[llvm] r316208 - [ValueTracking] Enabling ValueTracking patch by default

[Davide Italiano via llvm-commits]

This commit caused https://bugs.llvm.org/show_bug.cgi?id=35087
Can you please take a look?

Thanks,

--
Davide

On Fri, Oct 20, 2017 at 3:08 AM, Nikolai Bozhenov via llvm-commits
<llvm-commits at lists.llvm.org> wrote:
> Author: n.bozhenov
> Date: Fri Oct 20 03:08:47 2017
> New Revision: 316208
>
> URL: http://llvm.org/viewvc/llvm-project?rev=316208&view=rev
> Log:
> [ValueTracking] Enabling ValueTracking patch by default
> (recommit #2 after checking for timeout issue).
>
> The original patch was an improvement to IR ValueTracking on
> non-negative integers. It has been checked in to trunk (D18777,
> r284022). But was disabled by default due to performance regressions.
> Perf impact has improved. The patch would be enabled by default.
>
> Reviewers: reames, hfinkel
>
> Differential Revision: https://reviews.llvm.org/D34101
>
> Patch by: Olga Chupina <olga.chupina at intel.com>
>
> Modified:
>     llvm/trunk/lib/Analysis/ValueTracking.cpp
>     llvm/trunk/test/Transforms/LoopUnroll/runtime-unroll-remainder.ll
>
> Modified: llvm/trunk/lib/Analysis/ValueTracking.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/ValueTracking.cpp?rev=316208&r1=316207&r2=316208&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Analysis/ValueTracking.cpp (original)
> +++ llvm/trunk/lib/Analysis/ValueTracking.cpp Fri Oct 20 03:08:47 2017
> @@ -83,12 +83,6 @@ const unsigned MaxDepth = 6;
>  static cl::opt<unsigned> DomConditionsMaxUses("dom-conditions-max-uses",
>                                                cl::Hidden, cl::init(20));
>
> -// This optimization is known to cause performance regressions is some cases,
> -// keep it under a temporary flag for now.
> -static cl::opt<bool>
> -DontImproveNonNegativePhiBits("dont-improve-non-negative-phi-bits",
> -                              cl::Hidden, cl::init(true));
> -
>  /// Returns the bitwidth of the given scalar or pointer type. For vector types,
>  /// returns the element type's bitwidth.
>  static unsigned getBitWidth(Type *Ty, const DataLayout &DL) {
> @@ -1289,9 +1283,6 @@ static void computeKnownBitsFromOperator
>            Known.Zero.setLowBits(std::min(Known2.countMinTrailingZeros(),
>                                           Known3.countMinTrailingZeros()));
>
> -          if (DontImproveNonNegativePhiBits)
> -            break;
> -
>            auto *OverflowOp = dyn_cast<OverflowingBinaryOperator>(LU);
>            if (OverflowOp && OverflowOp->hasNoSignedWrap()) {
>              // If initial value of recurrence is nonnegative, and we are adding
>
> Modified: llvm/trunk/test/Transforms/LoopUnroll/runtime-unroll-remainder.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/LoopUnroll/runtime-unroll-remainder.ll?rev=316208&r1=316207&r2=316208&view=diff
> ==============================================================================
> --- llvm/trunk/test/Transforms/LoopUnroll/runtime-unroll-remainder.ll (original)
> +++ llvm/trunk/test/Transforms/LoopUnroll/runtime-unroll-remainder.ll Fri Oct 20 03:08:47 2017
> @@ -47,7 +47,7 @@ for.cond.cleanup:
>  ; CHECK: or i64 [[INDVAR0]], 1
>  ; CHECK: or i64 [[INDVAR0]], 2
>  ; CHECK: or i64 [[INDVAR0]], 3
> -; CHECK: add nsw i64 [[INDVAR0]], 4
> +; CHECK: add nuw nsw i64 [[INDVAR0]], 4
>  ; CHECK: [[SUB:%[a-z.0-9]+]] = add i64 [[ITER]], -4
>  ; CHECK: [[ITER_CMP:%[a-z.0-9]+]] = icmp eq i64 [[SUB]], 0
>  ; CHECK: br i1 [[ITER_CMP]], label %[[LOOP_EXIT:.*]], label %for.body
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits



-- 
Davide

"A mathematical theory is not to be considered complete until you have
made it so clear that you can explain it to the first man whom you
meet on the street." (D. Hilbert)