llvm/test/Transforms/LoopVectorize/X86, branch release_32 http://llvm.org https://git.amat.us/llvm/atom/test/Transforms/LoopVectorize/X86?h=release_32 2012-11-06T19:33:53Z Cost Model: add tables for some avx type-conversion hacks. 2012-11-06T19:33:53Z Nadav Rotem nrotem@apple.com 2012-11-06T19:33:53Z urn:sha1:b042868c01925dae3a1032890f591f1da78b19d3 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167480 91177308-0d34-0410-b5e6-96231b3b80d8 Code Model: Improve the accuracy of the zext/sext/trunc vector cost estimation. 2012-11-05T22:20:53Z Nadav Rotem nrotem@apple.com 2012-11-05T22:20:53Z urn:sha1:2d1528b3584ad85f2674f7db14b6ee607da678ec git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167412 91177308-0d34-0410-b5e6-96231b3b80d8 Implement the cost of abnormal x86 instruction lowering as a table. 2012-11-05T19:32:46Z Nadav Rotem nrotem@apple.com 2012-11-05T19:32:46Z urn:sha1:e623702c22e7cee4e02332b245a417a88ae6ffff git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167395 91177308-0d34-0410-b5e6-96231b3b80d8 LoopVectorize: Preserve NSW, NUW and IsExact flags. 2012-10-31T21:40:39Z Nadav Rotem nrotem@apple.com 2012-10-31T21:40:39Z urn:sha1:e57b2cbce652d45ed5516e52ad82991bfa03cfd7 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167174 91177308-0d34-0410-b5e6-96231b3b80d8 Fix a bug in the cost calculation of vector casts. Detect situations where bitcasts cost zero. 2012-10-31T20:52:26Z Nadav Rotem nrotem@apple.com 2012-10-31T20:52:26Z urn:sha1:0dba9a9a26f45119a82200d1db37b734b3e1bb22 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167170 91177308-0d34-0410-b5e6-96231b3b80d8 Add support for loops that don't start with Zero. 2012-10-31T00:45:26Z Nadav Rotem nrotem@apple.com 2012-10-31T00:45:26Z urn:sha1:462d1ca42831df53a3c6435d247776f0b85bd594 This is important for loops in the LAPACK test-suite. These loops start at 1 because they are auto-converted from fortran. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167084 91177308-0d34-0410-b5e6-96231b3b80d8 1. Fix a bug in getTypeConversion. When a *simple* type is split, we need to return the type of the split result. 2012-10-27T04:11:32Z Nadav Rotem nrotem@apple.com 2012-10-27T04:11:32Z urn:sha1:f065a8467785015336432e3e6e584798d8b48d8e 2. Change the maximum vectorization width from 4 to 8. 3. A test for both. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@166864 91177308-0d34-0410-b5e6-96231b3b80d8 Refactor the VectorTargetTransformInfo interface. 2012-10-26T23:49:28Z Nadav Rotem nrotem@apple.com 2012-10-26T23:49:28Z urn:sha1:a5a3a61c5fdcee972791d4e08441ba6edf131b88 Add getCostXXX calls for different families of opcodes, such as casts, arithmetic, cmp, etc. Port the LoopVectorizer to the new API. The LoopVectorizer now finds instructions which will remain uniform after vectorization. It uses this information when calculating the cost of these instructions. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@166836 91177308-0d34-0410-b5e6-96231b3b80d8 Move the target-specific tests, which require specific backends, to dirs that only run if the target is present. 2012-10-26T18:52:01Z Nadav Rotem nrotem@apple.com 2012-10-26T18:52:01Z urn:sha1:f56380114493ea501860aaff3c3b3b461c40233a git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@166796 91177308-0d34-0410-b5e6-96231b3b80d8