en Copyright 2015 Java http://172.16.0.5:8080/history/llvm-project-15.0.7/llvm/test/Transforms/StructurizeCFG/AMDGPU/backedge-id-bug.ll#cee313d2 Revert "Temporarily Revert "Add basic loop fusion pass.""The reversion apparently deleted the test/Transforms directory.Will be re-reverting again.llvm-svn: 358552 List of files: /llvm-project-15.0.7/llvm/test/Transforms/StructurizeCFG/AMDGPU/backedge-id-bug.ll Wed, 17 Apr 2019 04:52:47 +0000 Eric Christopher <echristo@gmail.com> http://172.16.0.5:8080/history/llvm-project-15.0.7/llvm/test/Transforms/StructurizeCFG/AMDGPU/backedge-id-bug.ll#4afb64e4 Revert r321751, "StructurizeCFG: Fix broken backedge detection"It causes regressions in various OpenGL test suites.Keep the test cases introduced by r321751 as XFAIL, and add a test casefor the regression.Change-Id: I90b4cc354f68cebe5fcef1f2422dc8fe1c6d3514Bugzilla: https://bugs.llvm.org/show_bug.cgi?id=36015llvm-svn: 323355 List of files: /llvm-project-15.0.7/llvm/test/Transforms/StructurizeCFG/AMDGPU/backedge-id-bug.ll Wed, 24 Jan 2018 18:02:05 +0000 Nicolai Haehnle <nhaehnle@gmail.com> http://172.16.0.5:8080/history/llvm-project-15.0.7/llvm/test/Transforms/StructurizeCFG/AMDGPU/backedge-id-bug.ll#35c4244a StructurizeCFG: xfail one of the testcases from r321751It fails with -verify-region-info. This seems to be a issuewith RegionInfo itself which existed before.llvm-svn: 321806 List of files: /llvm-project-15.0.7/llvm/test/Transforms/StructurizeCFG/AMDGPU/backedge-id-bug.ll Thu, 04 Jan 2018 17:23:24 +0000 Matt Arsenault <Matthew.Arsenault@amd.com> http://172.16.0.5:8080/history/llvm-project-15.0.7/llvm/test/Transforms/StructurizeCFG/AMDGPU/backedge-id-bug.ll#8070882b StructurizeCFG: Fix broken backedge detectionThe work order was changed in r228186 from SCC orderto RPO with an arbitrary sorting function. The sortingfunction attempted to move inner loop nodes earlier. Thiswas was apparently relying on an assumption that every blockin a given loop / the same loop depth would be seen beforevisiting another loop. In the broken testcase, a blockoutside of the loop was encountered before moving ontoanother block in the same loop. The testcase would thenstructurize such that one blocks unconditional successorcould never be reached.Revert to plain RPO for the analysis phase. This fixesdetecting edges as backedges that aren't really.The processing phase does use another visited set, andI'm unclear on whether the order there is as important.An arbitrary order doesn't work, and triggers some infiniteloops. The reversed RPO list seems to work and is closerto the order that was used before, minus the arbitarycustom sorting.A few of the changed tests now produce smaller code,and a few are slightly worse looking.llvm-svn: 321751 List of files: /llvm-project-15.0.7/llvm/test/Transforms/StructurizeCFG/AMDGPU/backedge-id-bug.ll Wed, 03 Jan 2018 18:45:37 +0000 Matt Arsenault <Matthew.Arsenault@amd.com>