This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: Kudos to the gcc developers


On Thu, Nov 30, 2000 at 11:45:25AM -0500, Brad Lucier wrote:
> 
> 1.  After Michael Matz's dominators calculation was integrated into
> gcc, I ran a profiled version to see what the next significant "hot
> spot" was in the compiler.  I couldn't find one.

Hm, this surprises me.  The profiles I do show significant time
(i.e. >5%) in the garbage collector and sometimes in CSE (not GCSE).
For example:

  %   cumulative   self              self     total           
 time   seconds   seconds    calls  ms/call  ms/call  name    
 10.99      1.57     1.57 13711554     0.00     0.00  ggc_set_mark
 10.50      3.07     1.50       21    71.43   167.08  ggc_mark_trees
  5.88      3.91     0.84        1   840.00 14218.24  yyparse
  4.27      4.52     0.61  1639379     0.00     0.00  memset

20% of runtime in garbage collection, ouch.  Incidentally, memset is
mostly being called by make_node:

                0.00    0.00    5420/1639379     make_tree_vec [444]
                0.01    0.00   28262/1639379     ggc_alloc_cleared [246]
                0.03    0.00   85756/1639379     ggc_collect [5]
                0.11    0.00  295255/1639379     calloc [55]
                0.46    0.00 1223609/1639379     make_node [17]
[34]     4.3    0.61    0.00 1639379         memset [34]

I think we want something akin to genrtl.c for trees.  I doubt it can
be done in 3.0, though.

zw

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]