[llvm] r324943 - [LSR] Avoid UB overflow when examining reuse opportunities

[Adam Nemet via llvm-commits]

I reverted this per Gerolf’s request in r324958.

> On Feb 12, 2018, at 3:05 PM, Eric Christopher via llvm-commits <llvm-commits at lists.llvm.org> wrote:
> 
> It fails with clang on linux as well.
> 
> Also, is there no way to do it outside of a builtin?
> 
> Probably best to revert for now :)
> 
> -eric
> 
> On Mon, Feb 12, 2018 at 2:37 PM Craig Topper via llvm-commits <llvm-commits at lists.llvm.org <mailto:llvm-commits at lists.llvm.org>> wrote:
> I'm getting a compile eror after this change. It also doesn't seem portable to MSVC
> 
> lib/Transforms/Scalar/LoopStrengthReduce.cpp:3944:86: error: cannot initialize a parameter of type 'long long *' with an rvalue of type 'int64_t *' (aka 'long *')
>     if (__builtin_saddll_overflow(Imms.begin()->first, std::prev(Imms.end())->first, &Res))
> 
> ~Craig
> 
> On Mon, Feb 12, 2018 at 1:49 PM, Gerolf Hoflehner via llvm-commits <llvm-commits at lists.llvm.org <mailto:llvm-commits at lists.llvm.org>> wrote:
> Author: ghoflehner
> Date: Mon Feb 12 13:49:32 2018
> New Revision: 324943
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=324943&view=rev <http://llvm.org/viewvc/llvm-project?rev=324943&view=rev>
> Log:
> [LSR] Avoid UB overflow when examining reuse opportunities
> 
> Added:
>     llvm/trunk/test/CodeGen/X86/loop-strength-reduce-overflow-check.ll
> Modified:
>     llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp
> 
> Modified: llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp?rev=324943&r1=324942&r2=324943&view=diff <http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp?rev=324943&r1=324942&r2=324943&view=diff>
> ==============================================================================
> --- llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp (original)
> +++ llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp Mon Feb 12 13:49:32 2018
> @@ -3939,6 +3939,11 @@ void LSRInstance::GenerateCrossUseConsta
>      if (Imms.size() == 1)
>        continue;
> 
> +    // UB: Check for overflow
> +    int64_t Res;
> +    if (__builtin_saddll_overflow(Imms.begin()->first, std::prev(Imms.end())->first, &Res))
> +      continue;
> +
>      DEBUG(dbgs() << "Generating cross-use offsets for " << *Reg << ':';
>            for (const auto &Entry : Imms)
>              dbgs() << ' ' << Entry.first;
> 
> Added: llvm/trunk/test/CodeGen/X86/loop-strength-reduce-overflow-check.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/loop-strength-reduce-overflow-check.ll?rev=324943&view=auto <http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/loop-strength-reduce-overflow-check.ll?rev=324943&view=auto>
> ==============================================================================
> --- llvm/trunk/test/CodeGen/X86/loop-strength-reduce-overflow-check.ll (added)
> +++ llvm/trunk/test/CodeGen/X86/loop-strength-reduce-overflow-check.ll Mon Feb 12 13:49:32 2018
> @@ -0,0 +1,26 @@
> +; RUN: llc < %s -O1 -mtriple=x86_64-unknown-unknown
> +
> +; LSR has to check for overflow to avoid UB when it examines reuse opportunities
> +; Clang built with UBSan would expose the issue in this test case
> +
> +define void @main() {
> +bb:
> +  br label %bb1
> +
> +bb1:                                              ; preds = %bb1, %bb
> +  %tmp = phi i64 [ 248268322795906120, %bb ], [ %tmp10, %bb1 ]
> +  %tmp2 = sub i64 %tmp, 248268322795906120
> +  %tmp3 = getelementptr i8, i8* undef, i64 %tmp2
> +  %tmp4 = sub i64 %tmp, 248268322795906120
> +  %tmp5 = getelementptr i8, i8* undef, i64 %tmp4
> +  %tmp6 = getelementptr i8, i8* %tmp5, i64 -9086989864993762928
> +  %tmp7 = load i8, i8* %tmp6, align 1
> +  %tmp8 = getelementptr i8, i8* %tmp3, i64 -1931736422337600660
> +  store i8 undef, i8* %tmp8
> +  %tmp9 = add i64 %tmp, 248268322795906121
> +  %tmp10 = add i64 %tmp9, -248268322795906120
> +  br i1 undef, label %bb11, label %bb1
> +
> +bb11:                                             ; preds = %bb1
> +  ret void
> +}
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at lists.llvm.org <mailto:llvm-commits at lists.llvm.org>
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits <http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits>
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at lists.llvm.org <mailto:llvm-commits at lists.llvm.org>
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits <http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits>
> _______________________________________________
> 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/20180212/92d047f4/attachment.html>