Kawrakow b66cecca45 Fused FFN_UP+FFN_GATE op (#741) ...

* Fused up+gate+unary for regular (not MoE) FFN - CPU

* WIP CUDA

* Seems to be working on CUDA

For a dense model we get 2-3% speedup for PP and ~0.6% for TG.

* Add command line option

This time the option is ON by default, and one needs to turn it
off via -no-fug or --no-fused-up-gate

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>

2025-08-31 18:16:36 +03:00

..

ggml-alloc.h

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

ggml-backend.h

Bug fixes from mainline (#439)

2025-05-20 17:03:14 +03:00

ggml-blas.h

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

ggml-cann.h

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

ggml-cuda.h

Merge mainline - Aug 12 2024 (#17)

2024-08-12 15:14:32 +02:00

ggml-kompute.h

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

ggml-metal.h

Merge mainline - Aug 12 2024 (#17)

2024-08-12 15:14:32 +02:00

ggml-rpc.h

Fix non rpc build error (#506)

2025-06-08 17:27:00 +03:00

ggml-sycl.h

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

ggml-vulkan.h

Vulkan: a fresh start (#608)

2025-07-15 08:03:13 +02:00

ggml.h

Fused FFN_UP+FFN_GATE op (#741)

2025-08-31 18:16:36 +03:00