Hi.
It seems to me also that the info about live vars printed out by CIL is
correct.
I have to agree that things are actually a bit more complex: we should
specify
clearly what live variable sets are we talking about. For example, Aho, Lam,
Sethi, Ullman
are talking in the dragon book on Compilers (2nd, ed 2006) about two types of
live
variable sets:
IN[B] = the set of vars with the val at entry in B used along
some path in the flow
starting from B onwards
OUT[B] = the set of vars with the val at entry in B used along
some path in the flow
starting from B onwards
CIL seems to present the IN sets. For all examples I ran the live variable
sets (the
IN sets) are OK.
A less important issue: it is not very clear how I can easily make CIL
print out the
OUT sets.
One more thing: does anybody know where are used the livenessVisitorClass
and
deadnessVisitorClass classes?
Best regards,
Alex
On 10/25/2012 4:56 PM, Jesse M Draper wrote:
> I think that Cil is correct here. It is reporting i as live because you use
> it before you set it
> (i+=n). If you change n=10 to n=10,i=0, Cil reports neither i nor n as live
> at 1. I think that this
> is correct for liveness analysis.
>
> Jesse Draper
>
> On 10/25/2012 06:13 AM, Alex Susu wrote:
>> Hello.
>> I have some issues with the liveness module of CIL.
>>
>> I used it a bit on a couple of simple C programs and the results I
>> get are a bit strange.
>>
>> More exactly, for the program below:
>> int main() {
>> int n;
>> int i;
>>
>> for (n = 10; n> 0; n--) {
>> i += n;
>> }
>> return 0;
>> }
>>
>> when I give:
>> cilly.asm.exe --out ./app2.cil.c --doLiveness --live_func main
>> --live_debug ./app2.i
>>
>> I get the following live variable sets for each Basic-Block (BB):
>> 1: i___0(int ), Note: here we should have n(int ), as
>> well
>> 2: n(int ),i___0(int ),
>> 3: n(int ),i___0(int ),
>> 4:
>> 5: n(int ),i___0(int ),
>> 6:
>> So the problem is for this program that for BB 1, we have live var
>> set {i} and we
>> should have {i, n}.
>>
>> Note that in the liveness module I added the instruction
>> (Cfg.printCfgFilename
>> "myfunc_cfg.dot" fd), which created the following myfunc_cfg.dot file:
>> digraph CFG_main {
>> 1 [label="1: instr"] Note: BB 1 is instr n = 10
>> 1 -> 2
>> 2 [label="2: loop"] Note: BB 2 is the while loop - not really a
>> BB, IMO
>> 2 -> 3
>> 3 [label="3: if"] Note: BB 3 is the if n> 0
>> 3 -> 5
>> 3 -> 4
>> 4 [label="4: break"]
>> 4 -> 6
>> 5 [label="5: instr"] Note: BB 5 is statements i += n; n--;
>> 5 -> 2
>> 6 [label="6: return"]
>> }
>>
>> Best regards,
>> Alex
>>
>> ------------------------------------------------------------------------------
>> Everyone hates slow websites. So do we.
>> Make your web apps faster with AppDynamics
>> Download AppDynamics Lite for free today:
>> http://p.sf.net/sfu/appdyn_sfd2d_oct
>> _______________________________________________
>> CIL-users mailing list
>> CIL-users@lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/cil-users
>
>
> ------------------------------------------------------------------------------
> Everyone hates slow websites. So do we.
> Make your web apps faster with AppDynamics
> Download AppDynamics Lite for free today:
> http://p.sf.net/sfu/appdyn_sfd2d_oct
> _______________________________________________
> CIL-users mailing list
> CIL-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/cil-users
>
------------------------------------------------------------------------------
Everyone hates slow websites. So do we.
Make your web apps faster with AppDynamics
Download AppDynamics Lite for free today:
http://p.sf.net/sfu/appdyn_sfd2d_oct
_______________________________________________
CIL-users mailing list
CIL-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/cil-users
