[Implemented] AllocateMemory without clearing

[HotteHC]

Hi Fred.

Would it be posible to make an optional Parameter to the
AllocateMemory-Procedure to dont clear the Allocated Mem.
I made some Tests and this would make an Speed-Increment at Factor 3
and more.

[Psychophanta]

+1

Looks like ReAllocateMemory() does not clear mem contents, just leave it intact, which AllocateMemory() should do by default :roll:

[netmaestro]

ReAllocateMemory leaves the contents intact for good reason, in fact if it didn't it would be more or less useless. It allows to resize the memory block up or down to fit the needed contents, which are usually at least partially existing. If AllocateMemory didn't clear the contents first, you'd be working in a memory block full of random junk which could easily masquerade as the stuff you're looking for in there, introducing some colossally hair-pulling bugs.

[Psychophanta]

If AllocateMemory didn't clear the contents first, you'd be working in a memory block full of random junk which could easily masquerade as the stuff you're looking for in there, introducing some colossally hair-pulling bugs.

The stuff to look for there is nothing, at least we firstly fill it with something.

[netmaestro]

I mean after you fill it, the block will contain your stuff + random junk.

[Psychophanta]

But is at the hand of the programmer to know or not to know what he is doing. So an optional parameter like #PB_ZeroMemory or so, would be great :roll: .

[Kaeru Gaman]

I agree with Psychophanta, that a Programmer should know halfways what's in Mem he allocated.

but I would strongly recommend to leave the AllocateMemory default to clearing, to help beginners avoid problems.
leaving allocated memory unchanged (filled with junk) to be more performant should be an explicitly requested feature.
the optional Flag should be set to leave the mem unchanged, not to clear it.

[HotteHC]

The fact, why I ask is, in some Situations the Memoryallocation is slow.
F.e. : I had to Allocate an Memoryblock with a size of 256000 very often and it seems to do so very slow. If i use HeapAlloc_ with the Parameter
#HEAP_ALIGN_16 it's much faster then HeapAlloc_ with Parameter
#HEAP_ALIGN_16|#HEAP_ZERO_MEMORY.

[PB]

> If AllocateMemory didn't clear the contents first, you'd be working in a
> memory block full of random junk which could easily masquerade as the
> stuff you're looking for in there

Rubbish. If you allocate but haven't then poked anything into it, then you
know it's just random stuff and not your own. :roll:

[ts-soft]

How often allocates you memory in a program. I see no real speed advantage

[netmaestro]

I mean after you fill it, the block will contain your stuff + random junk.

[PB]

Still doesn't matter, because if you fill it fully then it won't have junk, and if
you fill it partially then you know where you've filled and thus where the junk
still is. You know that, nm.

[Kaeru Gaman]

yeah, there is no reason to argue against the possibility to allocate without clearing...
only thing to talk about is, wich option should be default.

[Little John]

As Netmaestro wrote, the desired goal can already be achieved with ReAllocateMemory(), no?

Regards, Little John

[Trond]

ReAllocateMemory() also zeros the memory that is newly allocated.

Code: Select all

*A = AllocateMemory(24)
PokeS(*A, ReplaceString(Space(23), "a")) ; Fill *A with "a" characters
*B = ReAllocateMemory(*A, 1024)

The first 23 bytes of *B will now contain "a" characters, the 24th byte will contain the 0 from PokeS(), and the last 1000 bytes will contains zeros from ReAllocateMemory().