[llvm] r278584 - [IndVars] Ignore (s|z)exts that don't extend the induction variable

[David Majnemer via llvm-commits]

Hi Hans,

Can we get this merged in?

Thanks

On Fri, Aug 12, 2016 at 5:58 PM, Sanjoy Das via llvm-commits <
llvm-commits at lists.llvm.org> wrote:

> Author: sanjoy
> Date: Fri Aug 12 19:58:31 2016
> New Revision: 278584
>
> URL: http://llvm.org/viewvc/llvm-project?rev=278584&view=rev
> Log:
> [IndVars] Ignore (s|z)exts that don't extend the induction variable
>
> `IVVisitor::visitCast` used to have the invariant that if the
> instruction it was passed was a sext or zext instruction, the result of
> the instruction would be wider than the induction variable.  This is no
> longer true after rL275037, so this change teaches `IndVarSimplify` s
> implementation of `IVVisitor::visitCast` to work with the relaxed
> invariant.
>
> A corresponding change to SimplifyIndVar to preserve the said invariant
> after rL275037 would also work, but given how `IVVisitor::visitCast` is
> spelled (no indication of said invariant), I figured the current fix is
> cleaner.
>
> Fixes PR28935.
>
> Added:
>     llvm/trunk/test/Transforms/IndVarSimplify/pr28935.ll
> Modified:
>     llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp
>
> Modified: llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/
> IndVarSimplify.cpp?rev=278584&r1=278583&r2=278584&view=diff
> ============================================================
> ==================
> --- llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp (original)
> +++ llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp Fri Aug 12
> 19:58:31 2016
> @@ -816,6 +816,14 @@ static void visitIVCast(CastInst *Cast,
>    if (!Cast->getModule()->getDataLayout().isLegalInteger(Width))
>      return;
>
> +  // Check that `Cast` actually extends the induction variable (we rely
> on this
> +  // later).  This takes care of cases where `Cast` is extending a
> truncation of
> +  // the narrow induction variable, and thus can end up being narrower
> than the
> +  // "narrow" induction variable.
> +  uint64_t NarrowIVWidth = SE->getTypeSizeInBits(WI.NarrowIV->getType());
> +  if (NarrowIVWidth >= Width)
> +    return;
> +
>    // Cast is either an sext or zext up to this point.
>    // We should not widen an indvar if arithmetics on the wider indvar are
> more
>    // expensive than those on the narrower indvar. We check only the cost
> of ADD
>
> Added: llvm/trunk/test/Transforms/IndVarSimplify/pr28935.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/
> Transforms/IndVarSimplify/pr28935.ll?rev=278584&view=auto
> ============================================================
> ==================
> --- llvm/trunk/test/Transforms/IndVarSimplify/pr28935.ll (added)
> +++ llvm/trunk/test/Transforms/IndVarSimplify/pr28935.ll Fri Aug 12
> 19:58:31 2016
> @@ -0,0 +1,20 @@
> +; RUN: opt -S -indvars < %s | FileCheck %s
> +
> +target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
> +target triple = "x86_64-unknown-linux-gnu"
> +
> +declare i16 @fn1(i16 returned, i64)
> +
> +define void @fn2() {
> +; CHECK-LABEL: @fn2(
> +entry:
> +  br label %for.cond
> +
> +for.cond:
> +  %f.0 = phi i64 [ undef, %entry ], [ %inc, %for.cond ]
> +  %conv = trunc i64 %f.0 to i16
> +  %call = tail call i16 @fn1(i16 %conv, i64 %f.0)
> +  %conv2 = zext i16 %call to i32
> +  %inc = add nsw i64 %f.0, 1
> +  br label %for.cond
> +}
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160814/c777e7b9/attachment.html>