a unified hard register set 0~63 is permitted to hold either interger or float
value.Thanks.
At 2017-04-18 21:45:54, "Matthew Fortune" wrote:
>comp writes:
>> Hi all,
>> I recently have a problem with LRA.
>> 1 The Bug use case
>> int a=10;
>> float
It does not work when -fno-strict-aliasing is specified.
At 2017-04-18 21:02:45, "Segher Boessenkool" wrote:
>Hi,
>
>On Mon, Apr 17, 2017 at 09:11:56PM +0800, comp wrote:
>> float b;
>> *(int*)&b=a;
>
>This is undefined behaviour. Try with -fno-strict-aliasing, does it
>then work
comp writes:
> Hi all,
> I recently have a problem with LRA.
> 1 The Bug use case
> int a=10;
> float c=2.0,d;
> main()
> {
> float b;
> *(int*)&b=a;
> d=b+c;
> }
>
> 2 The problem description
> In the pass LRA, curr_insn_t
Hi,
On Mon, Apr 17, 2017 at 09:11:56PM +0800, comp wrote:
> float b;
> *(int*)&b=a;
This is undefined behaviour. Try with -fno-strict-aliasing, does it
then work as you expect?
Segher
Can you file this as a bug report in GCC Bugzilla?
I assume that the target is Alpha?
Thanks, David
On Mon, Apr 17, 2017 at 9:11 AM, comp wrote:
>
> Hi all,
> I recently have a problem with LRA.
> 1 The Bug use case
> int a=10;
> float c=2.0,d;
> main()
> {
>
Hi all,
I recently have a problem with LRA.
1 The Bug use case
int a=10;
float c=2.0,d;
main()
{
float b;
*(int*)&b=a;
d=b+c;
}
2 The problem description
In the pass LRA, curr_insn_transform () deal with the addition statement d = b
+ c, the corresponding rtx expres
