http://llvm.org https://git.amat.us/llvm/atom/lib/Transforms?h=release_31 2012-05-02T10:00:50Z 2012-05-02T10:00:50Z Bill Wendling isanbard@gmail.com 2012-05-02T10:00:50Z urn:sha1:f8c1b779c46ca9ed43fdaddecfa813d6623beb7b ------------------------------------------------------------------------ r155984 | void | 2012-05-02 02:59:45 -0700 (Wed, 02 May 2012) | 4 lines The value held in the vector may be RAUW'ed by some of the canonicalization methods. Use a weak value handle to keep up with this. PR12245 ------------------------------------------------------------------------ git-svn-id: https://llvm.org/svn/llvm-project/llvm/branches/release_31@155985 91177308-0d34-0410-b5e6-96231b3b80d8 2012-04-30T17:58:59Z Bill Wendling isanbard@gmail.com 2012-04-30T17:58:59Z urn:sha1:c70ebff18a5cbb863576a525ba3a5fe0cf9216a4 git-svn-id: https://llvm.org/svn/llvm-project/llvm/branches/release_31@155834 91177308-0d34-0410-b5e6-96231b3b80d8 2012-04-30T17:48:14Z Bill Wendling isanbard@gmail.com 2012-04-30T17:48:14Z urn:sha1:458cf3ee70dea5a733a5b16286c5d7eb98a108d3 ------------------------------------------------------------------------ r155817 | void | 2012-04-30 03:44:54 -0700 (Mon, 30 Apr 2012) | 9 lines Second attempt at PR12573: Allow the "SplitCriticalEdge" function to split the edge to a landing pad. If the pass is *sure* that it thinks it knows what it's doing, then it may go ahead and specify that the landing pad can have its critical edge split. The loop unswitch pass is one of these passes. It will split the critical edges of all edges coming from a loop to a landing pad not within the loop. Doing so will retain important loop analysis information, such as loop simplify. ------------------------------------------------------------------------ git-svn-id: https://llvm.org/svn/llvm-project/llvm/branches/release_31@155833 91177308-0d34-0410-b5e6-96231b3b80d8 2012-04-30T17:41:42Z Bill Wendling isanbard@gmail.com 2012-04-30T17:41:42Z urn:sha1:b394a7b7128ea7a64e96cd88b0dc656e681707ce ------------------------------------------------------------------------ r155813 | void | 2012-04-30 02:23:48 -0700 (Mon, 30 Apr 2012) | 1 line Remove hack from r154987. The problem persists even with it, so it's not even a good hack. ------------------------------------------------------------------------ git-svn-id: https://llvm.org/svn/llvm-project/llvm/branches/release_31@155831 91177308-0d34-0410-b5e6-96231b3b80d8 2012-04-30T17:40:59Z Bill Wendling isanbard@gmail.com 2012-04-30T17:40:59Z urn:sha1:461573a1d6dcd3fce56037ff34f1fd8b7ccc2e4b ------------------------------------------------------------------------ r155166 | void | 2012-04-19 16:31:07 -0700 (Thu, 19 Apr 2012) | 1 line Put this expensive check below the less expensive ones. ------------------------------------------------------------------------ git-svn-id: https://llvm.org/svn/llvm-project/llvm/branches/release_31@155830 91177308-0d34-0410-b5e6-96231b3b80d8 2012-04-30T17:12:42Z Bill Wendling isanbard@gmail.com 2012-04-30T17:12:42Z urn:sha1:8cd9fe4d54889d516451ee36d00f130a3140468b ------------------------------------------------------------------------ r155809 | rafael | 2012-04-29 20:53:06 -0700 (Sun, 29 Apr 2012) | 2 lines Make sure HoistInsertPosition finds a position that is dominated by all inputs. ------------------------------------------------------------------------ git-svn-id: https://llvm.org/svn/llvm-project/llvm/branches/release_31@155828 91177308-0d34-0410-b5e6-96231b3b80d8 2012-04-18T06:00:09Z Bill Wendling isanbard@gmail.com 2012-04-18T06:00:09Z urn:sha1:135604129646650ef9abcd12a38b2d8976d3493c If the loop contains invoke instructions, whose unwind edge escapes the loop, then don't try to unswitch the loop. Doing so may cause the unwind edge to be split, which not only is non-trivial but doesn't preserve loop simplify information. Fixes PR12573 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@154987 91177308-0d34-0410-b5e6-96231b3b80d8 2012-04-18T04:00:10Z Andrew Trick atrick@apple.com 2012-04-18T04:00:10Z urn:sha1:b5122635966a980a850c028895e275a43e0f946d This introduces a threshold of 200 IV Users, which is very conservative but should be sufficient to avoid serious compile time sink or stack overflow. The llvm test-suite with LTO never exceeds 190 users per loop. The bug doesn't relate to a specific type of loop. Checking in an arbitrary giant loop as a unit test would be silly. Fixes rdar://11262507. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@154983 91177308-0d34-0410-b5e6-96231b3b80d8 2012-04-17T23:05:54Z Joe Groff arcata@gmail.com 2012-04-17T23:05:54Z urn:sha1:d5bda5ec663f43710fe462f44b77ddbcf8fe9d9e also fix SimplifyLibCalls to use TLI rather than compile-time conditionals to enable optimizations on floor, ceil, round, rint, and nearbyint git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@154960 91177308-0d34-0410-b5e6-96231b3b80d8 2012-04-16T12:39:17Z Hal Finkel hfinkel@anl.gov 2012-04-16T12:39:17Z urn:sha1:05bc5087a25bbcf59936d71ebfc878b545ef3e5c git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@154810 91177308-0d34-0410-b5e6-96231b3b80d8