Question about Sparse parsing - return statement

Question about Sparse parsing - return statement

[Dibyendu Majumdar]

Hi,

I notice that the return statement has a symbol associated with - the
ret_target. What is the meaning of this symbol?

Thanks and Regards
Dibyendu

Re: Question about Sparse parsing - return statement

[Linus Torvalds]

On Mon, Aug 21, 2017 at 2:53 PM, Dibyendu Majumdar
<mobile@majumdar.org.uk> wrote:
>
> I notice that the return statement has a symbol associated with - the
> ret_target. What is the meaning of this symbol?

Sparse uses symbols for branch targets too.

There's several special target symbols - the return target, the
break/continue targets for loops and switch statements etc.

For example, a regular label (ie a target of a "goto") is a symbol
with the identifier being the label.  Amd the return point is
basically just another label that has the identifier "return".

So a "return statement" is basically the same as "goto return-label"
(plus the expression handling if there is one, of course).

Note that there can be *many* such return points as a function is
generated, since each inlined function will have its return point
pointing to the place in the calling function where it will return.
The same way each 'switch' statement will have its "break" label, and
each "for/.while/do" loop will have it's own "continue" and "break"
labels.

Note how each label is always described by a symbol, and how you can
look up what "break" means (regardless of whether you're in a switch
or a loop context) by just looking up the "break" label in the
NS_ITERATOR namespace.

In fact, it should be noted that the code to handle "break" and
"continue" is *EXACTLY*THE*SAME*. That should blow your mind. They
both use exactly the same logic, and they just do exactly the same
thing: look up the symbol associated with that ident, and "goto" it.

(I may be off in some of the details, it's way too long since I looked
at this code).

The return symbol is also the exit point of a compound statement, so
there's a lot of just commonalities here: any branch target always is
a symbol that is associated with the basic block that is the target.

         Linus

Re: Question about Sparse parsing - return statement

[Linus Torvalds]

On Mon, Aug 21, 2017 at 4:40 PM, Linus Torvalds
<torvalds@linux-foundation.org> wrote:
>
> Sparse uses symbols for branch targets too.

Side note: when you work with the linearized representation, I don't
think you should really need to care. You should be able to just work
with the basic block information, and the symbols that are used to
generate the branches between the basic blocks should be pretty much a
non-issue by the time it's all been linearized.

But maybe I'm missing or have forgotten something.

            Linus

Re: Question about Sparse parsing - return statement

[Dibyendu Majumdar]

Hi Linus,

On 22 August 2017 at 00:40, Linus Torvalds
<torvalds@linux-foundation.org> wrote:
> On Mon, Aug 21, 2017 at 2:53 PM, Dibyendu Majumdar
> <mobile@majumdar.org.uk> wrote:
>>
>> I notice that the return statement has a symbol associated with - the
>> ret_target. What is the meaning of this symbol?
>
> Sparse uses symbols for branch targets too.
>
> There's several special target symbols - the return target, the
> break/continue targets for loops and switch statements etc.
>
> For example, a regular label (ie a target of a "goto") is a symbol
> with the identifier being the label.  Amd the return point is
> basically just another label that has the identifier "return".
>
> So a "return statement" is basically the same as "goto return-label"
> (plus the expression handling if there is one, of course).
>

Okay thanks.

> Note that there can be *many* such return points as a function is
> generated, since each inlined function will have its return point
> pointing to the place in the calling function where it will return.
> The same way each 'switch' statement will have its "break" label, and
> each "for/.while/do" loop will have it's own "continue" and "break"
> labels.
>
> Note how each label is always described by a symbol, and how you can
> look up what "break" means (regardless of whether you're in a switch
> or a loop context) by just looking up the "break" label in the
> NS_ITERATOR namespace.
>
> In fact, it should be noted that the code to handle "break" and
> "continue" is *EXACTLY*THE*SAME*. That should blow your mind. They
> both use exactly the same logic, and they just do exactly the same
> thing: look up the symbol associated with that ident, and "goto" it.
>

Cool. I haven't got around to breaks and continues yet.

> (I may be off in some of the details, it's way too long since I looked
> at this code).
>
> The return symbol is also the exit point of a compound statement, so
> there's a lot of just commonalities here: any branch target always is
> a symbol that is associated with the basic block that is the target.
>

Okay makes sense.

Regards
Dibyendu