[Bug target/50572] New: unstable performance on Atom due to loop alignment

["sergos.gnu at gmail dot com" <gcc-bugzilla@gcc.gnu.org>]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50572

             Bug #: 50572
           Summary: unstable performance on Atom due to loop alignment
    Classification: Unclassified
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: major
          Priority: P3
         Component: target
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: sergos.gnu@gmail.com


After monitoring of Atom performance on trunk for some period of time I figured
out that we have a significant (up to 15%) instability because of loop
alignment. Currently for Atom we have the following alignments:

  {&atom_cost, 16, 7, 16, 7, 16}

for

struct ptt
{
  const struct processor_costs *cost;           /* Processor costs */
  const int align_loop;                         /* Default alignments.  */
  const int align_loop_max_skip;
  const int align_jump;
  const int align_jump_max_skip;
  const int align_func;
};

Which means we try to align by 16, although if it takes no more than 7 bytes to
insert. This 'if' is the source of instability. For a reduction loop I observed
almost twice slowdown because it did not fit into 16bytes after being aligned
by 8.

I used the -falign-loops=16 option to measure code size impact using -m32-O2
-msse2 -mfpmath=sse -ffast-math -march=atom for SPEC2000:

        SPEC2000
Test        .text section size
-----------------------------------------
    Aligned    Current    Increas    %% increase
wupwise    630324    630084    240    0,04%
swim_    602612    602548    64    0,01%
mgrid_    608388    608212    176    0,03%
applu_    641684    641412    272    0,04%
mesa_    941444    938116    3328    0,35%
galgel_    813508    811764    1744    0,21%
art_    437572    437412    160    0,04%
equake_    442228    442084    144    0,03%
facerec    694948    694596    352    0,05%
ammp_    561428    560292    1136    0,20%
lucas_    663236    662948    288    0,04%
fma3d_    1565348    1560228    5120    0,33%
sixtrac    1537844    1534228    3616    0,24%
apsi_    719172    718340    832    0,12%
gzip_    480452    480020    432    0,09%
vpr_    548164    547156    1008    0,18%
cc1_    1554052    1546532    7520    0,49%
mcf_    434036    433908    128    0,03%
crafty_    592084    590836    1248    0,21%
parser_    509476    508276    1200    0,24%
eon_    1189348    1188852    496    0,04%
perlbmk    894292    891268    3024    0,34%
gap_    845636    841124    4512    0,54%
vortex_    969988    968788    1200    0,12%
bzip2_    472596    472260    336    0,07%
twolf_    607140    605044    2096    0,35%

Will it be acceptable to put -falign-loops=16 under -mtune=atom for O2?