diff options
| author | Michael Kruse <llvm@meinersbur.de> | 2018-12-18 17:16:05 +0000 |
|---|---|---|
| committer | Michael Kruse <llvm@meinersbur.de> | 2018-12-18 17:16:05 +0000 |
| commit | 3284775b70865581df3f75f6731f7d6464b0f8aa (patch) | |
| tree | 41c37ad09f91aac2bed4a762606fdf9084f9d175 /llvm/test/Transforms | |
| parent | 53b5cfb08035ca02770d6c72752a55742fdd0f5e (diff) | |
| download | bcm5719-llvm-3284775b70865581df3f75f6731f7d6464b0f8aa.tar.gz bcm5719-llvm-3284775b70865581df3f75f6731f7d6464b0f8aa.zip | |
[LoopUnroll] Honor '#pragma unroll' even with -fno-unroll-loops.
When using clang with `-fno-unroll-loops` (implicitly added with `-O1`),
the LoopUnrollPass is not not added to the (legacy) pass pipeline. This
also means that it will not process any loop metadata such as
llvm.loop.unroll.enable (which is generated by #pragma unroll or
WarnMissedTransformationsPass emits a warning that a forced
transformation has not been applied (see
https://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20181210/610833.html).
Such explicit transformations should take precedence over disabling
heuristics.
This patch unconditionally adds LoopUnrollPass to the optimizing
pipeline (that is, it is still not added with `-O0`), but passes a flag
indicating whether automatic unrolling is dis-/enabled. This is the same
approach as LoopVectorize uses.
The new pass manager's pipeline builder has no option to disable
unrolling, hence the problem does not apply.
Differential Revision: https://reviews.llvm.org/D55716
llvm-svn: 349509
Diffstat (limited to 'llvm/test/Transforms')
| -rw-r--r-- | llvm/test/Transforms/LoopUnroll/disable-loop-unrolling_forced.ll | 30 |
1 files changed, 30 insertions, 0 deletions
diff --git a/llvm/test/Transforms/LoopUnroll/disable-loop-unrolling_forced.ll b/llvm/test/Transforms/LoopUnroll/disable-loop-unrolling_forced.ll new file mode 100644 index 00000000000..9a0900df1a1 --- /dev/null +++ b/llvm/test/Transforms/LoopUnroll/disable-loop-unrolling_forced.ll @@ -0,0 +1,30 @@ +; RUN: opt -disable-loop-unrolling -O1 -S < %s | FileCheck %s +; +; Check loop unrolling metadata is honored despite automatic unrolling +; being disabled in the pass builder. +; +target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" + +; CHECK-LABEL: @forced( +; CHECK: load +; CHECK: load +define void @forced(i32* nocapture %a) { +entry: + br label %for.body + +for.body: + %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ] + %arrayidx = getelementptr inbounds i32, i32* %a, i64 %indvars.iv + %0 = load i32, i32* %arrayidx, align 4 + %inc = add nsw i32 %0, 1 + store i32 %inc, i32* %arrayidx, align 4 + %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 + %exitcond = icmp eq i64 %indvars.iv.next, 64 + br i1 %exitcond, label %for.end, label %for.body, !llvm.loop !0 + +for.end: + ret void +} + +!0 = distinct !{!0, !{!"llvm.loop.unroll.enable"}, + !{!"llvm.loop.unroll.count", i32 2}} |
