llvm dev - May 2014 - [LLVMdev] Changing the design of GlobalAliases to represent what is actually possible in object files.

On Fri, May 23, 2014 at 3:06 PM, Rafael Avila de Espindola
<rafael.espindola at gmail.com> wrote:
>
>
> Sent from my iPhone
>
>> On May 23, 2014, at 17:51, Eric Christopher <echristo at
gmail.com> wrote:
>>
>> I'm not there yet, but at some point I'm going to need the
notion of a
>> global callable function like symbol that's resolved at runtime.
I've
>> not given it much thought but I may need a new callable entity here
>> (this is for the gnu ifunc stuff).
>>
>> Don't even know if this fits into the discussion, but since we were
>> talking about weird symbols...
>>
>
> It is a symbol or a value that is loaded? If it is an symbol, what does it
point to? That is, what is the value that shows up in the .o?
Relocation resolved at compile time to one of N symbols that will be
the call that's made.

-eric
>
>
>> -eric
>>
>>> On Fri, May 23, 2014 at 1:26 PM, Reid Kleckner <rnk at
google.com> wrote:
>>> On Thu, May 22, 2014 at 7:14 PM, Rafael Espíndola
>>> <rafael.espindola at gmail.com> wrote:
>>>>
>>>>> For example, it would not be possible to define an absolute
symbol to be
>>>>> the offset between two symbols.  In some restricted
circumstances —
>>>>> if both symbols are global variables in the same section
and defined
>>>>> in the same translation unit — this could be worked around.
>>>>>
>>>>> But I’ll gladly admit that I don’t have a use case in mind
for that
>>>>> feature.
>>>>> Absolute symbols are useful, and storing offsets between
symbols into
>>>>> global memory is useful, but I don’t know why you’d combine
them.
>>>>
>>>> That is funny. I, on the other hand, think that this is the
best
>>>> argument I have seen for keeping aliases pointing to
ConstantExpr so
>>>> far.
>>>
>>>
>>> IMO if we want to support defining symbols at absolute addresses,
we should
>>> add a separate construct for this.  So far everyone has gotten by
with
>>> linker flags and scripts, though.
>>>
>>> _______________________________________________
>>> LLVM Developers mailing list
>>> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
>>> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>>>

According to

http://www.airs.com/blog/archives/403
>From the compiler & assembler point of view this is just a special
symbol type.
Given that it cannot be used directly I would say it is not a
ConstantExpr, it is a property of a label.

Given that the special IFUNC label has the same value as the resolver,
it sounds like it is not a GlobalObject, just a special type of
GlobalAlias.

In summary, from a very quick look it looks like it is independent
from this discussion: All that is needed is a bit saying it is a
"IFUNC GlobalAlias", not a plain GlobalAlias, regardless of how we end
up representing GlobalAliases.
> On May 23, 2014, at 18:15, Eric Christopher <echristo at gmail.com>
wrote:
>
> On Fri, May 23, 2014 at 3:06 PM, Rafael Avila de Espindola
> <rafael.espindola at gmail.com> wrote:
>>
>>
>> Sent from my iPhone
>>
>>> On May 23, 2014, at 17:51, Eric Christopher <echristo at
gmail.com> wrote:
>>>
>>> I'm not there yet, but at some point I'm going to need the
notion of a
>>> global callable function like symbol that's resolved at
runtime. I've
>>> not given it much thought but I may need a new callable entity here
>>> (this is for the gnu ifunc stuff).
>>>
>>> Don't even know if this fits into the discussion, but since we
were
>>> talking about weird symbols...
>>
>> It is a symbol or a value that is loaded? If it is an symbol, what does
it point to? That is, what is the value that shows up in the .o?
>
> Relocation resolved at compile time to one of N symbols that will be
> the call that's made.
>
> -eric
>
>>
>>
>>> -eric
>>>
>>>> On Fri, May 23, 2014 at 1:26 PM, Reid Kleckner <rnk at
google.com> wrote:
>>>> On Thu, May 22, 2014 at 7:14 PM, Rafael Espíndola
>>>> <rafael.espindola at gmail.com> wrote:
>>>>>
>>>>>> For example, it would not be possible to define an
absolute symbol to be
>>>>>> the offset between two symbols.  In some restricted
circumstances —
>>>>>> if both symbols are global variables in the same
section and defined
>>>>>> in the same translation unit — this could be worked
around.
>>>>>>
>>>>>> But I’ll gladly admit that I don’t have a use case in
mind for that
>>>>>> feature.
>>>>>> Absolute symbols are useful, and storing offsets
between symbols into
>>>>>> global memory is useful, but I don’t know why you’d
combine them.
>>>>>
>>>>> That is funny. I, on the other hand, think that this is the
best
>>>>> argument I have seen for keeping aliases pointing to
ConstantExpr so
>>>>> far.
>>>>
>>>>
>>>> IMO if we want to support defining symbols at absolute
addresses, we should
>>>> add a separate construct for this.  So far everyone has gotten
by with
>>>> linker flags and scripts, though.
>>>>
>>>> _______________________________________________
>>>> LLVM Developers mailing list
>>>> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
>>>> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>>>>

On Fri, May 23, 2014 at 6:36 PM, Rafael Espíndola
<rafael.espindola at gmail.com> wrote:
> According to
>
> http://www.airs.com/blog/archives/403
>
> From the compiler & assembler point of view this is just a special
symbol type.
>
> Given that it cannot be used directly I would say it is not a
> ConstantExpr, it is a property of a label.
>
> Given that the special IFUNC label has the same value as the resolver,
> it sounds like it is not a GlobalObject, just a special type of
> GlobalAlias.
>
> In summary, from a very quick look it looks like it is independent
> from this discussion: All that is needed is a bit saying it is a
> "IFUNC GlobalAlias", not a plain GlobalAlias, regardless of how
we end
> up representing GlobalAliases.
Sounds fine. I hadn't thought about it more than "do we want an alias
or a new symbol type" but since the discussion was ongoing here I
thought I'd bring it up.

-eric
>
>> On May 23, 2014, at 18:15, Eric Christopher <echristo at
gmail.com> wrote:
>>
>> On Fri, May 23, 2014 at 3:06 PM, Rafael Avila de Espindola
>> <rafael.espindola at gmail.com> wrote:
>>>
>>>
>>> Sent from my iPhone
>>>
>>>> On May 23, 2014, at 17:51, Eric Christopher <echristo at
gmail.com> wrote:
>>>>
>>>> I'm not there yet, but at some point I'm going to need
the notion of a
>>>> global callable function like symbol that's resolved at
runtime. I've
>>>> not given it much thought but I may need a new callable entity
here
>>>> (this is for the gnu ifunc stuff).
>>>>
>>>> Don't even know if this fits into the discussion, but since
we were
>>>> talking about weird symbols...
>>>
>>> It is a symbol or a value that is loaded? If it is an symbol, what
does it point to? That is, what is the value that shows up in the .o?
>>
>> Relocation resolved at compile time to one of N symbols that will be
>> the call that's made.
>>
>> -eric
>>
>>>
>>>
>>>> -eric
>>>>
>>>>> On Fri, May 23, 2014 at 1:26 PM, Reid Kleckner <rnk at
google.com> wrote:
>>>>> On Thu, May 22, 2014 at 7:14 PM, Rafael Espíndola
>>>>> <rafael.espindola at gmail.com> wrote:
>>>>>>
>>>>>>> For example, it would not be possible to define an
absolute symbol to be
>>>>>>> the offset between two symbols.  In some restricted
circumstances —
>>>>>>> if both symbols are global variables in the same
section and defined
>>>>>>> in the same translation unit — this could be worked
around.
>>>>>>>
>>>>>>> But I’ll gladly admit that I don’t have a use case
in mind for that
>>>>>>> feature.
>>>>>>> Absolute symbols are useful, and storing offsets
between symbols into
>>>>>>> global memory is useful, but I don’t know why you’d
combine them.
>>>>>>
>>>>>> That is funny. I, on the other hand, think that this is
the best
>>>>>> argument I have seen for keeping aliases pointing to
ConstantExpr so
>>>>>> far.
>>>>>
>>>>>
>>>>> IMO if we want to support defining symbols at absolute
addresses, we should
>>>>> add a separate construct for this.  So far everyone has
gotten by with
>>>>> linker flags and scripts, though.
>>>>>
>>>>> _______________________________________________
>>>>> LLVM Developers mailing list
>>>>> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
>>>>> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>>>>>