From 5278f6def4248391a481382bd7b0042b04c829f2 Mon Sep 17 00:00:00 2001
From: Richard Qin <richard.qin.x@gmail.com>
Date: Tue, 21 Apr 2026 23:16:50 +0800
Subject: feat: add string manipulation functions derived from musl libc

This commit introduces a variety of string manipulation functions, including:
- `memmem`: Searches for a substring in a memory area.
- `memmove`: Safely copies memory areas, handling overlaps.
- `mempcpy`: Copies memory and returns a pointer to the end of the destination.
- `memrchr`: Finds the last occurrence of a character in a memory area.
- `memset`: Fills a memory area with a specified value.
- `rindex`: Returns the last occurrence of a character in a string.
- `stpcpy` and `stpncpy`: Copies strings and returns a pointer to the end.
- `strcat`, `strchr`, `strcspn`, `strlcat`, `strlcpy`, `strlen`, `strncat`, `strncmp`, `strncpy`, `strnlen`, `strpbrk`, `strrchr`, `strstr`, `strtok`, and `strtok_r`: Various string manipulation functions for concatenation, searching, and tokenization.

Assembly implementations for `memcpy`, `memmove`, and `memset` are also included for optimized performance on x86_64 architectures.
---
 na/lib/string/memcpy.c | 127 +++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 127 insertions(+)
 create mode 100644 na/lib/string/memcpy.c

(limited to 'na/lib/string/memcpy.c')

diff --git a/na/lib/string/memcpy.c b/na/lib/string/memcpy.c
new file mode 100644
index 0000000..de26a00
--- /dev/null
+++ b/na/lib/string/memcpy.c
@@ -0,0 +1,127 @@
+/* SPDX-License-Identifier: MIT */
+/* Derived from musl libc (Copyright © 2005-2020 Rich Felker, et al.) */
+
+#include <string.h>
+#include <stdint.h>
+#include <endian.h>
+
+void *memcpy(void *restrict dest, const void *restrict src, size_t n)
+{
+	unsigned char *d = dest;
+	const unsigned char *s = src;
+
+#ifdef __GNUC__
+
+#if __BYTE_ORDER == __LITTLE_ENDIAN
+#define LS >>
+#define RS <<
+#else
+#define LS <<
+#define RS >>
+#endif
+
+	typedef uint32_t __attribute__((__may_alias__)) u32;
+	uint32_t w, x;
+
+	for (; (uintptr_t)s % 4 && n; n--) *d++ = *s++;
+
+	if ((uintptr_t)d % 4 == 0) {
+		for (; n>=16; s+=16, d+=16, n-=16) {
+			*(u32 *)(d+0) = *(u32 *)(s+0);
+			*(u32 *)(d+4) = *(u32 *)(s+4);
+			*(u32 *)(d+8) = *(u32 *)(s+8);
+			*(u32 *)(d+12) = *(u32 *)(s+12);
+		}
+		if (n&8) {
+			*(u32 *)(d+0) = *(u32 *)(s+0);
+			*(u32 *)(d+4) = *(u32 *)(s+4);
+			d += 8; s += 8;
+		}
+		if (n&4) {
+			*(u32 *)(d+0) = *(u32 *)(s+0);
+			d += 4; s += 4;
+		}
+		if (n&2) {
+			*d++ = *s++; *d++ = *s++;
+		}
+		if (n&1) {
+			*d = *s;
+		}
+		return dest;
+	}
+
+	if (n >= 32) switch ((uintptr_t)d % 4) {
+	case 1:
+		w = *(u32 *)s;
+		*d++ = *s++;
+		*d++ = *s++;
+		*d++ = *s++;
+		n -= 3;
+		for (; n>=17; s+=16, d+=16, n-=16) {
+			x = *(u32 *)(s+1);
+			*(u32 *)(d+0) = (w LS 24) | (x RS 8);
+			w = *(u32 *)(s+5);
+			*(u32 *)(d+4) = (x LS 24) | (w RS 8);
+			x = *(u32 *)(s+9);
+			*(u32 *)(d+8) = (w LS 24) | (x RS 8);
+			w = *(u32 *)(s+13);
+			*(u32 *)(d+12) = (x LS 24) | (w RS 8);
+		}
+		break;
+	case 2:
+		w = *(u32 *)s;
+		*d++ = *s++;
+		*d++ = *s++;
+		n -= 2;
+		for (; n>=18; s+=16, d+=16, n-=16) {
+			x = *(u32 *)(s+2);
+			*(u32 *)(d+0) = (w LS 16) | (x RS 16);
+			w = *(u32 *)(s+6);
+			*(u32 *)(d+4) = (x LS 16) | (w RS 16);
+			x = *(u32 *)(s+10);
+			*(u32 *)(d+8) = (w LS 16) | (x RS 16);
+			w = *(u32 *)(s+14);
+			*(u32 *)(d+12) = (x LS 16) | (w RS 16);
+		}
+		break;
+	case 3:
+		w = *(u32 *)s;
+		*d++ = *s++;
+		n -= 1;
+		for (; n>=19; s+=16, d+=16, n-=16) {
+			x = *(u32 *)(s+3);
+			*(u32 *)(d+0) = (w LS 8) | (x RS 24);
+			w = *(u32 *)(s+7);
+			*(u32 *)(d+4) = (x LS 8) | (w RS 24);
+			x = *(u32 *)(s+11);
+			*(u32 *)(d+8) = (w LS 8) | (x RS 24);
+			w = *(u32 *)(s+15);
+			*(u32 *)(d+12) = (x LS 8) | (w RS 24);
+		}
+		break;
+	}
+	if (n&16) {
+		*d++ = *s++; *d++ = *s++; *d++ = *s++; *d++ = *s++;
+		*d++ = *s++; *d++ = *s++; *d++ = *s++; *d++ = *s++;
+		*d++ = *s++; *d++ = *s++; *d++ = *s++; *d++ = *s++;
+		*d++ = *s++; *d++ = *s++; *d++ = *s++; *d++ = *s++;
+	}
+	if (n&8) {
+		*d++ = *s++; *d++ = *s++; *d++ = *s++; *d++ = *s++;
+		*d++ = *s++; *d++ = *s++; *d++ = *s++; *d++ = *s++;
+	}
+	if (n&4) {
+		*d++ = *s++; *d++ = *s++; *d++ = *s++; *d++ = *s++;
+	}
+	if (n&2) {
+		*d++ = *s++; *d++ = *s++;
+	}
+	if (n&1) {
+		*d = *s;
+	}
+	return dest;
+#endif
+
+	for (; n; n--) *d++ = *s++;
+	return dest;
+}
-- 
cgit v1.2.3