Neon Intrinsics

Why don’t the math libraries supplied with the SDK take advantage of the SIMD Intrinsics that are provided with Imagination’s chip set? I was curious if there was a math lib available from you guy’s that utilizes these.

I understand for clarity that the demo’s might not have that for learning purposes. Is there any other reason?

Perhaps because a lot of hardware out there doesn’t have Neon oboard.

BTW … including Neon intristics inside typical C++ Vector implementations won’t buy you much.

Neon ops suffer from pretty high latency and where Neon really shines is with custom ASM code driving custom “streaming” like processing routines where one can operate at 2+ vectors/quats at a time with long data streams where, with proper coding, you can actually hide latencies.