Reapply the new LoopStrengthReduction code, with compile time and

bug fixes, and with improved heuristics for analyzing foreign-loop
addrecs.

This change also flattens IVUsers, eliminating the stride-oriented
groupings, which makes it easier to work with.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@95975 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Dan Gohman
2010-02-12 10:34:29 +00:00
parent 5cef638855
commit 572645cf84
35 changed files with 3523 additions and 2689 deletions

View File

@@ -25,7 +25,7 @@ bb1: ; preds = %bb2, %bb.nph
%j.01 = phi i64 [ %tmp9, %bb2 ], [ 0, %bb.nph ] ; <i64> [#uses=3]
%tmp3 = add i64 %j.01, %tmp1 ; <i64> [#uses=1]
%tmp4 = add i64 %j.01, %tmp2 ; <i64> [#uses=1]
%z0 = add i64 %tmp4, 5203
%z0 = add i64 %tmp3, 5203
%tmp5 = getelementptr double* %p, i64 %z0 ; <double*> [#uses=1]
%tmp6 = load double* %tmp5, align 8 ; <double> [#uses=1]
%tmp7 = fdiv double %tmp6, 2.100000e+00 ; <double> [#uses=1]