On Mar 1, 2009, at 10:58 AM, Anders Carlsson wrote:
> 1 mar 2009 kl. 10.51 skrev Mike Stump:
>
>> On Feb 28, 2009, at 5:09 PM, Anders Carlsson wrote:
>>> + // Check if the block can be global.
>>> + if (CanBlockBeGlobal(Info))
>>> + return CGM.GetAddrOfGlobalBlock(BE, Name.c_str());
>>
>> Like the last change, this also breaks things that were working. :-
>> ( Please, monotonic improvements only please. We lazily create
>> imports during codegen and your code doesn't account for that.
>
> I think it is better to just use one code path for creating global
> blocks.
Why? It replicates the other code entirely.
> I added code to look up imports beforehand.
And my point it, is is wrong. It is slow, I don't want to walk around
looking for all the data you walk around for and it is unnecessary and
you broke previously working code like:
#include <stdio.h>
void foo(int i) {
printf("%d\n", i);
}
int main() {
__block int i; int j;
j = 941;
i = 942;
&i;
printf ("i is %d, j is %d\n", i, j);
^{^{ static int i; i = 943; printf ("i is %d, j is %d\n", i,
j); }();
printf ("i is %d, j is %d\n", i, j); }();
^{ ^{ j; }(); }();
printf ("i is %d, j is %d\n", i, j);
}
_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
