llvm dev - Dec 2010 - [LLVMdev] How can I determine safely if a CallSite is "live" in a DSGraphs context

Hi,

I believe shouldHaveNodeForValue() should return false for
ConstantPointerNullValue.

Fixed in r121707.

Arushi


On Mon, Dec 13, 2010 at 12:10 PM, Kevin Streit
<kevin.streit at googlemail.com>wrote:
> I'm using BUDataStructures... But I tried LocalDatastructures and it
didn't
> work either...
>  On Dec 13, 2010 6:52 PM, "Arushi Aggarwal" <arushi987 at
gmail.com> wrote:
> > Hi,
> >
> > Which DSA pass are you using, local/bu/td? I can try looking at this
once
> I
> > know that.
> >
> > Arushi
> >
> >
> > On Mon, Dec 13, 2010 at 11:26 AM, Kevin Streit
> > <kevin.streit at googlemail.com>wrote:
> >
> >> Hi,
> >>
> >> I'm using DSAnalysis in order to get some points to
information. In my
> >> particular case for a CallSite.
> >> Everything is working well except for some special cases. Consider
the
> >> following example code:
> >>
> >> // =====================> >>
> >> void m(short *s) {
> >> return;
> >> }
> >>
> >> int main() {
> >> m(0);
> >> }
> >>
> >> // =====================> >>
> >> The call to method m in the main method is translated to
"call void
> @m(i16*
> >> null) nounwind ssp".
> >> If I try to call getDSCallSiteForCallSite(...) with the above
mentioned
> >> CallSite as argument on the DSGraph of the main method I get a
failing
> >> assertion when the getDSCallSiteForCallSite method tries to get
the
> DSNode
> >> (via getNodeForValue(...)) for the "i16* null" argument.
So it obviously
> >> thinks it should have a node for that value (since
> >> shouldHaveNodeForValue(i16* null) returns true) and thus calls
> >> getNodeForValue, which raises the error (The node is not contained
in
> the
> >> ScalarMap).
> >>
> >> I tried to verify if I'm using the analysis correctly by
invoking the
> >> poolalloc tool on the program which raises a completely different
error.
> >>
> >> So under which circumstances is it safe to call
> getDSCallSiteForCallSite?
> >>
> >> Or stated otherwise: Should shouldHaveNodeForValue() return true
for a
> >> ConstantPointerNullValue?
> >>
> >> Thanks, Kevin
> >>
> >>
> >> _______________________________________________
> >> LLVM Developers mailing list
> >> LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu
> >> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
> >>
> >>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20101213/da9dbf33/attachment.html>

Yeah. That was my guess (and workaround fix) too. Thanks for the quick
reply. I just wonder that this didn't occur more often since it seems like a
common situation to me to pass a constant zero pointer argument. Anyway,
thanks again.
On Dec 13, 2010 8:21 PM, "Arushi Aggarwal" <arushi987 at
gmail.com> wrote:
> Hi,
>
> I believe shouldHaveNodeForValue() should return false for
> ConstantPointerNullValue.
>
> Fixed in r121707.
>
> Arushi
>
>
> On Mon, Dec 13, 2010 at 12:10 PM, Kevin Streit
> <kevin.streit at googlemail.com>wrote:
>
>> I'm using BUDataStructures... But I tried LocalDatastructures and
it
didn't
>> work either...
>> On Dec 13, 2010 6:52 PM, "Arushi Aggarwal" <arushi987 at
gmail.com> wrote:
>> > Hi,
>> >
>> > Which DSA pass are you using, local/bu/td? I can try looking at
this
once
>> I
>> > know that.
>> >
>> > Arushi
>> >
>> >
>> > On Mon, Dec 13, 2010 at 11:26 AM, Kevin Streit
>> > <kevin.streit at googlemail.com>wrote:
>> >
>> >> Hi,
>> >>
>> >> I'm using DSAnalysis in order to get some points to
information. In my
>> >> particular case for a CallSite.
>> >> Everything is working well except for some special cases.
Consider the
>> >> following example code:
>> >>
>> >> // =====================>> >>
>> >> void m(short *s) {
>> >> return;
>> >> }
>> >>
>> >> int main() {
>> >> m(0);
>> >> }
>> >>
>> >> // =====================>> >>
>> >> The call to method m in the main method is translated to
"call void
>> @m(i16*
>> >> null) nounwind ssp".
>> >> If I try to call getDSCallSiteForCallSite(...) with the above
mentioned
>> >> CallSite as argument on the DSGraph of the main method I get a
failing
>> >> assertion when the getDSCallSiteForCallSite method tries to
get the
>> DSNode
>> >> (via getNodeForValue(...)) for the "i16* null"
argument. So it
obviously
>> >> thinks it should have a node for that value (since
>> >> shouldHaveNodeForValue(i16* null) returns true) and thus calls
>> >> getNodeForValue, which raises the error (The node is not
contained in
>> the
>> >> ScalarMap).
>> >>
>> >> I tried to verify if I'm using the analysis correctly by
invoking the
>> >> poolalloc tool on the program which raises a completely
different
error.
>> >>
>> >> So under which circumstances is it safe to call
>> getDSCallSiteForCallSite?
>> >>
>> >> Or stated otherwise: Should shouldHaveNodeForValue() return
true for a
>> >> ConstantPointerNullValue?
>> >>
>> >> Thanks, Kevin
>> >>
>> >>
>> >> _______________________________________________
>> >> LLVM Developers mailing list
>> >> LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu
>> >> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>> >>
>> >>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20101213/a65a118b/attachment.html>