[llvm] r247972 - Simplify SmallBitVector::applyMask by consolidating common code for 32-bit and 64-bit builds.

Yaron Keren via llvm-commits llvm-commits at lists.llvm.org
Fri Sep 18 09:03:09 PDT 2015


r247993

Compiled without warnings and passed tests on VC 2013 32 bit and 64 bit.

2015-09-18 16:13 GMT+03:00 Aaron Ballman <aaron at aaronballman.com>:

> On Fri, Sep 18, 2015 at 9:10 AM, Yaron Keren <yaron.keren at gmail.com>
> wrote:
> > I think this will work OK, will test:
> >
> >   void applyMask(const uint32_t *Mask, unsigned MaskWords) {
> >     assert(MaskWords <= sizeof(uintptr_t) && "Mask is larger than
> base!");
> >     uintptr_t M = Mask[0];
> >     if (NumBaseBits == 64)
> >       M |= uint64_t(Mask[1]) << 32;
> >     if (InvertMask) M = ~M;
> >     if (AddBits) setSmallBits(getSmallBits() | M);
> >     else         setSmallBits(getSmallBits() & ~M);
> >   }
>
> Formatting aside, that looks reasonable to me. ;-) Thank you!
>
> ~Aaron
>
> >
> >
> >
> > 2015-09-18 16:02 GMT+03:00 Yaron Keren <yaron.keren at gmail.com>:
> >>
> >> The condition is on sizeof(uintptr_t) * CHAR_BIT which is a compile time
> >> constant. If it's not possible to make the same code work for both 32
> and 64
> >> bits without warnings it could be done with #ifdef (ugly) or template
> >> specialization for 32 and 64 bits where the unrequired version will not
> be
> >> instantiated.
> >>
> >>
> >> 2015-09-18 15:40 GMT+03:00 Aaron Ballman <aaron at aaronballman.com>:
> >>>
> >>> On Fri, Sep 18, 2015 at 8:37 AM, Yaron Keren <yaron.keren at gmail.com>
> >>> wrote:
> >>> > Thanks, this code will not be reached in 32 bits as the assert will
> >>> > fail but
> >>> > that won't hold in release config, the shift code should probably be
> in
> >>> > a
> >>> > conditional of both NumBaseBits and MaskWords, as was the original
> >>> > code. It
> >>> > should be still possible to merge them but more carefully.
> >>>
> >>> In theory, the code doesn't have to be reached for UB to cause
> >>> problems anyway. I wasn't certain that upcasting to 64-bits was the
> >>> right approach either, that's why I punted on just fixing it myself.
> >>> It still seems like it is solvable though.
> >>>
> >>> ~Aaron
> >>>
> >>> >
> >>> >
> >>> > 2015-09-18 15:20 GMT+03:00 Aaron Ballman <aaron at aaronballman.com>:
> >>> >>
> >>> >> On Fri, Sep 18, 2015 at 2:35 AM, Yaron Keren via llvm-commits
> >>> >> <llvm-commits at lists.llvm.org> wrote:
> >>> >> > Author: yrnkrn
> >>> >> > Date: Fri Sep 18 01:35:12 2015
> >>> >> > New Revision: 247972
> >>> >> >
> >>> >> > URL: http://llvm.org/viewvc/llvm-project?rev=247972&view=rev
> >>> >> > Log:
> >>> >> > Simplify SmallBitVector::applyMask by consolidating common code
> for
> >>> >> > 32-bit and 64-bit builds.
> >>> >> > Extend mask value to 64 bits before taking its complement and
> assert
> >>> >> > when mask is
> >>> >> > too large to apply in the small case (previously the extra words
> >>> >> > were
> >>> >> > silently ignored).
> >>> >> >
> >>> >> > http://reviews.llvm.org/D11890
> >>> >> >
> >>> >> > Patch by James Touton!
> >>> >> >
> >>> >> >
> >>> >> > Modified:
> >>> >> >     llvm/trunk/include/llvm/ADT/SmallBitVector.h
> >>> >> >     llvm/trunk/unittests/ADT/BitVectorTest.cpp
> >>> >> >
> >>> >> > Modified: llvm/trunk/include/llvm/ADT/SmallBitVector.h
> >>> >> > URL:
> >>> >> >
> >>> >> >
> http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/ADT/SmallBitVector.h?rev=247972&r1=247971&r2=247972&view=diff
> >>> >> >
> >>> >> >
> >>> >> >
> ==============================================================================
> >>> >> > --- llvm/trunk/include/llvm/ADT/SmallBitVector.h (original)
> >>> >> > +++ llvm/trunk/include/llvm/ADT/SmallBitVector.h Fri Sep 18
> 01:35:12
> >>> >> > 2015
> >>> >> > @@ -553,17 +553,15 @@ public:
> >>> >> >  private:
> >>> >> >    template<bool AddBits, bool InvertMask>
> >>> >> >    void applyMask(const uint32_t *Mask, unsigned MaskWords) {
> >>> >> > -    if (NumBaseBits == 64 && MaskWords >= 2) {
> >>> >> > -      uint64_t M = Mask[0] | (uint64_t(Mask[1]) << 32);
> >>> >> > -      if (InvertMask) M = ~M;
> >>> >> > -      if (AddBits) setSmallBits(getSmallBits() | M);
> >>> >> > -      else         setSmallBits(getSmallBits() & ~M);
> >>> >> > -    } else {
> >>> >> > -      uint32_t M = Mask[0];
> >>> >> > -      if (InvertMask) M = ~M;
> >>> >> > -      if (AddBits) setSmallBits(getSmallBits() | M);
> >>> >> > -      else         setSmallBits(getSmallBits() & ~M);
> >>> >> > +    uintptr_t M = Mask[0];
> >>> >> > +    if (MaskWords != 1) {
> >>> >> > +      assert(NumBaseBits == 64 && MaskWords == 2 &&
> >>> >> > +             "Mask is larger than base!");
> >>> >> > +      M |= uintptr_t(Mask[1]) << 32;
> >>> >>
> >>> >> This patch introduces undefined behavior on platforms where
> uintptr_t
> >>> >> is 32-bits, such as on Windows when compiling with MSVC in 23-bit
> >>> >> mode.
> >>> >>
> >>> >> 67>E:\llvm\llvm\include\llvm/ADT/SmallBitVector.h(560): warning
> C4293:
> >>> >> '<<': shift count negative or too big, undefined behavior (compiling
> >>> >> source file
> >>> >>
> E:\llvm\llvm\tools\clang\lib\StaticAnalyzer\Core\FunctionSummary.cpp)
> >>> >> 67>  E:\llvm\llvm\include\llvm/ADT/SmallBitVector.h(521): note: see
> >>> >> reference to function template instantiation 'void
> >>> >> llvm::SmallBitVector::applyMask<true,false>(const uint32_t
> *,unsigned
> >>> >> int)' being compiled (compiling source file
> >>> >>
> E:\llvm\llvm\tools\clang\lib\StaticAnalyzer\Core\FunctionSummary.cpp)
> >>> >>
> >>> >> I am reverting the patch (and r247972) temporarily since this also
> >>> >> generates a lot of warnings (everywhere SmallBitVector.h ends up
> being
> >>> >> included from) in r247983.
> >>> >>
> >>> >> ~Aaron
> >>> >>
> >>> >> >      }
> >>> >> > +    if (InvertMask) M = ~M;
> >>> >> > +    if (AddBits) setSmallBits(getSmallBits() | M);
> >>> >> > +    else         setSmallBits(getSmallBits() & ~M);
> >>> >> >    }
> >>> >> >  };
> >>> >> >
> >>> >> >
> >>> >> > Modified: llvm/trunk/unittests/ADT/BitVectorTest.cpp
> >>> >> > URL:
> >>> >> >
> >>> >> >
> http://llvm.org/viewvc/llvm-project/llvm/trunk/unittests/ADT/BitVectorTest.cpp?rev=247972&r1=247971&r2=247972&view=diff
> >>> >> >
> >>> >> >
> >>> >> >
> ==============================================================================
> >>> >> > --- llvm/trunk/unittests/ADT/BitVectorTest.cpp (original)
> >>> >> > +++ llvm/trunk/unittests/ADT/BitVectorTest.cpp Fri Sep 18 01:35:12
> >>> >> > 2015
> >>> >> > @@ -235,12 +235,12 @@ TYPED_TEST(BitVectorTest, PortableBitMas
> >>> >> >    const uint32_t Mask1[] = { 0x80000000, 6, 5 };
> >>> >> >
> >>> >> >    A.resize(10);
> >>> >> > -  A.setBitsInMask(Mask1, 3);
> >>> >> > +  A.setBitsInMask(Mask1, 2);
> >>> >> >    EXPECT_EQ(10u, A.size());
> >>> >> >    EXPECT_FALSE(A.test(0));
> >>> >> >
> >>> >> >    A.resize(32);
> >>> >> > -  A.setBitsInMask(Mask1, 3);
> >>> >> > +  A.setBitsInMask(Mask1, 2);
> >>> >> >    EXPECT_FALSE(A.test(0));
> >>> >> >    EXPECT_TRUE(A.test(31));
> >>> >> >    EXPECT_EQ(1u, A.count());
> >>> >> >
> >>> >> >
> >>> >> > _______________________________________________
> >>> >> > 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/20150918/b2905e5a/attachment.html>


More information about the llvm-commits mailing list