Re: [Open64-devel] Source to source compilation

[Costin I. <CC...@lb...>]

As far as the internal representation per-se, I recall having to add  
new type attributes to keep around the original C types.
For example "long" becomes internally INT32 or INT64.  Hence a   
TY_is_long attribute.  "long" may have already been in there but
there were a lot of problems with the different flavors of "char"...


> Wondering if you could say what internal representation changes were  
> necessary to make whirl2c work?
>
> Thanks
> Vinod
>
>
>
> On Tue, Apr 29, 2008 at 3:20 PM, Costin Iancu <cc...@lb...> wrote:
> As Mike Murphy pointed out, the Berkeley sources fixed a lot of C  
> compatibility problems.
> This particular one has been fixed. From what I remember the changes  
> were not only in w2c, but also in the internal representation and  
> front-end in
> order to keep the original type info around.
>
> Depending on your intentions, one alternative for you is to use our  
> software directly. ( http://upc.lbl.gov )
> It's based on an earlier release of open64. We might be
> missing a lot of codegen (cg) changes but I think the lno, wopt and  
> front-ends are close enough in sync.
>
> Costin
>
> On Apr 29, 2008, at 2:07 PM, Dasarath Weeratunge wrote:
>
>>> Not so sure there is a "last" point where the C representation  
>>> would still
>>> be valid. I expect you have to fix whirl2c. What are the problems  
>>> you are
>>> seeing? Some small'ish things, or a large number of major problems?
>>>
>>> Robert
>>
>>  struct node *n = head;
>>
>>  if (n == NULL)
>>    head = m;
>>  else {
>>    while (n->next)
>>      n = n->next;
>>    n->next = m;
>>  }
>>
>>
>> Here's the generated code. In the while loop,
>> instead of referring to the next pointer, it does an
>> array reference and then tries to cast the structure anon3[1]
>> to an integer?
>>
>>
>> union var_pool {
>>  struct ty {
>>    _INT32 int_pool[16LL];
>>    struct node {
>>      _INT32 * item3;
>>      struct node * next;
>>    } node_pool[16LL];
>>  } u0;
>> } var_pool;
>>
>>
>>
>>    anon3 = head;
>>    if((_UINT64)(head) == (_UINT64)(0U))
>>    {
>>      head = (struct node *) rr1;
>>    }
>>    else
>>    {
>>      while((_UINT64)((anon3)[1LL]) != (_UINT64)(0U))
>>      {
>>        tmp2 = (anon3)[1LL];
>>        anon3 = *(struct node **) & tmp2;
>>      }
>>      tmp3 = (struct node *) rr1;
>>      (anon3)[1LL] = *(struct node *) & tmp3;
>>    }
>>
>>
>> 1.w2c.c:33: warning: assignment from incompatible pointer type
>> 1.w2c.c:83: error: aggregate value used where an integer was expected
>>
>> Thanks.
>> Dineth
>>
>>
>> Quoting Robert Hundt <rob...@gm...>:
>>
>>
>>>
>>> On Tue, Apr 29, 2008 at 1:34 PM, Dasarath Weeratunge <din...@gm... 
>>> >
>>> wrote:
>>>
>>>> I compiled a C program with opencc -clist -ipa -O3 and then tried
>>>> compiling
>>>> the generated C source. However, the generated source cannot be
>>>> recompiled.
>>>>
>>>> Could you please tell me what's that last point in be, that I may
>>>> generate legal C source through whirl2c.
>>>>
>>>> If opencc performs so far as loop analysis, that suffices for my  
>>>> current
>>>> work.
>>>>
>>>> Thank you.
>>>>
>>>> Dineth
>>>>
>>>> -------------------------------------------------------------------------
>>>> This SF.net email is sponsored by the 2008 JavaOne(SM) Conference
>>>> Don't miss this year's exciting event. There's still time to save  
>>>> $100.
>>>> Use priority code J8TL2D2.
>>>>
>>>>
>>> http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
>>>> _______________________________________________
>>>> Open64-devel mailing list
>>>> Ope...@li...
>>>> https://lists.sourceforge.net/lists/listinfo/open64-devel
>>>>
>>>
>>>
>>>
>>> --
>>> Robert Hundt
>>>
>>
>> -------------------------------------------------------------------------
>> This SF.net email is sponsored by the 2008 JavaOne(SM) Conference
>> Don't miss this year's exciting event. There's still time to save  
>> $100.
>> Use priority code J8TL2D2.
>> http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
>> _______________________________________________
>> Open64-devel mailing list
>> Ope...@li...
>> https://lists.sourceforge.net/lists/listinfo/open64-devel
>
> -- 
> Costin C.  
> Iancu                                                       cc...@lb...
> Future Technologies Group                                 Phone:  
> 510-495-2122
> Lawrence Berkeley National Laboratory           Fax:   510-486-6900
>
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by the 2008 JavaOne(SM) Conference
> Don't miss this year's exciting event. There's still time to save  
> $100.
> Use priority code J8TL2D2.
> http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
> _______________________________________________
> Open64-devel mailing list
> Ope...@li...
> https://lists.sourceforge.net/lists/listinfo/open64-devel
>
>

-- 
Costin C. Iancu                                                       cc...@lb...
Future Technologies Group                                 Phone:  
510-495-2122
Lawrence Berkeley National Laboratory           Fax:   510-486-6900