Improved scrolling and added better memcpy/memmove functions written in assembly.

targets/x86_64/src/cstdlib/memory.s

@@ -3,25 +3,64 @@
 .global memcpy
 .type memcpy @function
 
-// int memcpy(void* dest [%rdi], void* src [%rsi], size_t count [%rdx])
+// void* memcpy(void* dest [%rdi], void* src [%rsi], size_t count [%rdx])
 //
 memcpy:
-    cmpq %rdx, 4
-    jl memcpy_bytes
-    movq (%rsi), %rax
-    movq %rax, (%rdi)
-    addq $4, %rsi
-    addq $4, %rdi
-    subq $4, %rdx
-    jmp memcpy
+    movq %rdi, %rax // return value
+memcpy_qwords:
+    cmpq $8, %rdx      // check if the remaining bytes are at least 8
+    jl memcpy_bytes    // if not copy the remaining bytes per byte
+    movq (%rsi), %rcx  // if yes, copy using movq
+    movq %rcx, (%rdi)
+    addq $8, %rsi
+    addq $8, %rdi
+    subq $8, %rdx
+    jmp memcpy_qwords
 memcpy_bytes:
-    cmpq %rdx, 0
+    cmpq $0, %rdx
     je memcpy_end
-    movb (%rsi), %ah
-    movb %ah, (%rdi)
+    movb (%rsi), %ch
+    movb %ch, (%rdi)
     addq $1, %rsi
     addq $1, %rdi
     subq $1, %rdx
     jmp memcpy_bytes
 memcpy_end:
     ret
+
+.global memmove
+.type memmove @function
+// void* memmove(void* dest [%rdi], void* src [%rsi], size_t count [%rdx])
+//
+memmove:
+    movq %rdi, %rax     // preserve dest as return value
+    cmpq %rsi, %rdi     // check if dest > src
+    jg memmove_backward // if yes, do everything backwards
+    movq %rsi, %rcx     // check if (src - dest) < 8
+    subq %rdi, %rcx
+    cmpq $8, %rdi
+    jl memcpy_bytes     // if yes, we have to do a bytewise copy
+    jmp memcpy_qwords   // otherwise copy whole qwords
+memmove_backward:       // dest > src, copy backwards
+    addq %rdx, %rdi     // dest = dest + count
+    addq %rdx, %rsi     // src = src + count
+memmove_qwords:
+    cmpq $8, %rdx        // check if the remaining bytes are at least 8
+    jl memmove_bytes    // if not copy the remaining bytes per byte
+    movq -8(%rsi), %rcx // if yes, copy using movq
+    movq %rcx, -8(%rdi)
+    subq $8, %rsi
+    subq $8, %rdi
+    subq $8, %rdx
+    jmp memmove_qwords
+memmove_bytes:
+    cmpq $0, %rdx
+    je memmove_end
+    movb (%rsi), %ch
+    movb %ch, (%rdi)
+    subq $1, %rsi
+    subq $1, %rdi
+    subq $1, %rdx
+    jmp memmove_bytes
+memmove_end:
+    ret