~jameinel/meliae/proxy : changes

~jameinel/meliae/proxy » Changes from revision 123

From Revision 123 to 104

Rev	Summary	Authors	Tags	Date
123	Start playing around with an ObjProxy that can give prettier... Start playing around with an ObjProxy that can give prettier output. Using a proxy object, we can look up extra info in the ObjManager and generally give quick-info.	John Arbash Meinel		14 years ago
122	comment cleanups. comment cleanups.	John Arbash Meinel		14 years ago
121	Add a __getitem__ helper, which should make it easier to wal... Add a __getitem__ helper, which should make it easier to walk around.	John Arbash Meinel		14 years ago
120	A bit more cleanup. A bit more cleanup.	John Arbash Meinel		14 years ago
119	Create a real struct, which just makes it easier to work wit... Create a real struct, which just makes it easier to work with.	John Arbash Meinel		14 years ago
118	Change the internals of _loader a fair amount. Change the internals of _loader a fair amount. Rather than trying to pack things into C long types, we go ahead and keep the reference to them. Should be interesting.	John Arbash Meinel		14 years ago
117	Some quick fixes to handle refs that are bigger than the cur... Some quick fixes to handle refs that are bigger than the current bit width.	John Arbash Meinel		14 years ago
116	It seems setup.py had a different idea of the current versio... It seems setup.py had a different idea of the current version. Fix it so that it takes the value from meliae.__version__ instead. And release 0.1.2, since 0.1.1 seemed to have already been taken.	John Arbash Meinel	meliae-0.1.2	14 years ago
115	Release Meliae 0.1.0-final Release Meliae 0.1.0-final	John Arbash Meinel	meliae-0.1.0	14 years ago
114	add some flush and close calls, as well as clearing out gc o... add some flush and close calls, as well as clearing out gc object lists. This is in reference to bug #428165. I'm not positive that it fixes them, but it should at least make it better.	John Arbash Meinel		14 years ago
113	Fix bug #493531 from James Henstridge. Fix bug #493531 from James Henstridge.	John Arbash Meinel		14 years ago
112	Handle when there are fewer than 20 summaries. Handle when there are fewer than 20 summaries.	John O'Brien		14 years ago
111	When using dump_all_objects(), don't include the gc list its... When using dump_all_objects(), don't include the gc list itself.	John Arbash Meinel		14 years ago
110	make PerformanceCounter.get_timer() public. make PerformanceCounter.get_timer() public. Add a script to spawn a subprocess and watch its memory consumption. The only thing this really benefits is 'mem-secs', the idea of how many MB have been allocated for how long. This is slightly more useful than peak memory for stuff like codehosting. Since even if you peak very shortly, keeping the total memory consumption lower gives you better concurrency.	John Arbash Meinel		14 years ago
109	Do a little bit to help ensure we don't get a deadlock. Do a little bit to help ensure we don't get a deadlock.	John Arbash Meinel		14 years ago
108	Implement a PerformanceCounter module. Implement a PerformanceCounter module. This allows us to track memory consumption, etc for a child process. We might consider rewriting all of this as a pyrex extension. It would probably provide a bit better performance, if it is something we are going to be calling a lot. For now, this seems good enough.	John Arbash Meinel		14 years ago
107	pre-allocate the big dicts, if we are going to run out of me... pre-allocate the big dicts, if we are going to run out of memory, do it early.	John Arbash Meinel		14 years ago
106	Switch to using PyMem_Malloc/Free Switch to using PyMem_Malloc/Free	John Arbash Meinel		14 years ago
105	Document the new functionality Document the new functionality	John Arbash Meinel		14 years ago
104	Do some ugly hacks to keep memory low during 'compute_referr... Do some ugly hacks to keep memory low during 'compute_referrers'. 1) If there is only 1 referrer, keep the 'list' as a simple integer 2) If there is <10 referrers, use a tuple, and create new ones as necessary 3) If there is >=10 referrers, use a list as normal This should decrease memory a bit, when dealing with really big datasets. Quite a few objects will only have 1 reference, and this drops memory consumption down to a dict entry / pointer (because the address is already known to be a unique PyInt that we already have.)	John Arbash Meinel		14 years ago

Older »