My AMD M890 means I cannot use llama.cpp?

[resistorsoftware]

I built llama.cpp for my Arch based AMD system, and when I use llama-serve to run a model in OpenAI format and connect a chat to it using Opencode, the result is pretty much what I expected:

`rocBLAS error: Cannot read /opt/rocm/lib/../lib/rocblas/library/TensileLibrary.dat: No such file or directory for GPU arch : gfx1150'

So my machine has no support for gfx1150. I am sad. I am hopeful! Does anyone with any experience with llama.cpp have any guidance on when maybe this will be a filled in library? Is this mostly on AMD? I am curious! Would love to use llama.cpp!

Interesting Note

If I just start the llama-serve process for the same model, and do a curl POST with the same content, I get back a proper response with no error! So llama-serve does indeed accept content to process without incurring this Tensile Library error? How is this explained to a n00b like me? Does this mean it is truly something in Opencode passing content to the llama-serve? If so, what can I maybe alter in my settings so as to not incur this error?