LibC: Implement some mem* and str* functions in assembly

This made them a lot faster on modern cpus with optimized rep stos and
rep movs

kernel/CMakeLists.txt

@@ -148,6 +148,10 @@ set(BAN_SOURCES
 set(KLIBC_SOURCES
 	klibc/ctype.cpp
 	klibc/string.cpp
+
+	# Ehhh don't do this but for now libc uses the same stuff kernel can use
+	# This won't work after libc starts using sse implemetations tho
+	../userspace/libraries/LibC/arch/${BANAN_ARCH}/string.S
 )
 
 set(LIBFONT_SOURCES

userspace/libraries/LibC/CMakeLists.txt

@@ -45,6 +45,7 @@ set(LIBC_SOURCES
 	icxxabi.cpp
 
 	arch/${BANAN_ARCH}/setjmp.S
+	arch/${BANAN_ARCH}/string.S
 
 	../../../BAN/BAN/Assert.cpp
 )

userspace/libraries/LibC/arch/i686/string.S

@@ -0,0 +1,85 @@
+.global memchr
+memchr:
+	xchgl 4(%esp), %edi
+	movl  8(%esp), %eax
+	movl 12(%esp), %ecx
+	repne scasb
+	xorl %eax, %eax
+	testl %ecx, %ecx
+	cmovnzl %edi, %eax
+	movl 4(%esp), %edi
+	ret
+
+.global memcmp
+memcmp:
+	xchgl 4(%esp), %edi
+	xchgl 8(%esp), %esi
+	movl 12(%esp), %ecx
+	repe cmpsb
+	jne .memcmp_not_equal
+	xorl %eax, %eax
+	jmp .memcmp_done
+ .memcmp_not_equal:
+	movzbl -1(%edi), %eax
+	movzbl -1(%esi), %ecx
+	subl %ecx, %eax
+ .memcmp_done:
+	movl 4(%esp), %edi
+	movl 8(%esp), %esi
+	ret
+
+.global memcpy
+memcpy:
+	xchgl 4(%esp), %edi
+	xchgl 8(%esp), %esi
+	movl 12(%esp), %ecx
+	movl %edi, %edx
+	rep movsb
+	movl 4(%esp), %edi
+	movl 8(%esp), %esi
+	movl %edx, %eax
+	ret
+
+.global memmove
+memmove:
+	xchgl 4(%esp), %edi
+	xchgl 8(%esp), %esi
+	movl 12(%esp), %ecx
+	movl %edi, %edx
+	cmpl %edi, %esi
+	jb .memmove_slow
+	rep movsb
+ .memmove_done:
+	movl 4(%esp), %edi
+	movl 8(%esp), %esi
+	movl %edx, %eax
+	ret
+ .memmove_slow:
+	leal -1(%edi, %ecx), %edi
+	leal -1(%esi, %ecx), %esi
+	std
+	rep movsb
+	cld
+	jmp .memmove_done
+
+.global memset
+memset:
+	xchgl 4(%esp), %edi
+	movl  8(%esp), %eax
+	movl 12(%esp), %ecx
+	movl %edi, %edx
+	rep stosb
+	movl 4(%esp), %edi
+	movl %edx, %eax
+	ret
+
+.global strlen
+strlen:
+	xchgl 4(%esp), %edi
+	xorb %al, %al
+	movl $-1, %ecx
+	repne scasb
+	movl 4(%esp), %edi
+	movl $-2, %eax
+	subl %ecx, %eax
+	ret

userspace/libraries/LibC/arch/x86_64/string.S

@@ -0,0 +1,62 @@
+.global memchr
+memchr:
+	movb %sil, %al
+	movq %rdx, %rcx
+	repne scasb
+	xorq %rax, %rax
+	testq %rcx, %rcx
+	cmovnzq %rdi, %rax
+	ret
+
+.global memcmp
+memcmp:
+	movq %rdx, %rcx
+	repe cmpsb
+	jne .memcmp_not_equal
+	xorq %rax, %rax
+	ret
+ .memcmp_not_equal:
+	movzbl -1(%rdi), %eax
+	movzbl -1(%rsi), %ecx
+	subq %rcx, %rax
+	ret
+
+.global memcpy
+memcpy:
+	movq %rdi, %rax
+	movq %rdx, %rcx
+	movq %rdi, %rdx
+	rep movsb
+	movq %rdx, %rax
+	ret
+
+.global memmove
+memmove:
+	cmpq %rdi, %rsi
+	jae memcpy
+	leaq -1(%rdi, %rdx), %rdi
+	leaq -1(%rsi, %rdx), %rsi
+	movq %rdx, %rcx
+	std
+	rep movsb
+	cld
+	leaq 1(%rdi), %rax
+	ret
+
+.global memset
+memset:
+	movq %rdi, %r8
+	movb %sil, %al
+	movq %rdx, %rcx
+	rep stosb
+	movq %r8, %rax
+	ret
+
+.global strlen
+strlen:
+	xorb %al, %al
+	movq $-1, %rcx
+	repne scasb
+	movq $-2, %rax
+	subq %rcx, %rax
+	ret