libc: optimized memcpy()

If the cpu supports "Enhanced REP MOVS / STOS" (ERMS), use an rep movsb
instruction to implement memcpy.  This speeds up copies significantly,
especially large misaligned ones.

arch/x64/string.cc

+
+#include <string.h>
+#include "processor.hh"
+
+extern "C"
+void *memcpy_base(void *__restrict dest, const void *__restrict src, size_t n);
+
+extern "C"
+void *memcpy_repmov(void *__restrict dest, const void *__restrict src, size_t n)
+{
+    auto ret = dest;
+    asm volatile("rep movsb" : "+D"(dest), "+S"(src), "+c"(n) : : "memory");
+    return ret;
+}
+
+extern "C"
+void *(*resolve_memcpy())(void *__restrict dest, const void *__restrict src, size_t n)
+{
+    using processor::cpuid;
+    if (cpuid(0).a >= 7 && (cpuid(7, 0).b & 0x200)) {
+        return memcpy_repmov;
+    }
+    return memcpy_base;
+}
+
+void *memcpy(void *__restrict dest, const void *__restrict src, size_t n)
+    __attribute__((ifunc("resolve_memcpy")));

build.mak

@@ -409,6 +409,7 @@ objects += arch/x64/arch-setup.o
 objects += arch/x64/smp.o
 objects += arch/x64/signal.o
 objects += arch/x64/cpuid.o
+objects += arch/x64/string.o
 objects += core/mutex.o
 objects += core/lfmutex.o
 objects += core/semaphore.o

libc/string/memcpy.c

@@ -6,7 +6,7 @@
 #define ALIGN (sizeof(size_t)-1)
 #define ONES ((size_t)-1/UCHAR_MAX)
 
-void *memcpy(void *__restrict dest, const void *__restrict src, size_t n)
+void *memcpy_base(void *__restrict dest, const void *__restrict src, size_t n)
 {
 	unsigned char *d = dest;
 	const unsigned char *s = src;