MacPorts clang performance (and arm_runtime)

[Lawrence Velázquez]

On Dec 20, 2014, at 3:31 PM, Ryan Schmidt <ryandesign at macports.org> wrote:

> On Dec 20, 2014, at 12:17 PM, René J.V. Bertin wrote:
> 
>>> port variants llvm-3.6
>>> 
>>> llvm-3.6 has the variants:
>>> [+]assertions: Enable assertions for error detection (has performance
>>> impacts, especially on JIT)
>> 
>> So the default variant, and the only one available as a binary package, is built to be slow and crash upon certain errors.
>> Why is that the default?
> 
> Presumably because llvm-3.6 is a development version, and if problems are encountered, the port maintainer would like to receive the most helpful bug report possible. In past llvm versions, the maintainer has removed this as a default variant once the port reaches stable status.

If one cares about stability and performance in a compiler, one should not be using a developmental compiler.

>> Same with +arm_runtime, btw. I don't know how useful a MacPorts compiler is for building for arm (iOS development I presume), and how many people do that, but it'd be nice to have a binary version without. I can remember having had issues with the presence of an arm runtime in the past.
> 
> I'm not aware of any issues caused by having that variant selected. If it's causing you problems, file a ticket and explain.

I'm actually seriously thinking about forcing +arm_runtime if the iOS SDK is present and forcing -arm_runtime otherwise. The only reason Jeremy added that variant was to deal with configurations that are missing the iOS SDK (i.e., Xcode older than 4.x); the ability to deselect the variant is incidental.

vq