Subject:
|
Re: detecting memory leaks
|
Newsgroups:
|
lugnet.robotics.rcx.legos
|
Date:
|
Fri, 12 Jul 2002 23:58:36 GMT
|
Viewed:
|
5153 times
|
| |
|
|
Jason,
Excellent points! I like where this is going.
"Jason Clark" wrote
<snip>
> Please, post.
See below.
> I'm guessing you defrag using mm_try_join()? malloc() uses this to defrag
> the beginning of the pool until it finds a big enough block (which can • then
> be split again if its too big);
Yes, it uses mm_try_join() --- here it is:
//! defragment free blocks
/*! use mm_try_join on each free block of memory
*/
void mm_defrag() {
size_t *ptr = &mm_start;
#ifdef CONF_TM
sem_wait(&mm_semaphore); // tasksafe
#endif
while(ptr >= &mm_start) {
if(*ptr == MM_FREE)
mm_try_join(ptr+1);
ptr += *(ptr+1);
ptr += MM_HEADER_SIZE;
}
#ifdef CONF_TM
sem_post(&mm_semaphore); // tasksafe
#endif
}
> Defragging during idle would save this
> overhead on every malloc if malloc is rewired to only mm_try_join() • *after*
> checking for a big enough block (and failing to get one).
I had done that with an older version of LegOS. It works fine (as one might
expect) but doesn't really improve performance much.
> Another thought... defrag is meaningless if no malloc() or free() has
> occurred since last defrag (calloc() calls malloc() internally)... might • be
> worth having a flag in mm.c that mm_defrag() clears and malloc/free set.
> Not only do extraneous calls to mm_defrag() cost less by checking the flag
> and exiting, you can auto-defrag with a thread of PRIO_MIN that sleeps • until
> the flag is set, defrags, repeats. No need to insert code before the
idle.
You hit the nail on the head. I will give this a try.
// Joe
That reminds me, there was a bug in mm.c concerning mm_try_join() attempting
to join the last block of memory. I will try to find that solution as well
(so it can be included)
|
|
Message has 1 Reply:
 | | Re: detecting memory leaks
|
| In case anyone attempts to use the mm_defrag() posted previously... you will need the following update to mm_try_join() //! check for free blocks after this one and join them if possible /* \param ptr pointer to size field of current block \return (...) (24 years ago, 13-Jul-02, to lugnet.robotics.rcx.legos)
|
Message is in Reply To:
| | Re: detecting memory leaks
|
| (...) Please, post. (...) I'm guessing you defrag using mm_try_join()? malloc() uses this to defrag the beginning of the pool until it finds a big enough block (which can then be split again if its too big); Defragging during idle would save this (...) (24 years ago, 12-Jul-02, to lugnet.robotics.rcx.legos)
|
5 Messages in This Thread:
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
This Message and its Replies on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
