Support 128 bit integer MAD benchmark? (NV only)

[oscarbg]

Hi,
don't know if a good idea to add to this benchmark, as it's NV only for the moment,
but NV OpenCL 3.0 implementation since 510.xx drivers supports int128 type, and altough it's a non native integer format it's supported, and supposedly uses efficient PTX instructions for carry,etc..
curious to know perf on a MAD benchmark..

requires using new OpenCL NVVM 7.0 compiler using:
set NVCL_USE_NVVM70_COMPILER=1 or
export “NVCL_USE_NVVM70_COMPILER=1 on Linux

shown in "511.23-win11-win10-release-notes.pdf":
"128-bit integer types or “(un)signed long long” is available as a native data type in the new
compiler. This type is enabled by default and does not require any macros to be defined"