[PATCH] AArch32 v8 NEON intrinsics for numeric max/min and directed rounding to integral

[Tim Northover]

Hi Graham,

This mostly looks fine. Just one comment on the tests:

+float32x2_t test_vrnda_f32(float32x2_t a) {
+  // CHECK-LABEL: test_vrnda_f32
+  // CHECK-LABEL: call <2 x float> @llvm.arm.neon.vrinta.v2f32(<2 x float> %a)
+  return vrnda_f32(a);
+}

That second CHECK-LABEL should probably be just CHECK. The CHECK-LABEL
directive is there to split the test up into distinct regions (usually
functions) so that we don't accidentally find what we're looking for
in an unrelated piece of output. And to make sure we get told about it
more accurately when things go wrong.

Cheers.

Tim.