[PATCH] [RFC] Move to schedule trees to store our schedule

[Tobias Grosser]

The previous patch had some correctness issues slipping in, which invalidated
the original performance results. This patch should be more correct, but
again slower. We now convert from the schedule tree back to flat dependences
to allow our normal reduction dependence analysis to work.


http://reviews.llvm.org/D9226

Files:
  include/polly/ScopInfo.h
  lib/Analysis/DependenceInfo.cpp
  lib/Analysis/ScopInfo.cpp
  lib/CodeGen/IslAst.cpp
  lib/CodeGen/IslCodeGeneration.cpp
  lib/Exchange/JSONExporter.cpp
  lib/Transform/ScheduleOptimizer.cpp
  test/DeadCodeElimination/chained_iterations.ll
  test/DeadCodeElimination/chained_iterations_2.ll
  test/DeadCodeElimination/computeout.ll
  test/DeadCodeElimination/dead_iteration_elimination.ll
  test/DeadCodeElimination/non-affine-affine-mix.ll
  test/DeadCodeElimination/null_schedule.ll
  test/Isl/Ast/reduction_modulo_schedule_multiple_dimensions_4.ll
  test/Isl/Ast/simple-run-time-condition.ll
  test/Isl/CodeGen/loop_with_condition.ll
  test/Isl/CodeGen/loop_with_condition_2.ll
  test/Isl/CodeGen/loop_with_condition_ineq.ll
  test/Isl/CodeGen/loop_with_condition_nested.ll
  test/Isl/CodeGen/sequential_loops.ll
  test/ScheduleOptimizer/computeout.ll
  test/ScheduleOptimizer/line-tiling.ll
  test/ScheduleOptimizer/one-dimensional-band.ll
  test/ScheduleOptimizer/prevectorization.ll
  test/ScheduleOptimizer/rectangular-tiling.ll
  test/ScopInfo/NonAffine/non-affine-loop-condition-dependent-access_2.ll
  test/ScopInfo/NonAffine/non_affine_loop_used_later.ll
  test/ScopInfo/pointer-type-expressions.ll

EMAIL PREFERENCES
  http://reviews.llvm.org/settings/panel/emailpreferences/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: D9226.24381.patch
Type: text/x-patch
Size: 47166 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150424/c751ea49/attachment.bin>