BDEPEND=>=dev-util/nvidia-cuda-toolkit-13.2.2 app-alternatives/ninja >=dev-build/cmake-3.28.5
DEFINED_PHASES=compile configure install prepare pretend setup test
DEPEND=>=dev-util/nvidia-cuda-toolkit-13.2.2:= openmp? ( llvm-runtimes/openmp:= )
DESCRIPTION=CUDA llama.cpp runtime with tiered KV memory for long contexts
EAPI=8
HOMEPAGE=https://github.com/kvmem/kvmem-llama.cpp
IUSE=cpu_flags_x86_avx cpu_flags_x86_avx2 cpu_flags_x86_bmi2 cpu_flags_x86_f16c cpu_flags_x86_fma3 cpu_flags_x86_sse4_2 +openmp test
KEYWORDS=~amd64
LICENSE=Apache-2.0 MIT
RDEPEND=>=dev-util/nvidia-cuda-toolkit-13.2.2:= openmp? ( llvm-runtimes/openmp:= ) x11-drivers/nvidia-drivers
RESTRICT=!test? ( test )
SLOT=0
SRC_URI=https://github.com/kvmem/kvmem-llama.cpp/archive/refs/tags/v0.17.0.tar.gz -> kvmem-llama-cpp-0.17.0.tar.gz https://github.com/ggml-org/llama.cpp/archive/b81c99b479d4c24e5eeca10de99032ebd343ef8f.tar.gz -> llama.cpp-b81c99b479d4c24e5eeca10de99032ebd343ef8f.tar.gz
_eclasses_=cmake	6c98cdedd575b64a24f3e3a7a688c689	cuda	8b660e223a1695e3884ee4c7dc2c5059	flag-o-matic	fd3558f73f6503093adee69adf41020d	multiprocessing	45d0620f25c2f063ad6276ecd9ed9399	ninja-utils	d063ea1900a793f93eb095010d6d9156	toolchain-funcs	58a918e3a42d0156f010dcfd43725be5	xdg-utils	42869b3c8d86a70ef3cf75165a395e09
_md5_=e63226df800b48767802e78beebaaeff