video: don't use NEON blitters when SDL_HasNEON returns false

[madebr]

• I confirm that I am the author of this code and release it to the SDL project under the Zlib license. This contribution does not contain code from other sources, including code generated by a Large Language Model ("AI").

Description

Existing Issue(s)

[icculus]

Does this only need the SDL_NEON_INTRINSICS check and not all the arm8 etc stuff? Those extra checks are generally "we know this target always supports NEON," but we're changing this to check at runtime, so they probably need to go.

[madebr]

Does this only need the SDL_NEON_INTRINSICS check and not all the arm8 etc stuff? Those extra checks are generally "we know this target always supports NEON," but we're changing this to check at runtime, so they probably need to go.

I think so. The SDL_NEON_INTRINSICS macro is also used as guard in other locations.
But an actual arm user should verify this change.

[madebr]

I suppose the build failure answers your question :).
The blitters use 128-bit neon which apparently is arm8+.

[ccawley2011]

The __ARM_ARCH >= 8 check can be removed, since that should already be implied on AArch64.

[cgutman]

What's the motivation for this change? All of the platforms that pass the ifdef (ARMv8) should be guaranteed to have NEON.

[madebr]

SDL3 has a hint that allows disabling simd features at runtime: SDL_HINT_CPU_FEATURE_MASK.