Re: [aspectj-users] Garbage collector behavior

[Mathieu LEMAIRE <mlemaire@xxxxxxx>]

Hi all,

I had to bench clearly memory consumptions and cpu usage overheads in 
those configs (aspect scalability, LTW / compile time W, no aspect at 
all) and here is what i got :

benchmark
	weaving 	glM (KB) 	tgPk (KB) 	ccM (KB) 	tgM (KB) 	glT (ms) 	gcT (ms)
bench1 	CT few aspects 	70,085.41 	97,510.18 	26.71 	65,470.21 
15,555.11 	4,590.33
CT many aspects 	70,064.73 	97,497.00 	26.11 	65,470.54 	15,956.42 
4,509.00
LT few aspects
	117,904.67 	112,381.06 	161.26 	112,381.06 	16,512.71 	2,179.00
no aspects
	70,067.72 	97,515.85 	25.37 	65,469.83 	15,737.30 	4,551.67
LT many aspects
	119,323.28 	109,141.54 	107.46 	109,141.54 	18,595.90 	2,380.67
bench2 	CT few aspects 	77,874.13 	72,646.21 	31.59 	72,646.21 
40,059.81 	6,447.67
CT many aspects 	75,584.27 	70,877.49 	29.85 	70,877.49 	41,238.57 
6,379.33
LT few aspects
	103,814.15 	103,671.09 	267.99 	93,940.62 	38,536.98 	4,707.67
no aspects
	77,716.08 	72,481.75 	29.85 	72,481.75 	40,638.22 	6,409.00
LT many aspects
	102,670.77 	104,942.94 	238.74 	95,081.41 	43,023.24 	4,337.00


quick legend :

   * *glM : *globalMemoryConsumed
   * *tgPk : *peakMemoryConsumed
   * *ccM : *codeCacheMemoryConsumed
   * *tgM : *tenuredGenMemoryConsumed
   * *glT : *globalTime
   * *gcT : *gcTime

Those marks have been done using hotspot 1.5 and new JMX features..
Well those results should not stand as direct proofs but maybe just 
hints on the LTW consumption.

My aspects are really simple : for profiling I do not use any runtime 
check nor advanced aspectj features such as thisJointPoint** and when I 
say *many* aspects, i just mean 15, compared to *few* (1) :)

The important thing is that the overhead resides in the tenured gen 
pool, quite an old space.. that means that aspectj ltw still holds hard 
references somewhere and forcing collections should not help with any 
thing... Other interesting stuff is code cache consumption ; well I do 
not know much about that, but ltw needs much more than simply 
duplicating class defs.

The very good news is that compile time weaving do not introduce any 
distorsion ; that is great for my profiling !!

hope it helps.
--
Mathieu