[Bug 70613] Repeated function call gives HUGE increases to compile time

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=70613

--- Comment #3 from Petri Latvala <petri.latvala at intel.com> ---
Tested with Mesa commit e6c2afa9ceacc188690deba9da7b46bf1a0a9027 (2013-10-13,
10.0.0-devel)

Slow compile times reproduced in this version. Profiling gives

(pprof) top10
Total: 1537 samples
     299  19.5%  19.5%      299  19.5% fs_visitor::virtual_grf_interferes
     217  14.1%  33.6%     1132  73.6% fs_visitor::assign_regs
     185  12.0%  45.6%      185  12.0% pq_test (inline)
     155  10.1%  55.7%      267  17.4% ra_add_node_adjacency
     136   8.8%  64.5%      138   9.0% instruction_scheduler::add_dep [clone
.part.4]
      76   4.9%  69.5%       76   4.9% ra_get_spill_benefit.isra.1
      46   3.0%  72.5%      151   9.8% ra_add_node_interference
      33   2.1%  74.6%       33   2.1% get_assignment_entry
      29   1.9%  76.5%       40   2.6% _int_malloc
      24   1.6%  78.1%      100   6.5% __GI___libc_realloc

Further:

INTEL_DEBUG=perf prints

16-wide shader failed to compile, falling back to 8-wide at a 10-20%
performance cost: FS compile failed: Failure to register allocate.  Reduce
number of live scalar values to avoid this.
Fragment shader triggered register spilling.  Try reducing the number of live
scalar values to improve performance.

while compiling the shader, and afterwards (presumably triggered by
glUseProgram)

16-wide shader failed to compile, falling back to 8-wide at a 10-20%
performance cost: FS compile failed: Failure to register allocate.  Reduce
number of live scalar values to avoid this.
Recompiling fragment shader for program 1
  fragment color clamping 1->0
FS compile took 7057.783 ms and stalled the GPU
Fragment shader triggered register spilling.  Try reducing the number of live
scalar values to improve performance.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/intel-3d-bugs/attachments/20131021/d9a6d683/attachment.html>