Changes between Version 2 and Version 3 of RTSsummaryEvents

Show
Ignore:
Timestamp:
12/21/11 07:02:52 (3 years ago)
Author:
MikolajKonarski (IP: 95.160.111.162)
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • RTSsummaryEvents

    v2 v3  
    3737}}} 
    3838                   
    39 No idea, TODO. 
     39Ask JaffaCake what does the formula mean: 
     40 
     41{{{ 
     42 (peak_mblocks_allocated * BLOCKS_PER_MBLOCK * BLOCK_SIZE_W - hw_alloc_blocks * BLOCK_SIZE_W) / (1024 * 1024 / sizeof(W_)) 
     43}}} 
     44 
     45in particular what is the difference between peak_mblocks_allocated and hw_alloc_blocks. 
    4046 
    4147{{{ 
     
    4854(if RequestParGC is not enough to tell seq/par). We don't want to report 
    4955the CPU time, only the elapsed time, and that's fine. 
     56With the current GC events only summarized information is available. 
    5057 
    5158{{{ 
     
    5966the events. Perhaps +RTS -s could be modified to make this figure simpler? 
    6067 
    61 To be continued. 
     68{{{ 
     69                        MUT time (elapsed)       GC time  (elapsed) 
     70  Task  0 (worker) :    0.04s    (  0.38s)       0.13s    (  0.17s) 
     71  Task  1 (worker) :    0.00s    (  0.55s)       0.00s    (  0.00s) 
     72  Task  2 (bound)  :    0.41s    (  0.41s)       0.13s    (  0.13s) 
     73  Task  3 (worker) :    0.00s    (  0.55s)       0.00s    (  0.00s) 
     74}}} 
     75 
     76Ask JaffaCake how the "tasks" relate to the "threads" for which we generate  
     77events. Perhaps in the existing GC evens add the info about which task 
     78does the job. BTW, is the time between event GCIdle and GCWork counted 
     79as GC time of the task? 
     80 
     81{{{ 
     82  SPARKS: 300001 (17739 converted, 282262 overflowed, 0 dud, 0 GC'd, 0 fizzled) 
     83}}} 
     84 
     85We have enough events for that (we calculate this using the SparkCounters 
     86events, but we could also use the precise per-spark events).