|
From: Florian K. <br...@ac...> - 2012-09-05 18:17:46
|
On 09/03/2012 01:51 PM, Julian Seward wrote:
>
>>> And
>>> also, for people using the Intel compilers, the Intel runtime library
>>> messes with the FPU settings in a way which cause an emwarn at startup
>>> each time. Along the lines of "it is asking for a fast-but-inaccurate
>>> FP mode (RAZ/DAZ), but I can't do that, so I'm going to remain in the
>>> slower-but-accurate mode".
>>
>> OK, I get that and see why you don't want to change the default.
>
> I must admit I am not enthusiastic to do so. Although it occurred to
> me later, if it is important for you, what we could do is have per-emwarn
> show/no-show defaults, so we can hide the x86 etc ones by default but
> display the s390 ones.
It is not all that important, as those warnings could only be triggered
by hand-written assembly.
Having per-emwarn show/no-show defaults adds too much complexity imho.
If we want to provide more flexibility here I'd say: let's move the
control of emulation warnings to the various frontends. They would know
best what the proper default for a warning is and how often it should be
issued.
There would be a little code duplication, of course. But nothing to
worry about. Perhaps a bit more complex is to get the initial value of
VG_(clo_show_emwarns) from VEX (because the frontends now control it) at
the right time. Not sure. Haven't looked at it.
Florian
|