changed behaviour of malloc(0) and optimized calloc

[ZERICO2005]

calloc (when __TICE__ is defined) now uses an inlined implementation of bzero which uses the $E40000 address to speed up the zero filling of memory.

Otherwise, it will use the previous memset implementation when __TICE__ is undefined.

[runer112]

Seems like the add here should never carry?

Suggested change

	add hl, bc
	or a, a
	sbc hl, bc
	add hl, bc
	sbc hl, bc

[calc84maniac]

Only if you also move up the pop of the size (the pop of the parameter is undefined data because malloc may clobber it)

[ZERICO2005]

does this go off the assumption that malloc won't return an address higher than 0xE40000 on the CE and that the allocation size is less than 0x1C0000 bytes or etc

[runer112]

This relies on the seemingly sound assumption that the pointer to the allocated memory plus the allocation size does not overflow.

[runer112]

I went down a bit of a rabbit hole here... I thought: "What's the point of malloc'ing 0 bytes? Could we treat that as an allocation failure and return null?" And sure enough, the C standard allows for exactly this:

If the size of the space requested is zero, the behavior is implementation-defined: either a null pointer is returned to indicate an error, or the behavior is as if the size were some nonzero value, except that the returned pointer shall not be used to access an object.

If we push the zero-size check into malloc, we can eliminate the check here. We'd have to conditionally keep the check here to be safe if _custom_malloc is being used, but otherwise the check can be cheaply pushed into __simple_malloc:

 	ex	(sp),hl
 	push	bc
 	ld	de,(_heap_ptr)
+	dec hl
 	add	hl,de
 	jr	c,.null
- 	ld	bc,___heaptop
+ 	ld	bc,___heaptop-1
 	sbc	hl,bc
 	jr	nc,.null
 	add	hl,bc

And _standard_malloc:

 
     /* add size of block header to real size */
     const size_t size = alloc_size + sizeof(block_t);
+    /* abort if alloc_size is 0 or size overflowed */
-    if (size < alloc_size)
+    if (size <= alloc_size)
     {
         return NULL;
     }

The same idea can be applied to the no hardware assumptions version of calloc.

[ZERICO2005]

I assume it is okay for cpd to read nonnull_ptr_from_malloc + size