__clear_cache fix for ARM 64 devices plus ANDROID cleanup

[fzurita]

This implements a fix for a bug in some arm 64 bit CPUs as seen here:

dolphin-emu/dolphin#4204

Also, removed some ANDROID specific code that is no longer needed.

@Gillou68310 you should review this please. I didn't see any performance penalties.

Should we attribute the source of the fix in the code?

[Gillou68310]

Well I don't like intrinsics but I guess I won't ask you to write it in assembly ;-)
You already gave the source in this PR I don't think it's necessary to add it in the source code.

[fzurita]

Feel free to rewrite it in assembly :-)

[Gillou68310]

Maybe later

[fzurita]

Ok to merge then?

[Gillou68310]

Yep

[fzurita]

@Gillou68310 I'm getting a warning from static analysis with clang tidy (It was added on Android Studio 3.3):

Line 4453: Local variable 'memtarget' might not have been initialized

It may explain all the breakpoint crashes I have been seeing in the play store.
signal 5 (SIGTRAP), code 1 (TRAP_BRKPT)
Thoughts?

This is in the storelr_assemble function.

[Gillou68310]

That uninitialized variable is bad and should be fixed but I don't think it's the cause of the breakpoint.

Currently SWL/SWR/SDL/SDR instructions have no fallback to interpreted opcode, which means that a store with no valid TLB map or an address outside RDRAM (eg: 0xa4000000) will trigger a breakpoint. This is true for all platform except ARM which has no breakpoint.

We really need to find the game that triggers this so I can implement interpreted fallback properly.

[fzurita]

Also, I'm still seeing some of these crashes with bad core dumps in arm64 devices:

signal 4 (SIGILL), code 1 (ILL_ILLOPC)

Any chance we are emitting some bad instruction? I'm not seeing it in arm32.

[Gillou68310]

that's possible indeed

[fzurita]

Too bad I can't find which game

[Gillou68310]

¯_(ツ)_/¯