How to change the llama.cpp to T-mac(LUT-based)?I successfully run the code from the readme in my M1 computer, but the method is using llama.cpp don't the T-mac(LUT-based) from the thesis.

[jixiuy]

[jixiuy]

How to change the llama.cpp to T-mac(LUT-based)?I successfully run the code from the readme in my M1 computer, but the method is using llama.cpp don't the T-mac(LUT-based) from the thesis. #91

[Promortyus]

I met the same problem on M3 Mac. Have you solved it？

[type-shangshu]

The ./3rdparty/llama.cpp after successfully running README is already a LUT-based method. You can verify this with perf in Linux: the overhead of function "tbl_..." exceeds that of other functions, ranking 1st. The "tbl..." function is the LUT-based method.