Re: [PATCH v2] rs6000: Fix some issues in rs6000_can_inline_p [PR102059]

[Peter Bergner <bergner@linux.ibm.com>]

On 12/6/21 7:06 AM, Segher Boessenkool wrote:
> On Fri, Dec 03, 2021 at 11:46:53AM +0800, Kewen.Lin wrot
>> Without this fix, bar inlines foo but get the error as below:
>>
>> In function ‘foo’,
>>     inlined from ‘bar’ at test.c:8:8:
>> test.c:3:9: error: ‘__builtin_ttest’ requires the ‘-mhtm’ option
>>     3 |   *b += __builtin_ttest();
>>       |         ^~~~~~~~~~~~~~~~~
>>
>> Since bar doesn't support foo's required HTM feature.
>>
>> With this fix, bar doesn't inline foo and the case gets compiled well.
> 
> And if you inline something no-htm into something that allows htm?  That
> should work fine, be allowed just fine.

It is only ok to inline a function to be compiled with no-htm into a
function that will be compiled with htm if the no-htm function is
being compiled with no-htm by default.  If the user explicitly used
-fno-htm on the function, then we should not inline the htm function
into it.  Ditto for the other way around.  Meaning, the caller's
option flags should be a superset of the callee's flags, but the
explicit flags used on both functions need to match exactly.

Peter