On Wed, Jul 28, 2021 at 11:20:35AM +0100, Mark Brown wrote: > On Wed, Jul 28, 2021 at 10:41:30AM +0100, Dave Martin wrote: > > On Tue, Jul 27, 2021 at 07:06:47PM +0100, Mark Brown wrote: > > > SVE provides an instruction RDVL which reports the currently configured > > > vector length. In order to validate that our vector length configuration > > > interfaces are working correctly without having to build the C code for > > > > +.globl rdvl_sve > > > +rdvl_sve: > > > + rdvl x0, #1 > > > + ret > > > > This works, but can we use an ACLE intrinsic for this? I'm pretty GCC > > and LLVM have been up to date with that stuff for some time, though > > you'd have to check with the compiler folks. > > If we put SVE into C source files we need to build the C code with SVE > support enabled, I felt it was safer to avoid any possibility that we > might get some interference with the tests due to interactions with SVE > code generated by the compiler. It should be fine, but it's even easier > to not have to think about it at all. Good point, stick with the out-of-line asm in that case. (Also saves having to find out about the intrinsics!) Cheers ---Dave
