Improve comments of mem.jou (#529)

stdlib/mem.jou

@@ -2,12 +2,25 @@
 
 # Heap allocations
 # TODO: write a tutorial about using these and add a link
-declare malloc(size: long) -> void*
-declare calloc(a: long, b: long) -> void*
-declare realloc(ptr: void*, size: long) -> void*
-declare free(ptr: void*) -> None
+declare malloc(size: long) -> void*  # allocate memory
+declare calloc(a: long, b: long) -> void*  # allocate a*b bytes of memory and zero it
+declare realloc(ptr: void*, new_size: long) -> void*  # grow/shrink allocated memory
+declare free(ptr: void*) -> None  # release allocated memory so it can be reused
 
-# TODO: explain what each of these does
-declare memset(dest: void*, fill_byte: int, count: long) -> void*
-declare memcpy(dest: void*, source: void*, count: long) -> void*
-declare memmove(dest: void*, source: void*, count: long) -> void*
+# This function fills a memory region with the given byte.
+# The most common way use case for this function is zeroing memory:
+#
+#   memset(&foo, 0, sizeof(foo))
+#
+declare memset(dest: void*, fill_byte: int, size: long) -> void*
+
+# These functions copy memory from one place to another. Source and destination
+# are of the same size, and pointers to their start are given.
+#
+# The difference between these two is how they handle overlapping memory. If
+# source and destination may overlap, use memmove(). If you know that source
+# and destination will never overlap, use memcpy(), because
+#   - it is a hint to people reading the code that there will be no overlap
+#   - it may be slightly faster.
+declare memcpy(dest: void*, source: void*, size: long) -> void*  # copy memory, overlaps are UB
+declare memmove(dest: void*, source: void*, size: long) -> void*  # copy memory, overlaps are ok