Changes between Version 27 and Version 28 of Frisby2013Q1


Ignore:
Timestamp:
Feb 25, 2013 12:16:17 AM (3 years ago)
Author:
nfrisby
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • Frisby2013Q1

    v27 v28  
    213213Binary sizes increase +5.5% with the first, and only +2.5 with the second. It's so consistent that it's probably in the base library.
    214214
    215 Allocation was within 0.1%, except for cacheprof (worse by 1%), but cacheprof is always wiggly.
     215Allocation was within 0.2%, except for cacheprof (worse by 1%), but cacheprof is always wiggly.
     216
     217====== With baseline libraries ======
     218
     219Reproducible elapsed time changes
     220
     221{{{
     222                   baseline       baseline2        it10             lam10
     223           anna     0.44           -0.2%           +0.2%           -3.8%
     224         genfft     0.21           +0.0%           +3.9%           +5.5%
     225          event     1.02           +0.1%           -0.8%           +0.0%
     226            scs     3.57           -0.2%           +1.1%           -0.4%
     227}}}
     228
     229 * event - a joinpoint Int# -> [...] gets floated, gains LSSELiS parameters, called 103999 times. Also, one of the abstracted variables keeps a case binder alive.
     230
     231 * scs - LinearAlgebra - several changes. uses of accumArray and listArray get inlined, and then joinpoints get floated out; adding 10 abstracted variables. Other similar functions (like unit) don't do this, so I'm not totally sure what triggers it.
     232
     233 * scs - Simulate - changes here too TODO
     234
     235TODO inspect the event CMM to see the sort of difference that this makes; find that 1%!
     236
     237TODO Why is the it10 float good in event but bad in scs? Maybe it's mostly the Simulate changes?
     238
     239TODO switching on protect-ignore seems to even out scs, but it does not affect event... ACK it's that darn "inlining putStr some how slows things down" again. Why's it being inlined if the libraries are baseline?
     240
     241====== With matching libraries ======
    216242
    217243Elapsed time changes