[PATCH] Add a "probe-stack" attribute

[John Kåre Alsaker]

Bump

On Sat, Apr 11, 2015 at 12:11 AM, John Kåre Alsaker
<john.mailinglists at gmail.com> wrote:
> Bump.
>
> On Sat, Mar 21, 2015 at 8:33 PM, John Kåre Alsaker
> <john.mailinglists at gmail.com> wrote:
>> Here's a rebased patchset.
>>
>> On Sun, Feb 22, 2015 at 5:01 PM, John Kåre Alsaker
>> <john.mailinglists at gmail.com> wrote:
>>> Bump
>>>
>>> On Sat, Jan 17, 2015 at 9:57 AM, John Kåre Alsaker
>>> <john.mailinglists at gmail.com> wrote:
>>>> I have a patch to add __probestack to compiler-rt. No other such
>>>> function have a configurable name, so I don't see why __probestack
>>>> should be different.
>>>>
>>>> I'm not sure how to test for #3 as I don't think any calling
>>>> convention has RAX as live-in.
>>>>
>>>> I don't know if #4 even handles the edge case(s), which is why I added
>>>> an assertion, so I don't know how to test for those. I can add a test
>>>> for the non-edge case though.
>>>>
>>>> The stack-probe-size attribute interacts as expected (it just controls
>>>> the stack allocation size for which probes are required).
>>>>
>>>> On Fri, Jan 16, 2015 at 10:30 PM, Reid Kleckner <rnk at google.com> wrote:
>>>>> On Fri, Jan 16, 2015 at 12:43 PM, David Majnemer <david.majnemer at gmail.com>
>>>>> wrote:
>>>>>>
>>>>>> Reid, I'm not so sure.
>>>>>>
>>>>>> We already have a probe-stack-size attribute, how does this attribute
>>>>>> interact with that one?  Which targets have a __probestack that can be
>>>>>> called?
>>>>>
>>>>>
>>>>> There aren't any targets that use __probestack. It's new this patch. Maybe
>>>>> the attribute should take the symbol name as a value so it's user
>>>>> controlled?