Hi Segher, > On 15 May 2019, at 18:24, Segher Boessenkool <seg...@kernel.crashing.org> > wrote: >
> On Wed, May 15, 2019 at 12:54:03PM +0100, Iain Sandoe wrote: >> If we build Darwin with a modern assembler, then it might well >> recognise insns that cannot be used on current Darwin systems. >> >> The patch augments the tests for feature support for VSX, >> power8 and power9 to exclude Darwin even if the assembler can >> handle the instructions. > > Cannot be used, in what way? *_ok test only if we can *compile* for > some subtarget. What blows up? It cISTR there were other problems too - I could revert the patch locally and remind myself (I am motivated to have parity between the cctools and newer assemblers in coverage on Darwin for now, and then to try expanding the horizons when the basics are working well). > The current tests aren't great, we should compile some actual code to > test if the compiler allows building for p8 for example -- more than > just the assembler. Maybe it is enough to have some register input > that requires VSX registers on the asm in the test, for example. Yes, that could be more general (AIX also has similar exclusions to Darwin), if the test results are cached it doesn’t hurt to try a bit harder. Iain
