OS 4 memory allocation

Anyone with inside (OS4) knowledge has noticed this thread on AmigaWorld ?
What's the problem here ? I did not have the opportunity yet to run the program on my A1 so can't confirm the figures but there is really something wrong there.
Anyone can explain us the problem in order to avoid replicating it in our own software ?
I really surprised because in everyday usage OS 4 does not seem to be that slow so either MOS is way time faster (and then OMG !! Good job MOS devs) either this piece of "sillybenchmark" is doing something the OS 4 allocator does not like and have poor performance in ? Or either OS 4 on Peg II (I guess it's on Peg II because I know Fab used to have a PegII) is really not optimized about the memory handling on this platform.

@abalaban

I've noticed a long time ago (when I was developing a file indexing app) that allocation of memory is fast but freeing/deallocation is very very slow. I've thought to report this to somebody but haven't done that. I think/hope OS4 devs knows this already (?). (Elapsed time should be reported between allocation and freeing also in any testing tools.) I would like to hear a comment from OS4 devs for this. (Or maybe it's taken care in the "ambitious project".)

I dont get it either. I have never tried MOS , but you say "I really surprised because in everyday usage OS 4 does not seem to be that slow"

How is the priority set in MOS? Because the test states that is was done with multitasking. Does MOS allow this test software to get all ressourses since it seems busy? While in OS4 it has to stand in que among the rest of the running software?

Perhaps a new test has to be run, with a measureable amount of other tasks running. Maybe there will be a different result.

Quote:
The priority is just 0 if you run the test from shell (or icon). The only way to change it is to use ChangeTaskPri command to change the priority of given shell window (so all programs launched from this window inherit the priority), or use SetTaskPri() from inside the source. Standard MorphOS scheduler is typical round-robin one, so priority is not changed by the system. It means, when this memtest runs at priority 0, any process running at higher priority may take all the CPU. Other processes at priority 0 will share the CPU with memtest. All tasks below 0 will be just effectively stopped, while memtest runs.