This is a respin of the series at
https://gcc.gnu.org/ml/gcc-patches/2015-04/msg01332.html, plus the two ARM
patches on which these depend
(https://gcc.gnu.org/ml/gcc-patches/2015-04/msg01333.html). These two somewhat
duplicate Charles Baylis' lane-bounds-checking patch at
https://gcc.gnu.org/ml/gcc-patches/2015-07/msg00140.html, in that they both port
some of the same AArch64 infrastructure onto ARM; while each has some parts the
other doesn't, there don't look to be any serious conflicts; if Charles' patches
were to go in first, I would not expect any major problems in rebasing mine over
his.
Changes since the first version of the float16 series are
* to separate out the (non-vector) tests from gcc.testsuite/aarch64 into a
.../fp16 subdirectory, as per
https://gcc.gnu.org/ml/gcc-patches/2015-05/msg00656.html
* dropped the patch rewriting advsimd-intrinsics.exp, following other changes
along similar lines by Sandra Loosemore and Christopher Lyon;
* Rebased over other testsuite changes, including dropping expected values from
many tests of intrinsics with no fp16 variant, and introducing a
CHECK_RESULTS_NO_FP16 macro.
* Changed the mechanism on ARM by which we passed in -mfpu=neon-fp16: we now try
to pass this into all tests, but fail the vcvt_f16.c if float16 is still not
supported (e.g. there was a conflicting -mfpu=neon passed to the compiler, or we
are running on HW which does not support the instructions).
Are these OK for trunk?
Thanks, Alan
