On Thu, Sep 7, 2017 at 12:18 PM, Dibyendu Majumdar
<mobile@xxxxxxxxxxxxxxx> wrote:
> Hi Luc,
>
> On 7 September 2017 at 02:31, Luc Van Oostenryck
> <luc.vanoostenryck@xxxxxxxxx> wrote:
>> On Thu, Sep 07, 2017 at 12:05:58AM +0100, Dibyendu Majumdar wrote:
>>> I recently found another test case that fails:
>>>
>>> static int test_do(void) {
>>> int a = 0;
>>> int count = 27;
>>> switch (count % 8) {
>>> case 0: do { a++;
>>> case 7: a++;
>>> case 6: a++;
>>> case 5: a++;
>>> case 4: a++;
>>> case 3: a++;
>>> case 2: a++;
>>> case 1: a++;
>>> } while ((count -= 8) > 0);
>>> }
>>> if (27 != a) return 1;
>>> return 0;
>>> }
>>
>
> What's interesting is that I find that this code compiles OK with the
> old SSA implementation, with or without single store short-cut, and
> with out without simplifications.
>
> But the newssa implementation ( your original sssa-mini) fails when
> simplifications are turned on.
It's possible but it's just a coincidence. The code generated by the rc5
version is wrong and rc5 or newssa need the patch I send because of
what is done on phi-node operands in insert_branch()/remove_parent().
-- Luc
--
To unsubscribe from this list: send the line "unsubscribe linux-sparse" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
