skip optimized memcpy/memset for small builds

This saves 100 bytes on trinket_m0 at the price of making many memset &
memcpy calls slower.

Author: jepler

shared/libc/string0.c

@@ -28,13 +28,16 @@
 #include <stddef.h>
 #include <string.h>
 
+#include "py/mpconfig.h"
+
 #ifndef likely
 #define likely(x) __builtin_expect((x), 1)
 #endif
 
 #pragma GCC diagnostic push
 #pragma GCC diagnostic ignored "-Wcast-align"
 void *memcpy(void *dst, const void *src, size_t n) {
+#if CIRCUITPY_FULL_BUILD
     if (likely(!(((uintptr_t)dst) & 3) && !(((uintptr_t)src) & 3))) {
         // pointers aligned
         uint32_t *d = dst;
@@ -56,7 +59,9 @@ void *memcpy(void *dst, const void *src, size_t n) {
             // copy byte
             *((uint8_t*)d) = *((const uint8_t*)s);
         }
-    } else {
+    } else
+#endif
+    {
         // unaligned access, copy bytes
         uint8_t *d = dst;
         const uint8_t *s = src;
@@ -93,6 +98,7 @@ void *memmove(void *dest, const void *src, size_t n) {
 }
 
 void *memset(void *s, int c, size_t n) {
+#if CIRCUITPY_FULL_BUILD
     if (c == 0 && ((uintptr_t)s & 3) == 0) {
         // aligned store of 0
         uint32_t *s32 = s;
@@ -106,7 +112,9 @@ void *memset(void *s, int c, size_t n) {
         if (n & 1) {
             *((uint8_t*)s32) = 0;
         }
-    } else {
+    } else
+#endif
+    {
         uint8_t *s2 = s;
         for (; n > 0; n--) {
             *s2++ = c;