CreatePool MEMF_FAST

[Wepl]

maybe somebody will fail with same approach like me, used:

exec.CreatePool(MEMF_FAST,a,b)

and expected this will fail on a machine with chip ram only.
but no - this succeeds. you will get a memory pool which is unable to allocate any memory. because the pool structure is allocated with MEMF_ANY! all subsequent allocations in the pool are tried with MEMF_FAST.

[hooverphonique]

well, since the pool is grown dynamically, no memory is probably allocated before the first AllocPooled is called, and therefore CreatePool never even discovers that the system has no fastmem.. This may or may not appear stupid, depending on how you look at it..

the attributes used for allocating the pool structure itself rightfully has nothing to do with the memflags argument given to CreatePool..

[Wepl]

yes, but but was it returns under my conditions is useless.

[Asman]

Originally Posted by Wepl

maybe somebody will fail with same approach like me, used:

exec.CreatePool(MEMF_FAST,a,b)

and expected this will fail on a machine with chip ram only.
but no - this succeeds. you will get a memory pool which is unable to allocate any memory. because the pool structure is allocated with MEMF_ANY! all subsequent allocations in the pool are tried with MEMF_FAST.

Sound strange. If I understand correctly exec.AllocMem(a, MEMF_FAST) fails on that machine, right ?

[StingRay]

Originally Posted by Asman

If I understand correctly exec.AllocMem(a, MEMF_FAST) fails on that machine, right ?

Of course it will fail on a chipram only machine.

As for CreatePool, to me it somehow makes sense that it doesn't fail. Because all it does it allocating memory for the pool. And AllocPooled will later fail anyway if you try to alloc fast mem. So I don't really see a problem here.

[thomas]

What's the reason for using MEMF_FAST? Fast RAM is preferred anyway so MEMF_ANY should serve well on a machine with Fast RAM as well as on a machine without.

[Wepl]

because I don't want use memory pools on a machine which has no fast memory

[Wepl]

Originally Posted by StingRay

As for CreatePool, to me it somehow makes sense that it doesn't fail. Because all it does it allocating memory for the pool. And AllocPooled will later fail anyway if you try to alloc fast mem. So I don't really see a problem here.

typical answer of a programmer
I understand too why it is working that way.
From the customer point of view it simply does not deliver what it promises. It tells it can deliver fast mem and will never be able to do so.

[StingRay]

I guess it really depends on how you look at it. I would have expected CreatePool to fail too if you request fast mem for the pool on a chip only machine. On the other hand I can perfectly understand why it successfully creates the pool. :P

But I think it's better to use other approaches to check for fast mem anyway.

[BastyCDGS]

Hi!

Originally Posted by StingRay

I guess it really depends on how you look at it. I would have expected CreatePool to fail too if you request fast mem for the pool on a chip only machine. On the other hand I can perfectly understand why it successfully creates the pool. :P

But I think it's better to use other approaches to check for fast mem anyway.

Just curious, but is it possible to hotplug fast memory on (certain) Amiga(s)? In that case, the AllocPooled call would be successful if you hotplug some fast RAM after CreatePool has been called.

BTW, what happens if you do a AllocPooled with 0 bytes with MEMF_FAST? Does it fail, too? If yes, you could use that to check whether there is really some fast memory available.

[StingRay]

Originally Posted by BastyCDGS

BTW, what happens if you do a AllocPooled with 0 bytes with MEMF_FAST? Does it fail, too?

It doesn't fail.

Originally Posted by BastyCDGS

If yes, you could use that to check whether there is really some fast memory available.

There are better/easier ways to check if fast memory is available.