Oops! Didn't do this quite right the first time round.
This commit fixes RETRO_DEVICE_ID_POINTER_PRESSED, which would always
return 0 due to to an implicit case to int16_t. Basically, we'd do
(val & 0x00080000) & 0xFFFF; which would return 0 every time. Fixed that
by wrapping it in a ternary. Yes, I know we could use a rotation, but
for a boolean value it really doesn't matter.
I also rewrote scaleTP to deal entirely in integers. While the
floating-point math was theoretically faster on PowerPC; it gets awkward
to cast -0x7FFF to a float.
Speaking of, the driver now actually conforms to the libretro API. Not
sure how I managed to not see the spec; but hey, now its fixed.
RETRO_DEVICE_POINTER_ID_X/Y will now return values between -0x7FFF and
0x7FFF like they're supposed to.
Big thanks to @r-type for hounding me to fix this.
Partially addresses #5294; we still need mouse emulation.
a3e4a7612f