forked from OSchip/llvm-project
eb7311ffb1
Fixes cases like the new test @nonuniform. In that test, %cc itself is a uniform value; however, when reading it after the end of the loop in basic block %if, its value is effectively non-uniform, so the branch is non-uniform. This problem was encountered in https://bugs.freedesktop.org/show_bug.cgi?id=103743; however, this change in itself is not sufficient to fix that bug, as there is another issue in the AMDGPU backend. As discovered after committing an earlier version of this change, this exposes a subtle interaction between this pass and DivergenceAnalysis: since we remove and re-create branch instructions, we can no longer rely on DivergenceAnalysis for branches in subregions that were already processed by the pass. Explicitly remove branch instructions from DivergenceAnalysis to avoid dangling pointers as a matter of defensive programming, and change how we detect non-uniform subregions. Change-Id: I32bbffece4a32f686fab54964dae1a5dd72949d4 Differential Revision: https://reviews.llvm.org/D43743 llvm-svn: 329165 |
||
|---|---|---|
| .. | ||
| AMDGPU | ||
| branch-on-argument.ll | ||
| bug36015.ll | ||
| invert-constantexpr.ll | ||
| loop-multiple-exits.ll | ||
| nested-loop-order.ll | ||
| no-branch-to-entry.ll | ||
| one-loop-multiple-backedges.ll | ||
| post-order-traversal-bug.ll | ||
| rebuild-ssa-infinite-loop.ll | ||
| switch.ll | ||