<div dir="ltr"><div style>Hi folks,</div><div style><br></div><div style>I just "fixed" a bug on ARM LNT regarding lowering of a VMUL.f32 as NEON and not VFP. The former is not IEEE 754 compliant, while the latter is, and that was failing TSVC.</div>
<div style><br></div><div style>The question is:</div><div style> * is this a problem with the test, that shouldn't be expecting values below FLT_MIN, or</div><div style> * is it a bug in the lowering, that should only be lowering to NEON's VMUL when unsafe-math is on, or</div>
<div style> * neither, and people should disable that when they want correctness?</div><div><br></div><div style>The bugs in question...</div><div style><br></div><div style>Possible unsafe maths bug in NEON VMUL.f32<br></div>
<a href="http://llvm.org/bugs/show_bug.cgi?id=15546">http://llvm.org/bugs/show_bug.cgi?id=15546</a><br><div><br></div><div>MultiSource/Benchmarks/TSVC/Reductions-flt tests beyond FLT_MIN<br></div><div><a href="http://llvm.org/bugs/show_bug.cgi?id=15545">http://llvm.org/bugs/show_bug.cgi?id=15545</a><br>
</div><div><br></div><div style>cheers,</div><div style>--renato</div></div>