Changes between Initial Version and Version 1 of Ticket #1600, comment 36


Ignore:
Timestamp:
Jan 13, 2014 5:05:24 PM (6 years ago)
Author:
nomeata
Comment:

Legend:

Unmodified
Added
Removed
Modified
  • Ticket #1600, comment 36

    initial v1  
    1 New numbers, after uprooting a bug where things unrelated to CPR (namely things alrady returning an unboxed tuple, with nothing to be CPRed inside) would suddenly get an `INLINE` flag, including some thunks. Finally a measurable positive change in the geometric mean!
    2 
    3 {{{
    4 --------------------------------------------------------------------------------
    5         Program           Size    Allocs   Runtime   Elapsed  TotalMem
    6 --------------------------------------------------------------------------------
    7            ansi          +0.3%     -0.1%      0.00      0.00     +0.0%
    8          awards          +0.3%     -0.1%      0.00      0.00     +0.0%
    9       compress2          +0.5%     -0.8%      0.11      0.11     -8.0%
    10        fibheaps          +0.3%     -0.3%      0.03      0.03     +0.0%
    11          gamteb          +0.3%     -0.2%      0.04      0.04     +0.0%
    12            grep          +0.3%     -0.1%      0.00      0.00     +0.0%
    13             hpg          +0.3%     -3.0%      0.13      0.13     +0.0%
    14           infer          +0.3%     -1.2%      0.04      0.04     +0.0%
    15    k-nucleotide          +0.1%     -6.9%     -1.3%     -1.1%     +0.0%
    16           kahan         -11.9%     -0.1%      0.17      0.17     +0.0%
    17        maillist          +0.3%     -0.8%      0.04      0.04    -22.5%
    18         mkhprog          +0.3%     -0.3%      0.00      0.00     +0.0%
    19             pic          +0.3%     -0.6%      0.00      0.00     +0.0%
    20          pretty          +0.3%     -0.1%      0.00      0.00     +0.0%
    21            rfib          +0.3%     -0.1%      0.01      0.01     +0.0%
    22             scc          +0.3%     -0.1%      0.00      0.00     +0.0%
    23   spectral-norm          +0.3%     -0.1%     +0.2%     +0.3%     +0.0%
    24          sphere          +0.3%     -4.7%      0.04      0.04     +0.0%
    25          symalg          +0.3%     -0.1%      0.01      0.01     +0.0%
    26             tak          +0.3%     -0.3%      0.01      0.01     +0.0%
    27       transform          +0.3%     +0.2%     -0.7%     -0.7%     +0.0%
    28       wave4main          +0.4%    +12.2%      0.21      0.21     +7.7%
    29 --------------------------------------------------------------------------------
    30             Min         -11.9%     -6.9%     -4.5%    -26.8%    -22.5%
    31             Max          +0.6%    +12.2%     +2.1%     +1.9%    +50.0%
    32  Geometric Mean          +0.2%     -0.1%     -0.3%     -2.1%     +0.1%
    33 }}}
    34 
    35 The increase of `wave4main` is due to a lost join point, as discussed in [wiki:NestedCPR/wave4main]. The increase in transform is not yet investigated (but will be, and then be discussed [wiki:NestedCPR#Motivatingexamples here]. I also did not look into the code size changes yet; `kahan` certainly looks interesting here...
    36 
    37 I was not especially careful about the runtime numbers, but I believe that the machine was unloaded when doing either ran. It certainly was when I was doing the baseline. So these are maybe a bit reliable.
     1(Removed post; looks like one of the trees was dirty in `libraries/base` from my experiments with inlining `runSTRep`. Re-running nofib....)