processors do not. If CONFIG_64BIT, we definitely have 64-bit loads,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 80) * but it's possible to have 64-bit loads without 64-bit pointers (e.g.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 81) * x32 ABI). Are there any cases the kernel needs to worry about?
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 82) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 83) static void swap_words_64(void *a, void *b, size_t n)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 84) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 85) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 86) #ifdef CONFIG_64BIT
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 87) u64 t = *(u64 *)(a + (n -= 8));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 88) *(u64 *)(a + n) = *(u64 *)(b + n);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 89) *(u64 *)(b + n) = t;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 90) #else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 91) /* Use two 32-bit transfers to avoid base+index+4 addressing */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 92) u32 t = *(u32 *)(a + (n -= 4));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 93) *(u32 *)(a + n) = *(u32 *)(b + n);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 94) *(u32 *)(b + n) = t;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 95)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 96) t = *(u32 *)(a + (n -= 4));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 97) *(u32 *)(a + n) = *(u32 *)(b + n);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 98) *(u32 *)(b + n) = t;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 99) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 100) } while (n);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 101) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 102)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 103) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 104) * swap_bytes - swap two elements a byte at a time
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 105) * @a: pointer to the first element to swap
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 106) * @b: pointer to the second element to swap
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 107) * @n: element size
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 108) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 109) * This is the fallback if alignment doesn't allow using larger chunks.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 110) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 111) static void swap_bytes(void *a, void *b, size_t n)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 112) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 113) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 114) char t = ((char *)a)[--n];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 115) ((char *)a)[n] = ((char *)b)[n];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 116) ((char *)b)[n] = t;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 117) } while (n);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 118) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 119)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 120) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 121) * The values are arbitrary as long as they can't be confused with
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 122) * a pointer, but small integers make for the smallest compare
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 123) * instructions.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 124) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 125) #define SWAP_WORDS_64 (swap_func_t)0
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 126) #define SWAP_WORDS_32 (swap_func_t)1
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 127) #define SWAP_BYTES (swap_func_t)2
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 128)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 129) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 130) * The function pointer is last to make tail calls most efficient if the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 131) * compiler decides not to inline this function.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 132) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 133) static void do_swap(void *a, void *b, size_t size, swap_func_t swap_func)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 134) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 135) if (swap_func == SWAP_WORDS_64)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 136) swap_words_64(a, b, size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 137) else if (swap_func == SWAP_WORDS_32)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 138) swap_words_32(a, b, size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 139) else if (swap_func == SWAP_BYTES)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 140) swap_bytes(a, b, size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 141) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 142) swap_func(a, b, (int)size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 143) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 144)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 145) #define _CMP_WRAPPER ((cmp_r_func_t)0L)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 146)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 147) static int do_cmp(const void *a, const void *b, cmp_r_func_t cmp, const void *priv)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 148) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 149) if (cmp == _CMP_WRAPPER)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 150) return ((cmp_func_t)(priv))(a, b);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 151) return cmp(a, b, priv);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 152) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 153)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 154) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 155) * parent - given the offset of the child, find the offset of the parent.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 156) * @i: the offset of the heap element whose parent is sought. Non-zero.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 157) * @lsbit: a precomputed 1-bit mask, equal to "size & -size"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 158) * @size: size of each element
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 159) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 160) * In terms of array indexes, the parent of element j = @i/@size is simply
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 161) * (j-1)/2. But when working in byte offsets, we can't use implicit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 162) * truncation of integer divides.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 163) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 164) * Fortunately, we only need one bit of the quotient, not the full divide.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 165) * @size has a least significant bit. That bit will be clear if @i is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 166) * an even multiple of @size, and set if it's an odd multiple.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 167) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 168) * Logically, we're doing "if (i & lsbit) i -= size;", but since the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 169) * branch is unpredictable, it's done with a bit of clever branch-free
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 170) * code instead.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 171) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 172) __attribute_const__ __always_inline
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 173) static size_t parent(size_t i, unsigned int lsbit, size_t size)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 174) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 175) i -= size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 176) i -= size & -(i & lsbit);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 177) return i / 2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 178) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 179)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 180) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 181) * sort_r - sort an array of elements
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 182) * @base: pointer to data to sort
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 183) * @num: number of elements
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 184) * @size: size of each element
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 185) * @cmp_func: pointer to comparison function
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 186) * @swap_func: pointer to swap function or NULL
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 187) * @priv: third argument passed to comparison function
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 188) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 189) * This function does a heapsort on the given array. You may provide
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 190) * a swap_func function if you need to do something more than a memory
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 191) * copy (e.g. fix up pointers or auxiliary data), but the built-in swap
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 192) * avoids a slow retpoline and so is significantly faster.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 193) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 194) * Sorting time is O(n log n) both on average and worst-case. While
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 195) * quicksort is slightly faster on average, it suffers from exploitable
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 196) * O(n*n) worst-case behavior and extra memory requirements that make
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 197) * it less suitable for kernel use.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 198) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 199) void sort_r(void *base, size_t num, size_t size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 200) cmp_r_func_t cmp_func,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 201) swap_func_t swap_func,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 202) const void *priv)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 203) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 204) /* pre-scale counters for performance */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 205) size_t n = num * size, a = (num/2) * size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 206) const unsigned int lsbit = size & -size; /* Used to find parent */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 207)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 208) if (!a) /* num < 2 || size == 0 */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 209) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 210)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 211) if (!swap_func) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 212) if (is_aligned(base, size, 8))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 213) swap_func = SWAP_WORDS_64;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 214) else if (is_aligned(base, size, 4))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 215) swap_func = SWAP_WORDS_32;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 216) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 217) swap_func = SWAP_BYTES;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 218) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 219)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 220) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 221) * Loop invariants:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 222) * 1. elements [a,n) satisfy the heap property (compare greater than
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 223) * all of their children),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 224) * 2. elements [n,num*size) are sorted, and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 225) * 3. a <= b <= c <= d <= n (whenever they are valid).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 226) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 227) for (;;) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 228) size_t b, c, d;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 229)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 230) if (a) /* Building heap: sift down --a */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 231) a -= size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 232) else if (n -= size) /* Sorting: Extract root to --n */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 233) do_swap(base, base + n, size, swap_func);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 234) else /* Sort complete */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 235) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 236)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 237) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 238) * Sift element at "a" down into heap. This is the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 239) * "bottom-up" variant, which significantly reduces
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 240) * calls to cmp_func(): we find the sift-down path all
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 241) * the way to the leaves (one compare per level), then
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 242) * backtrack to find where to insert the target element.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 243) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 244) * Because elements tend to sift down close to the leaves,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 245) * this uses fewer compares than doing two per level
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 246) * on the way down. (A bit more than half as many on
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 247) * average, 3/4 worst-case.)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 248) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 249) for (b = a; c = 2*b + size, (d = c + size) < n;)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 250) b = do_cmp(base + c, base + d, cmp_func, priv) >= 0 ? c : d;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 251) if (d == n) /* Special case last leaf with no sibling */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 252) b = c;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 253)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 254) /* Now backtrack from "b" to the correct location for "a" */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 255) while (b != a && do_cmp(base + a, base + b, cmp_func, priv) >= 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 256) b = parent(b, lsbit, size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 257) c = b; /* Where "a" belongs */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 258) while (b != a) { /* Shift it into place */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 259) b = parent(b, lsbit, size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 260) do_swap(base + b, base + c, size, swap_func);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 261) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 262) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 263) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 264) EXPORT_SYMBOL(sort_r);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 265)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 266) void sort(void *base, size_t num, size_t size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 267) cmp_func_t cmp_func,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 268) swap_func_t swap_func)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 269) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 270) return sort_r(base, num, size, _CMP_WRAPPER, swap_func, cmp_func);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 271) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 272) EXPORT_SYMBOL(sort);
Orange Pi5 kernel
Deprecated Linux kernel 5.10.110 for OrangePi 5/5B/5+ boards
3 Commits
0 Branches
0 Tags