Re: [UFRaw-Devel] I'm back

[Simon K. <si...@ho...>]

On Mon, May 14, 2007 at 09:49:33AM -0700, Simon Kirby wrote:

> anything is possible.  I need to recompile lcms with debugging symbols
> and frame pointers and I'll try an oprofile again to see exactly where
> the problem is.

Ok, so it's definitely within licms:

CPU: AMD64 processors, speed 1890.06 MHz (estimated)
Counted CPU_CLK_UNHALTED events (Cycles outside of halt state) with a unit mask of 0x00 (No unit mask) count 100000
CPU_CLK_UNHALT...|
  samples|      %|
------------------
   242086 74.3300 liblcms.so.1.0.16
    45039 13.8288 libm-2.5.so
    13000  3.9915 ufraw
     4680  1.4369 vmlinux
     2580  0.7922 Xorg

samples  %        linenr info                 image name               app name                 symbol name
46900    14.4002  cmsintrp.c:517              liblcms.so.1.0.16        liblcms.so.1.0.16        cmsReverseLinearInterpLUT16
45039    13.8288  (no location information)   libm-2.5.so              libm-2.5.so              (no symbols)
42011    12.8990  cmsintrp.c:372              liblcms.so.1.0.16        liblcms.so.1.0.16        cmsLinearInterpLUT16
24557     7.5400  cmsmtrx.c:720               liblcms.so.1.0.16        liblcms.so.1.0.16        MAT3evalW
20643     6.3382  cmssamp.c:95                liblcms.so.1.0.16        liblcms.so.1.0.16        cmsSample3DGrid
14871     4.5660  cmspcs.c:290                liblcms.so.1.0.16        liblcms.so.1.0.16        cmsLab2XYZEncoded
11982     3.6789  cmspcs.c:149                liblcms.so.1.0.16        liblcms.so.1.0.16        f
9913      3.0437  cmssamp.c:34                liblcms.so.1.0.16        liblcms.so.1.0.16        _cmsQuantizeVal
9012      2.7670  cmslut.c:351                liblcms.so.1.0.16        liblcms.so.1.0.16        cmsEvalLUT
6912      2.1223  cmsxform.c:191              liblcms.so.1.0.16        liblcms.so.1.0.16        NormalXFORM
6900      2.1186  cmsmatsh.c:361              liblcms.so.1.0.16        liblcms.so.1.0.16        cmsEvalMatShaper
6660      2.0449  cmspcs.c:207                liblcms.so.1.0.16        liblcms.so.1.0.16        cmsXYZ2LabEncoded
5717      1.7553  cmsintrp.c:701              liblcms.so.1.0.16        liblcms.so.1.0.16        cmsTrilinearInterp16
3874      1.1895  (no location information)   liblcms.so.1.0.16        liblcms.so.1.0.16        .plt
3630      1.1146  nikon_curve.c:1003          ufraw                    ufraw                    CurveDataSample
3516      1.0796  cmsmtrx.c:168               liblcms.so.1.0.16        liblcms.so.1.0.16        FixedMul
3462      1.0630  cmscnvrt.c:251              liblcms.so.1.0.16        liblcms.so.1.0.16        XYZ2XYZ
3154      0.9684  nikon_curve.c:745           ufraw                    ufraw                    spline_cubic_val
2909      0.8932  cmscgats.c:0                liblcms.so.1.0.16        liblcms.so.1.0.16        __i686.get_pc_thunk.bx
2764      0.8487  cmsintrp.c:489              liblcms.so.1.0.16        liblcms.so.1.0.16        cmsLinearInterpFixed
2698      0.8284  ufraw_developer.c:265       ufraw                    ufraw                    developer_prepare

With "Disable soft proofing", liblcms does not even show up in the
profile.  As you might expect, things are then much faster. :)

What is the next step?  I don't understand how this would not affect
everyone...

Simon-