mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-12-15 07:39:31 +00:00
43c0d4789b
Summary: Use the new LoopVersioning facility (D16712) to add noalias metadata in the vector loop if we versioned with memchecks. This can enable some optimization opportunities further down the pipeline (see the included test or the benchmark improvement quoted in D16712). The test also covers the bug I had in the initial version in D16712. The vectorizer did not previously use LoopVersioning. The reason is that the vectorizer performs its transformations in single shot. It creates an empty single-block vector loop that it then populates with the widened, if-converted instructions. Thus creating an intermediate versioned scalar loop seems wasteful. So this patch (rather than bringing in LoopVersioning fully) adds a special interface to LoopVersioning to allow the vectorizer to add no-alias annotation while still performing its own versioning. As the vectorizer propagates metadata from the instructions in the original loop to the vector instructions we also check the pointer in the original instruction and see if LoopVersioning can add no-alias metadata based on the issued memchecks. Reviewers: hfinkel, nadav, mzolotukhin Subscribers: mzolotukhin, llvm-commits Differential Revision: http://reviews.llvm.org/D17191 llvm-svn: 263744
60 lines
1.5 KiB
LLVM
60 lines
1.5 KiB
LLVM
; RUN: opt -basicaa -scoped-noalias -loop-vectorize -licm -force-vector-width=2 \
|
|
; RUN: -force-vector-interleave=1 -S < %s | FileCheck %s
|
|
|
|
target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
|
|
|
|
; In order to vectorize the inner loop, it needs to be versioned with
|
|
; memchecks between {A} x {B, C} first:
|
|
;
|
|
; for (i = 0; i < n; i++)
|
|
; for (j = 0; j < m; j++)
|
|
; A[j] += B[i] + C[j];
|
|
;
|
|
; Since in the versioned vector loop A and B can no longer alias, B[i] can be
|
|
; LICM'ed from the inner loop.
|
|
|
|
|
|
define void @f(i32* %a, i32* %b, i32* %c) {
|
|
entry:
|
|
br label %outer
|
|
|
|
outer:
|
|
%i.2 = phi i64 [ 0, %entry ], [ %i, %inner.end ]
|
|
%arrayidxB = getelementptr inbounds i32, i32* %b, i64 %i.2
|
|
br label %inner.ph
|
|
|
|
inner.ph:
|
|
; CHECK: vector.ph:
|
|
; CHECK: load i32, i32* %arrayidxB,
|
|
; CHECK: br label %vector.body
|
|
br label %inner
|
|
|
|
inner:
|
|
%j.2 = phi i64 [ 0, %inner.ph ], [ %j, %inner ]
|
|
|
|
%arrayidxA = getelementptr inbounds i32, i32* %a, i64 %j.2
|
|
%loadA = load i32, i32* %arrayidxA, align 4
|
|
|
|
%loadB = load i32, i32* %arrayidxB, align 4
|
|
|
|
%arrayidxC = getelementptr inbounds i32, i32* %c, i64 %j.2
|
|
%loadC = load i32, i32* %arrayidxC, align 4
|
|
|
|
%add = add nuw i32 %loadA, %loadB
|
|
%add2 = add nuw i32 %add, %loadC
|
|
|
|
store i32 %add2, i32* %arrayidxA, align 4
|
|
|
|
%j = add nuw nsw i64 %j.2, 1
|
|
%cond1 = icmp eq i64 %j, 20
|
|
br i1 %cond1, label %inner.end, label %inner
|
|
|
|
inner.end:
|
|
%i = add nuw nsw i64 %i.2, 1
|
|
%cond2 = icmp eq i64 %i, 30
|
|
br i1 %cond2, label %outer.end, label %outer
|
|
|
|
outer.end:
|
|
ret void
|
|
}
|