llama.cpp
c54bba86 - ggml : optimize cuda cumsum fallback kernel (#18343)

Commit

16 days ago

ggml : optimize cuda cumsum fallback kernel (#18343)

References

#18343 - ggml : optimize cuda cumsum fallback (~2.5x speedup vs CUB)

Author

Aadeshveer

Aadeshveer

Parents

Loading