optimization for t194 should use gcc or clang? 
i find out is maybe better to use clang cause by clang-11 and higher have mcpu/mtune=carmel for t194(xavier)
however gcc does not, even gcc-13 does not
but this means users maybe have to install llvm stuff
so i open this issue to confirm should i change to mtune/mcpu=carmel or keep like now(cortex-a76) or make both options(like TEGRA_T194_CLANG and TEGRA_T194_GCC, this is just example and t234 is not affected since is cortex-a based(note: t2XX are cortex-a based, but t1XX is not)

cortex-a76 is suggested by nvidia staff on dev fourm
https://forums.developer.nvidia.com/t/gcc-mcpu-mtune-for-carmel-cpu/146345/6?u=leonpano