Changes between Version 1 and Version 2 of RTSsummaryEvents

Show
Ignore:
Timestamp:
12/21/11 06:43:43 (3 years ago)
Author:
MikolajKonarski (IP: 95.160.111.162)
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • RTSsummaryEvents

    v1 v2  
    11== The analysis of the semantics of +RTS -s and the list of needed new events == 
     2 
     3A sample output of +RTS -s, annotated with new require events. More concrete proposals will follow. 
     4 
     5{{{ 
     6     237,179,528 bytes allocated in the heap 
     7}}} 
     8 
     9An event for memory allocation and freeing memory via GC. 
     10 
     11{{{ 
     12      52,785,584 bytes copied during GC 
     13}}} 
     14 
     15An event for copying during GC. 
     16 
     17{{{ 
     18      17,272,336 bytes maximum residency (5 sample(s)) 
     19}}} 
     20 
     21An and event for that ("only checked during major garbage collections"). 
     22 
     23{{{ 
     24       6,493,328 bytes maximum slop 
     25}}} 
     26              
     27An event for slop, probably emitted rarely, also. 
     28 
     29{{{ 
     30              45 MB total memory in use 
     31}}} 
     32 
     33Can probably be calculated from allocation and deallocation. 
     34 
     35{{{ 
     36                  (0 MB lost due to fragmentation) 
     37}}} 
     38                   
     39No idea, TODO. 
     40 
     41{{{ 
     42                                    Tot time (elapsed)  Avg pause  Max pause 
     43  Gen  0       448 colls,     0 par    0.04s    0.04s     0.0001s    0.0016s 
     44  Gen  1         5 colls,     0 par    0.07s    0.07s     0.0149s    0.0386s 
     45}}} 
     46                               
     47Split the current GC events into generations and into seq/par 
     48(if RequestParGC is not enough to tell seq/par). We don't want to report 
     49the CPU time, only the elapsed time, and that's fine. 
     50 
     51{{{ 
     52  Parallel GC work balance: 1.00 (6391526 / 6375794, ideal 2) 
     53}}} 
     54   
     55This is quite convoluted: total words copied during parallel GCs divided 
     56by the average over all parallel GCs of maximal number of words copied 
     57by any thread in a single par GC.Revents needed: the events added above suffice, 
     58but quite a bit of extra state will have to be maintained when reading 
     59the events. Perhaps +RTS -s could be modified to make this figure simpler? 
     60 
     61To be continued.