llvm dev - Jan 2018 - Does OpenMP hints bypass the vectorisation legality check in llvm

Hi all,

I am currently looking into how "#pragma omp for simd" is actually 
recognized in llvm. To my knowledge, clang will parse it and set 
metadata in IR to indicate this force-vectorization hint and later 
optimization passes would read it and vectorize the marked loop. 
Therefore, the loop should be vectorized even the compiler think it 
might not be safe to do so?

So my assumption is that such force-vectorization hints should bypass 
both the vectorization legality and cost model check. However, in 
LoopVectorize.cpp, I can't see how this is done. All loops will be sent 
to a legality check of LVL.canVectorize() and, if this condition does 
not fit, it return to false directly without actually reaching the 
vectorization stage.

Is there anything wrong with my assumption made on the use of 
force-vectorization hints?

Thanks in advance,

T

-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20180119/fa24bfb5/attachment.html>

Tom, your understanding is correct per OpenMP SIMD model. Our implementation
behaves as you stated. Which is not part of LLVM main trunk yet.

Xinmin

From: llvm-dev [mailto:llvm-dev-bounces at lists.llvm.org] On Behalf Of Tom Sun
via llvm-dev
Sent: Friday, January 19, 2018 12:39 AM
To: llvm-dev at lists.llvm.org
Subject: [llvm-dev] Does OpenMP hints bypass the vectorisation legality check in
llvm


Hi all,

I am currently looking into how "#pragma omp for simd" is actually
recognized in llvm. To my knowledge, clang will parse it and set metadata in IR
to indicate this force-vectorization hint and later optimization passes would
read it and vectorize the marked loop. Therefore, the loop should be vectorized
even the compiler think it might not be safe to do so?

So my assumption is that such force-vectorization hints should bypass both the
vectorization legality and cost model check. However, in LoopVectorize.cpp, I
can't see how this is done. All loops will be sent to a legality check of
LVL.canVectorize() and, if this condition does not fit, it return to false
directly without actually reaching the vectorization stage.

Is there anything wrong with my assumption made on the use of
force-vectorization hints?

Thanks in advance,

T
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20180119/94ac9dc6/attachment.html>

Hi, Tom,

Yes, OpenMP SIMD hints are currently translated into metadata that is
used by the vectorizer. If you'd like to know how that's done, see:

http://llvm.org/docs/LangRef.html#llvm-mem-parallel-loop-access-metadata

(and should also enable the vectorizer:
http://llvm.org/docs/LangRef.html#llvm-loop-vectorize-enable-metadata)

The relevant Clang code is in clang/lib/CodeGen/CGLoopInfo.cpp and
lib/CodeGen/CGStmtOpenMP.cpp (see, in particular, the comments around
emitSimdlenSafelenClause and EmitOMPSimdInit regarding safelen and
semantics).

The relevant check for the vectorizer is in LoopAccessAnalysis.cpp:

  const bool IsAnnotatedParallel = TheLoop->isAnnotatedParallel();

where isAnnotatedParallel checks for the parallel-loop-access metadata
(that indicates that the loop is data parallel) and is part of LoopInfo.

 -Hal


On 01/19/2018 02:38 AM, Tom Sun via llvm-dev wrote:
>
> Hi all,
>
> I am currently looking into how "#pragma omp for simd" is
actually
> recognized in llvm. To my knowledge, clang will parse it and set
> metadata in IR to indicate this force-vectorization hint and later
> optimization passes would read it and vectorize the marked loop.
> Therefore, the loop should be vectorized even the compiler think it
> might not be safe to do so?
>
> So my assumption is that such force-vectorization hints should bypass
> both the vectorization legality and cost model check. However, in
> LoopVectorize.cpp, I can't see how this is done. All loops will be
> sent to a legality check of LVL.canVectorize() and, if this condition
> does not fit, it return to false directly without actually reaching
> the vectorization stage. 
>
> Is there anything wrong with my assumption made on the use of
> force-vectorization hints?
>
> Thanks in advance,
>
> T
>
>
>
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
-- 
Hal Finkel
Lead, Compiler Technology and Programming Languages
Leadership Computing Facility
Argonne National Laboratory

-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20180119/750ff4fe/attachment-0001.html>