From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: by sourceware.org (Postfix, from userid 1791) id 9D8533858000; Tue, 31 Oct 2023 17:24:00 +0000 (GMT) DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 9D8533858000 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=sourceware.org; s=default; t=1698773040; bh=PFhN2iisGsrLi2YeyHNjYvH6Iq4PG2lxalD+jmUYg0c=; h=From:To:Subject:Date:From; b=LRTZfgOGUEHB/vbpzA9RwmBiUtilBpEpXmYAWWiEcTB7AcJkcy1q/YVzaTHZ21fbs hA6Fc72vr6Z9SgrY7W0ojGjsL/AP2cG4KrrMZdwJx1rXTwKtTnERLuYdM1RxiGRtEG CThs6r06hdt0SgH6gHwNy3hwEA0fkD5dMB7jAPi8= Content-Type: text/plain; charset="us-ascii" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit From: Adhemerval Zanella To: glibc-cvs@sourceware.org Subject: [glibc] stdlib: Implement introsort for qsort (BZ 19305) X-Act-Checkin: glibc X-Git-Author: Adhemerval Zanella X-Git-Refname: refs/heads/master X-Git-Oldrev: d097f3c79be55d646d86efb7ce876bf84d5ebe4e X-Git-Newrev: 274a46c9b25ab733a1fb9fb1497f1beecae30193 Message-Id: <20231031172400.9D8533858000@sourceware.org> Date: Tue, 31 Oct 2023 17:24:00 +0000 (GMT) List-Id: https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=274a46c9b25ab733a1fb9fb1497f1beecae30193 commit 274a46c9b25ab733a1fb9fb1497f1beecae30193 Author: Adhemerval Zanella Date: Tue Oct 3 09:22:49 2023 -0300 stdlib: Implement introsort for qsort (BZ 19305) This patch makes the quicksort implementation to acts as introsort, to avoid worse-case performance (and thus making it O(nlog n)). It switch to heapsort when the depth level reaches 2*log2(total elements). The heapsort is a textbook implementation. Checked on x86_64-linux-gnu and aarch64-linux-gnu. Reviewed-by: Noah Goldstein Diff: --- stdlib/qsort.c | 89 +++++++++++++++++++++++++++++++++++++++++++++++++++++----- 1 file changed, 82 insertions(+), 7 deletions(-) diff --git a/stdlib/qsort.c b/stdlib/qsort.c index 80706b3357..d5f205affc 100644 --- a/stdlib/qsort.c +++ b/stdlib/qsort.c @@ -98,6 +98,7 @@ typedef struct { char *lo; char *hi; + size_t depth; } stack_node; /* The stack needs log (total_elements) entries (we could even subtract @@ -107,22 +108,85 @@ typedef struct enum { STACK_SIZE = CHAR_BIT * sizeof (size_t) }; static inline stack_node * -push (stack_node *top, char *lo, char *hi) +push (stack_node *top, char *lo, char *hi, size_t depth) { top->lo = lo; top->hi = hi; + top->depth = depth; return ++top; } static inline stack_node * -pop (stack_node *top, char **lo, char **hi) +pop (stack_node *top, char **lo, char **hi, size_t *depth) { --top; *lo = top->lo; *hi = top->hi; + *depth = top->depth; return top; } +/* NB: N is inclusive bound for BASE. */ +static inline void +siftdown (void *base, size_t size, size_t k, size_t n, + enum swap_type_t swap_type, __compar_d_fn_t cmp, void *arg) +{ + while (k <= n / 2) + { + size_t j = 2 * k; + if (j < n && cmp (base + (j * size), base + ((j + 1) * size), arg) < 0) + j++; + + if (cmp (base + (k * size), base + (j * size), arg) >= 0) + break; + + do_swap (base + (size * j), base + (k * size), size, swap_type); + k = j; + } +} + +static inline void +heapify (void *base, size_t size, size_t n, enum swap_type_t swap_type, + __compar_d_fn_t cmp, void *arg) +{ + size_t k = n / 2; + while (1) + { + siftdown (base, size, k, n, swap_type, cmp, arg); + if (k-- == 0) + break; + } +} + +/* A non-recursive heapsort, used on introsort implementation as a fallback + routine with worst-case performance of O(nlog n) and worst-case space + complexity of O(1). It sorts the array starting at BASE and ending at + END, with each element of SIZE bytes. The SWAP_TYPE is the callback + function used to swap elements, and CMP is the function used to compare + elements. */ +static void +heapsort_r (void *base, void *end, size_t size, enum swap_type_t swap_type, + __compar_d_fn_t cmp, void *arg) +{ + const size_t count = ((uintptr_t) end - (uintptr_t) base) / size; + + if (count < 2) + return; + + size_t n = count - 1; + + /* Build the binary heap, largest value at the base[0]. */ + heapify (base, size, n, swap_type, cmp, arg); + + /* On each iteration base[0:n] is the binary heap, while base[n:count] + is sorted. */ + while (n > 0) + { + do_swap (base, base + (n * size), size, swap_type); + n--; + siftdown (base, size, 0, n, swap_type, cmp, arg); + } +} static inline void insertion_sort_qsort_partitions (void *const pbase, size_t total_elems, @@ -208,7 +272,7 @@ _quicksort (void *const pbase, size_t total_elems, size_t size, const size_t max_thresh = MAX_THRESH * size; - if (total_elems == 0) + if (total_elems <= 1) /* Avoid lossage with unsigned arithmetic below. */ return; @@ -220,15 +284,26 @@ _quicksort (void *const pbase, size_t total_elems, size_t size, else swap_type = SWAP_BYTES; + /* Maximum depth before quicksort switches to heapsort. */ + size_t depth = 2 * (sizeof (size_t) * CHAR_BIT - 1 + - __builtin_clzl (total_elems)); + 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; + stack_node *top = push (stack, NULL, NULL, depth); while (stack < top) { + if (depth == 0) + { + heapsort_r (lo, hi, size, swap_type, cmp, arg); + top = pop (top, &lo, &hi, &depth); + continue; + } + char *left_ptr; char *right_ptr; @@ -292,7 +367,7 @@ _quicksort (void *const pbase, size_t total_elems, size_t size, { if ((size_t) (hi - left_ptr) <= max_thresh) /* Ignore both small partitions. */ - top = pop (top, &lo, &hi); + top = pop (top, &lo, &hi, &depth); else /* Ignore small left partition. */ lo = left_ptr; @@ -303,13 +378,13 @@ _quicksort (void *const pbase, size_t total_elems, size_t size, else if ((right_ptr - lo) > (hi - left_ptr)) { /* Push larger left partition indices. */ - top = push (top, lo, right_ptr); + top = push (top, lo, right_ptr, depth - 1); lo = left_ptr; } else { /* Push larger right partition indices. */ - top = push (top, left_ptr, hi); + top = push (top, left_ptr, hi, depth - 1); hi = right_ptr; } }