Andrew Santosa via llvm-dev
2017-Jun-04 02:31 UTC
[llvm-dev] Get elements from a getelementptr constant expression
If what is needed is to simply extract the operands of a getelementptr
constant expression, I wrote a simple example here:
https://github.com/analysis-examples/gep-expr-example
Best,
Andrew
On Wednesday, 31 May 2017, 19:53, via llvm-dev <llvm-dev at
lists.llvm.org> wrote:
Message: 8
Date: Tue, 30 May 2017 22:13:09 -0400
From: Damilola Orikogbo via llvm-dev <llvm-dev at lists.llvm.org>
To: llvm-dev at lists.llvm.org
Subject: [llvm-dev] Get elements from a getelementptr constant
expression
Message-ID:
<CAJJs-5kRu1r1L8zy56Vv1g78Uijj-EoJ4_z689cgXd=2vMugWA at
mail.gmail.com>
Content-Type: text/plain; charset="utf-8"
Hi,
I have this instruction,
store i32 (...)** bitcast (i8** getelementptr inbounds ([5 x i8*], [5 x
i8*]* @_ZTV9Rectangle, i32 0, i32 2) to i32 (...)**), i32 (...)*** %5,
align 8
and was wondering how to access the indices of the getelementptr. How would
i print out all pointers stored from index 0 to 5.
Cheers,
Dami
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20170530/4bd0d076/attachment-0001.html>
------------------------------
Message: 9
Date: Tue, 30 May 2017 23:20:02 -0400
From: Damilola Orikogbo via llvm-dev <llvm-dev at lists.llvm.org>
To: Tim Northover <t.p.northover at gmail.com>, llvm-dev at lists.llvm.org
Subject: Re: [llvm-dev] Get elements from a getelementptr constant
expression
Message-ID:
<CAJJs-5=nn1JRNWoENCB3Q2coqmr=V1rDJzcQTMTRQTnF6pg1dg at
mail.gmail.com>
Content-Type: text/plain; charset="utf-8"
Hello,
I apologize, I am new to the list-serve (joined on the 25th) and am not
quite sure how it works.
I didn't reply to Hans because the llvm/ clang distribution I have
installed doesn't allow me to use that command line option, I will respond
in a bit. And I asked anther question simply because no one replied to my
first question.
I have been using llvm for about a week, so I am not sure how to ask my
question succinctly but I am trying, please bear with me.
In reply to your response, I will try that out tomorrow and hopefully be
able to make some headway.
I have been searching previous emails and have found that some people have
cast the getelementptr in the Instruction i posted in this original email
thread to a constantexpr or GEPOperator and have been able to get
information that way, but I can't quite seem to figure it out.
*Ex email thread:*
https://groups.google.com/forum/#!topic/llvm-dev/_F-PFACaHbU
<https://groups.google.com/forum/#!topic/llvm-dev/_F-PFACaHbU>
If anyone has tried this approach and could provide some more information
going this route, I would appreciate that as well.
Cheers,
Dami
On Tue, May 30, 2017 at 10:37 PM, Tim Northover <t.p.northover at
gmail.com>
wrote:
> Hi,
>
> On 30 May 2017 at 19:13, Damilola Orikogbo via llvm-dev
> <llvm-dev at lists.llvm.org> wrote:
> > store i32 (...)** bitcast (i8** getelementptr inbounds ([5 x i8*], [5
x
> > i8*]* @_ZTV9Rectangle, i32 0, i32 2) to i32 (...)**), i32 (...)*** %5,
> align
> > 8
> >
> > and was wondering how to access the indices of the getelementptr. How
> would
> > i print out all pointers stored from index 0 to 5.
>
> You really need to start replying to your own threads. The one I
> noticed this morning was particularly useless compared to your message
> from 25th May, and you didn't even clarify your problem for Hans after
> he replied.
>
> That said, if you'd identified a vtable you would insert code to GEP
> element N and then pass the result to something like a printf
> expecting a %p. Of course you need to know in advance that your vtable
> contains 5 entries.
>
> So you'd create a bunch of instructions looking vaguely like:
>
> call void @printf(@WHATEVER_FOR_PERCENT_P, bitcast(i8**
> getelementptr inbounds([5 x i8*], [5 x i8*]* @_ZTV9Rectangle, i32 0,
> i32 N) to i8*)
>
> The chances are you won't always have a @_ZTV* global as a base though
> (that's the entire point of dynamic dispatch). In which case you'd
> have to create a bunch of instructions like:
>
> %tmp = getelementptr inbounds [5 * i8*], [5 x i8*]* %vtable, i32 0,
> i32 N
> %tmp1 = bitcast i8** %tmp to i8*
> call void @printf(@WHATEVER_FOR_PERCENT_P, %tmp1)
>
> Tim.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20170530/dafec7d9/attachment-0001.html>
------------------------------
Message: 10
Date: Tue, 30 May 2017 23:23:42 -0400
From: Damilola Orikogbo via llvm-dev <llvm-dev at lists.llvm.org>
To: Hans Wennborg <hans at chromium.org>, llvm-dev at lists.llvm.org
Subject: Re: [llvm-dev] Dumping the Vtable
Message-ID:
<CAJJs-5k1+GJPrDanMXE3-VPvFa6dBDhTqSnoJidyxhqyjzMXrw at
mail.gmail.com>
Content-Type: text/plain; charset="utf-8"
Hi Hans,
Thanks for the reply. I apologize for the unclear question. I am new to
using clang and llvm.
I have tried this but the version I have for some reason won't allow me to
use this command line option. Furthermore, I am trying to get this
information via the IR and writing a pass to get this information to use
at a later point.
-Dami
On Tue, May 30, 2017 at 12:56 PM, Hans Wennborg <hans at chromium.org>
wrote:
> Hi Damilola,
>
> On Mon, May 29, 2017 at 5:20 AM, Damilola Orikogbo via llvm-dev
> <llvm-dev at lists.llvm.org> wrote:
> > Can someone help me with how to dump the vtable and the vptr addresses
> > within it?
>
> I'm not sure if this answers your question exactly, but you can pass
> "-Xclang -fdump-vtable-layouts" to ask Clang to dump the layout
of
> generated vtables. For example:
>
> $ cat /tmp/a.cc
> struct S {
> virtual void f() {};
> };
>
> void f(S * s) {
> s->f();
> }
> $ bin/clang++ -c /tmp/a.cc -Xclang -fdump-vtable-layouts
> Vtable for 'S' (3 entries).
> 0 | offset_to_top (0)
> 1 | S RTTI
> -- (S, 0) vtable address --
> 2 | void S::f()
>
> VTable indices for 'S' (1 entries).
> 0 | void S::f()
>
>
> - Hans
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20170530/b8cc0e82/attachment-0001.html>