sci-misc/llama-cpp
Port of Facebook's LLaMA model in C/C++
Screenshots
-
llama-cpp-9999curl openblas +openmp blis rocm cuda opencl openssl vulkan flexiblas hip-graphs +vmm examples +amdgpu_targets_gfx908 +amdgpu_targets_gfx90a +amdgpu_targets_gfx942 +amdgpu_targets_gfx1030 +amdgpu_targets_gfx1100 +amdgpu_targets_gfx1101 +amdgpu_targets_gfx1200 +amdgpu_targets_gfx1201 amdgpu_targets_gfx803 amdgpu_targets_gfx900 amdgpu_targets_gfx906 amdgpu_targets_gfx940 amdgpu_targets_gfx941 amdgpu_targets_gfx1010 amdgpu_targets_gfx1011 amdgpu_targets_gfx1012 amdgpu_targets_gfx1031 amdgpu_targets_gfx1102 amdgpu_targets_gfx1103 amdgpu_targets_gfx1150 amdgpu_targets_gfx1151
View 
Download 
Browse License: MIT Overlay: inode64-overlay -
llama-cpp-9999blis cpu_flags_x86_avx cpu_flags_x86_avx2 cpu_flags_x86_f16c cpu_flags_x86_fma3 cpu_flags_x86_sse4_2 cuda flexiblas openblas opencl +openmp openssl rocm rpc vulkan wmma +amdgpu_targets_gfx908 +amdgpu_targets_gfx90a +amdgpu_targets_gfx942 +amdgpu_targets_gfx1030 +amdgpu_targets_gfx1100 +amdgpu_targets_gfx1101 +amdgpu_targets_gfx1200 +amdgpu_targets_gfx1201 amdgpu_targets_gfx803 amdgpu_targets_gfx900 amdgpu_targets_gfx906 amdgpu_targets_gfx940 amdgpu_targets_gfx941 amdgpu_targets_gfx1010 amdgpu_targets_gfx1011 amdgpu_targets_gfx1012 amdgpu_targets_gfx1031 amdgpu_targets_gfx1102 amdgpu_targets_gfx1103 amdgpu_targets_gfx1150 amdgpu_targets_gfx1151
View Download Browse License: MIT Overlay: vklimovs -
llama-cpp-9999~amd64curl openblas +openmp blis hip cuda opencl vulkan xdna +amdgpu_targets_gfx908 +amdgpu_targets_gfx90a +amdgpu_targets_gfx942 +amdgpu_targets_gfx1030 +amdgpu_targets_gfx1100 +amdgpu_targets_gfx1101 +amdgpu_targets_gfx1200 +amdgpu_targets_gfx1201 amdgpu_targets_gfx803 amdgpu_targets_gfx900 amdgpu_targets_gfx906 amdgpu_targets_gfx940 amdgpu_targets_gfx941 amdgpu_targets_gfx1010 amdgpu_targets_gfx1011 amdgpu_targets_gfx1012 amdgpu_targets_gfx1031 amdgpu_targets_gfx1102 amdgpu_targets_gfx1103 amdgpu_targets_gfx1150 amdgpu_targets_gfx1151
View Download Browse License: MIT Overlay: FireBurn -
llama-cpp-9999curl openblas +openmp blis rocm cuda opencl openssl vulkan flexiblas wmma examples +amdgpu_targets_gfx908 +amdgpu_targets_gfx90a +amdgpu_targets_gfx942 +amdgpu_targets_gfx1030 +amdgpu_targets_gfx1100 +amdgpu_targets_gfx1101 +amdgpu_targets_gfx1200 +amdgpu_targets_gfx1201 amdgpu_targets_gfx803 amdgpu_targets_gfx900 amdgpu_targets_gfx906 amdgpu_targets_gfx940 amdgpu_targets_gfx941 amdgpu_targets_gfx1010 amdgpu_targets_gfx1011 amdgpu_targets_gfx1012 amdgpu_targets_gfx1031 amdgpu_targets_gfx1102 amdgpu_targets_gfx1103 amdgpu_targets_gfx1150 amdgpu_targets_gfx1151
View Download Browse License: MIT Overlay: booboo -
llama-cpp-9999curl openblas +openmp blis rocm cuda opencl openssl vulkan flexiblas wmma examples +amdgpu_targets_gfx908 +amdgpu_targets_gfx90a +amdgpu_targets_gfx942 +amdgpu_targets_gfx1030 +amdgpu_targets_gfx1100 +amdgpu_targets_gfx1101 +amdgpu_targets_gfx1200 +amdgpu_targets_gfx1201 amdgpu_targets_gfx803 amdgpu_targets_gfx900 amdgpu_targets_gfx906 amdgpu_targets_gfx940 amdgpu_targets_gfx941 amdgpu_targets_gfx1010 amdgpu_targets_gfx1011 amdgpu_targets_gfx1012 amdgpu_targets_gfx1031 amdgpu_targets_gfx1102 amdgpu_targets_gfx1103 amdgpu_targets_gfx1150 amdgpu_targets_gfx1151
View Download Browse License: MIT Overlay: guru -
llama-cpp-9999openblas +openmp blis rocm cuda opencl +openssl vulkan flexiblas examples +webui sycl cpu_flags_x86_avx cpu_flags_x86_avx2 cpu_flags_x86_avx512f cpu_flags_x86_avx512vbmi cpu_flags_x86_bmi2 cpu_flags_x86_f16c cpu_flags_x86_fma3 cpu_flags_x86_sse4_2 +amdgpu_targets_gfx908 +amdgpu_targets_gfx90a +amdgpu_targets_gfx942 +amdgpu_targets_gfx950 +amdgpu_targets_gfx1030 +amdgpu_targets_gfx1100 +amdgpu_targets_gfx1101 +amdgpu_targets_gfx1200 +amdgpu_targets_gfx1201 amdgpu_targets_gfx803 amdgpu_targets_gfx900 amdgpu_targets_gfx906 amdgpu_targets_gfx940 amdgpu_targets_gfx941 amdgpu_targets_gfx1010 amdgpu_targets_gfx1011 amdgpu_targets_gfx1012 amdgpu_targets_gfx1031 amdgpu_targets_gfx1102 amdgpu_targets_gfx1103 amdgpu_targets_gfx1150 amdgpu_targets_gfx1151
View Download Browse License: MIT Overlay: stuff -
llama-cpp-0_pre10595~amd64 ~arm64openblas +openmp blis rocm cuda opencl +openssl vulkan flexiblas examples +webui sycl cpu_flags_x86_avx cpu_flags_x86_avx2 cpu_flags_x86_avx512f cpu_flags_x86_avx512vbmi cpu_flags_x86_bmi2 cpu_flags_x86_f16c cpu_flags_x86_fma3 cpu_flags_x86_sse4_2 +amdgpu_targets_gfx908 +amdgpu_targets_gfx90a +amdgpu_targets_gfx942 +amdgpu_targets_gfx950 +amdgpu_targets_gfx1030 +amdgpu_targets_gfx1100 +amdgpu_targets_gfx1101 +amdgpu_targets_gfx1200 +amdgpu_targets_gfx1201 amdgpu_targets_gfx803 amdgpu_targets_gfx900 amdgpu_targets_gfx906 amdgpu_targets_gfx940 amdgpu_targets_gfx941 amdgpu_targets_gfx1010 amdgpu_targets_gfx1011 amdgpu_targets_gfx1012 amdgpu_targets_gfx1031 amdgpu_targets_gfx1102 amdgpu_targets_gfx1103 amdgpu_targets_gfx1150 amdgpu_targets_gfx1151
View Download Browse License: MIT Overlay: stuff -
llama-cpp-0_pre10582~amd64 ~arm64openblas +openmp blis rocm cuda opencl +openssl vulkan flexiblas examples +webui sycl cpu_flags_x86_avx cpu_flags_x86_avx2 cpu_flags_x86_avx512f cpu_flags_x86_avx512vbmi cpu_flags_x86_bmi2 cpu_flags_x86_f16c cpu_flags_x86_fma3 cpu_flags_x86_sse4_2 +amdgpu_targets_gfx908 +amdgpu_targets_gfx90a +amdgpu_targets_gfx942 +amdgpu_targets_gfx950 +amdgpu_targets_gfx1030 +amdgpu_targets_gfx1100 +amdgpu_targets_gfx1101 +amdgpu_targets_gfx1200 +amdgpu_targets_gfx1201 amdgpu_targets_gfx803 amdgpu_targets_gfx900 amdgpu_targets_gfx906 amdgpu_targets_gfx940 amdgpu_targets_gfx941 amdgpu_targets_gfx1010 amdgpu_targets_gfx1011 amdgpu_targets_gfx1012 amdgpu_targets_gfx1031 amdgpu_targets_gfx1102 amdgpu_targets_gfx1103 amdgpu_targets_gfx1150 amdgpu_targets_gfx1151
View Download Browse License: MIT Overlay: stuff -
llama-cpp-0_pre10453~amd64openblas +openmp blis rocm cuda opencl openssl vulkan flexiblas examples +amdgpu_targets_gfx908 +amdgpu_targets_gfx90a +amdgpu_targets_gfx942 +amdgpu_targets_gfx1030 +amdgpu_targets_gfx1100 +amdgpu_targets_gfx1101 +amdgpu_targets_gfx1200 +amdgpu_targets_gfx1201 amdgpu_targets_gfx803 amdgpu_targets_gfx900 amdgpu_targets_gfx906 amdgpu_targets_gfx940 amdgpu_targets_gfx941 amdgpu_targets_gfx1010 amdgpu_targets_gfx1011 amdgpu_targets_gfx1012 amdgpu_targets_gfx1031 amdgpu_targets_gfx1102 amdgpu_targets_gfx1103 amdgpu_targets_gfx1150 amdgpu_targets_gfx1151
View Download Browse License: MIT Overlay: hossie -
llama-cpp-0_pre10235~amd64blis cpu_flags_x86_avx cpu_flags_x86_avx2 cpu_flags_x86_f16c cpu_flags_x86_fma3 cpu_flags_x86_sse4_2 cuda flexiblas openblas opencl +openmp openssl rocm rpc vulkan wmma +amdgpu_targets_gfx908 +amdgpu_targets_gfx90a +amdgpu_targets_gfx942 +amdgpu_targets_gfx1030 +amdgpu_targets_gfx1100 +amdgpu_targets_gfx1101 +amdgpu_targets_gfx1200 +amdgpu_targets_gfx1201 amdgpu_targets_gfx803 amdgpu_targets_gfx900 amdgpu_targets_gfx906 amdgpu_targets_gfx940 amdgpu_targets_gfx941 amdgpu_targets_gfx1010 amdgpu_targets_gfx1011 amdgpu_targets_gfx1012 amdgpu_targets_gfx1031 amdgpu_targets_gfx1102 amdgpu_targets_gfx1103 amdgpu_targets_gfx1150 amdgpu_targets_gfx1151
View Download Browse License: MIT Overlay: vklimovs -
llama-cpp-0_pre9941~amd64openblas +openmp blis rocm cuda opencl openssl vulkan flexiblas wmma examples +amdgpu_targets_gfx908 +amdgpu_targets_gfx90a +amdgpu_targets_gfx942 +amdgpu_targets_gfx1030 +amdgpu_targets_gfx1100 +amdgpu_targets_gfx1101 +amdgpu_targets_gfx1200 +amdgpu_targets_gfx1201 amdgpu_targets_gfx803 amdgpu_targets_gfx900 amdgpu_targets_gfx906 amdgpu_targets_gfx940 amdgpu_targets_gfx941 amdgpu_targets_gfx1010 amdgpu_targets_gfx1011 amdgpu_targets_gfx1012 amdgpu_targets_gfx1031 amdgpu_targets_gfx1102 amdgpu_targets_gfx1103 amdgpu_targets_gfx1150 amdgpu_targets_gfx1151
View Download Browse License: MIT Overlay: hossie -
llama-cpp-0_pre9888~amd64curl openblas +openmp blis rocm cuda opencl openssl vulkan flexiblas wmma examples +amdgpu_targets_gfx908 +amdgpu_targets_gfx90a +amdgpu_targets_gfx942 +amdgpu_targets_gfx950 +amdgpu_targets_gfx1030 +amdgpu_targets_gfx1100 +amdgpu_targets_gfx1101 +amdgpu_targets_gfx1200 +amdgpu_targets_gfx1201 amdgpu_targets_gfx803 amdgpu_targets_gfx900 amdgpu_targets_gfx906 amdgpu_targets_gfx940 amdgpu_targets_gfx941 amdgpu_targets_gfx1010 amdgpu_targets_gfx1011 amdgpu_targets_gfx1012 amdgpu_targets_gfx1031 amdgpu_targets_gfx1102 amdgpu_targets_gfx1103 amdgpu_targets_gfx1150 amdgpu_targets_gfx1151
View Download Browse License: MIT Overlay: guru -
llama-cpp-0_pre9747~amd64openblas +openmp blis rocm cuda opencl openssl vulkan flexiblas wmma examples +amdgpu_targets_gfx908 +amdgpu_targets_gfx90a +amdgpu_targets_gfx942 +amdgpu_targets_gfx1030 +amdgpu_targets_gfx1100 +amdgpu_targets_gfx1101 +amdgpu_targets_gfx1200 +amdgpu_targets_gfx1201 amdgpu_targets_gfx803 amdgpu_targets_gfx900 amdgpu_targets_gfx906 amdgpu_targets_gfx940 amdgpu_targets_gfx941 amdgpu_targets_gfx1010 amdgpu_targets_gfx1011 amdgpu_targets_gfx1012 amdgpu_targets_gfx1031 amdgpu_targets_gfx1102 amdgpu_targets_gfx1103 amdgpu_targets_gfx1150 amdgpu_targets_gfx1151
View Download Browse License: MIT Overlay: hossie -
llama-cpp-0_pre9700~amd64curl openblas +openmp blis rocm cuda opencl openssl vulkan flexiblas wmma examples +amdgpu_targets_gfx908 +amdgpu_targets_gfx90a +amdgpu_targets_gfx942 +amdgpu_targets_gfx1030 +amdgpu_targets_gfx1100 +amdgpu_targets_gfx1101 +amdgpu_targets_gfx1200 +amdgpu_targets_gfx1201 amdgpu_targets_gfx803 amdgpu_targets_gfx900 amdgpu_targets_gfx906 amdgpu_targets_gfx940 amdgpu_targets_gfx941 amdgpu_targets_gfx1010 amdgpu_targets_gfx1011 amdgpu_targets_gfx1012 amdgpu_targets_gfx1031 amdgpu_targets_gfx1102 amdgpu_targets_gfx1103 amdgpu_targets_gfx1150 amdgpu_targets_gfx1151
View Download Browse License: MIT Overlay: booboo -
llama-cpp-0_pre8838~amd64curl openblas +openmp blis rocm cuda opencl openssl vulkan flexiblas wmma examples +amdgpu_targets_gfx908 +amdgpu_targets_gfx90a +amdgpu_targets_gfx942 +amdgpu_targets_gfx1030 +amdgpu_targets_gfx1100 +amdgpu_targets_gfx1101 +amdgpu_targets_gfx1200 +amdgpu_targets_gfx1201 amdgpu_targets_gfx803 amdgpu_targets_gfx900 amdgpu_targets_gfx906 amdgpu_targets_gfx940 amdgpu_targets_gfx941 amdgpu_targets_gfx1010 amdgpu_targets_gfx1011 amdgpu_targets_gfx1012 amdgpu_targets_gfx1031 amdgpu_targets_gfx1102 amdgpu_targets_gfx1103 amdgpu_targets_gfx1150 amdgpu_targets_gfx1151
View Download Browse License: MIT Overlay: guru -
llama-cpp-0_pre8628~amd64curl openblas +openmp blis rocm cuda opencl vulkan flexiblas wmma examples +amdgpu_targets_gfx908 +amdgpu_targets_gfx90a +amdgpu_targets_gfx942 +amdgpu_targets_gfx1030 +amdgpu_targets_gfx1100 +amdgpu_targets_gfx1101 +amdgpu_targets_gfx1200 +amdgpu_targets_gfx1201 amdgpu_targets_gfx803 amdgpu_targets_gfx900 amdgpu_targets_gfx906 amdgpu_targets_gfx940 amdgpu_targets_gfx941 amdgpu_targets_gfx1010 amdgpu_targets_gfx1011 amdgpu_targets_gfx1012 amdgpu_targets_gfx1031 amdgpu_targets_gfx1102 amdgpu_targets_gfx1103 amdgpu_targets_gfx1150 amdgpu_targets_gfx1151
View Download Browse License: MIT Overlay: guru -
llama-cpp-0_pre8198~amd64curl openblas +openmp blis rocm cuda opencl vulkan flexiblas wmma examples +amdgpu_targets_gfx908 +amdgpu_targets_gfx90a +amdgpu_targets_gfx942 +amdgpu_targets_gfx1030 +amdgpu_targets_gfx1100 +amdgpu_targets_gfx1101 +amdgpu_targets_gfx1200 +amdgpu_targets_gfx1201 amdgpu_targets_gfx803 amdgpu_targets_gfx900 amdgpu_targets_gfx906 amdgpu_targets_gfx940 amdgpu_targets_gfx941 amdgpu_targets_gfx1010 amdgpu_targets_gfx1011 amdgpu_targets_gfx1012 amdgpu_targets_gfx1031 amdgpu_targets_gfx1102 amdgpu_targets_gfx1103 amdgpu_targets_gfx1150 amdgpu_targets_gfx1151
View Download Browse License: MIT Overlay: guru -
llama-cpp-0_pre7924~amd64curl openblas +openmp blis rocm cuda opencl vulkan flexiblas wmma examples +amdgpu_targets_gfx908 +amdgpu_targets_gfx90a +amdgpu_targets_gfx942 +amdgpu_targets_gfx1030 +amdgpu_targets_gfx1100 +amdgpu_targets_gfx1101 +amdgpu_targets_gfx1200 +amdgpu_targets_gfx1201 amdgpu_targets_gfx803 amdgpu_targets_gfx900 amdgpu_targets_gfx906 amdgpu_targets_gfx940 amdgpu_targets_gfx941 amdgpu_targets_gfx1010 amdgpu_targets_gfx1011 amdgpu_targets_gfx1012 amdgpu_targets_gfx1031 amdgpu_targets_gfx1102 amdgpu_targets_gfx1103 amdgpu_targets_gfx1150 amdgpu_targets_gfx1151
View Download Browse License: MIT Overlay: guru -
llama-cpp-0_pre7611~amd64curl openblas +openmp blis hip cuda opencl vulkan flexiblas +amdgpu_targets_gfx908 +amdgpu_targets_gfx90a +amdgpu_targets_gfx942 +amdgpu_targets_gfx1030 +amdgpu_targets_gfx1100 +amdgpu_targets_gfx1101 +amdgpu_targets_gfx1200 +amdgpu_targets_gfx1201 amdgpu_targets_gfx803 amdgpu_targets_gfx900 amdgpu_targets_gfx906 amdgpu_targets_gfx940 amdgpu_targets_gfx941 amdgpu_targets_gfx1010 amdgpu_targets_gfx1011 amdgpu_targets_gfx1012 amdgpu_targets_gfx1031 amdgpu_targets_gfx1102 amdgpu_targets_gfx1103 amdgpu_targets_gfx1150 amdgpu_targets_gfx1151
View Download Browse License: MIT Overlay: guru -
llama-cpp-0_pre7276~amd64curl openblas +openmp blis hip cuda opencl vulkan flexiblas +amdgpu_targets_gfx908 +amdgpu_targets_gfx90a +amdgpu_targets_gfx942 +amdgpu_targets_gfx1030 +amdgpu_targets_gfx1100 +amdgpu_targets_gfx1101 +amdgpu_targets_gfx1200 +amdgpu_targets_gfx1201 amdgpu_targets_gfx803 amdgpu_targets_gfx900 amdgpu_targets_gfx906 amdgpu_targets_gfx940 amdgpu_targets_gfx941 amdgpu_targets_gfx1010 amdgpu_targets_gfx1011 amdgpu_targets_gfx1012 amdgpu_targets_gfx1031 amdgpu_targets_gfx1102 amdgpu_targets_gfx1103 amdgpu_targets_gfx1150 amdgpu_targets_gfx1151
View Download Browse License: MIT Overlay: guru -
llama-cpp-0_pre6980~amd64curl openblas +openmp blis hip cuda opencl vulkan flexiblas +amdgpu_targets_gfx908 +amdgpu_targets_gfx90a +amdgpu_targets_gfx942 +amdgpu_targets_gfx1030 +amdgpu_targets_gfx1100 +amdgpu_targets_gfx1101 +amdgpu_targets_gfx1200 +amdgpu_targets_gfx1201 amdgpu_targets_gfx803 amdgpu_targets_gfx900 amdgpu_targets_gfx906 amdgpu_targets_gfx940 amdgpu_targets_gfx941 amdgpu_targets_gfx1010 amdgpu_targets_gfx1011 amdgpu_targets_gfx1012 amdgpu_targets_gfx1031 amdgpu_targets_gfx1102 amdgpu_targets_gfx1103 amdgpu_targets_gfx1150 amdgpu_targets_gfx1151
View Download Browse License: MIT Overlay: guru -
llama-cpp-0.3.0~amd64curl openblas +openmp blis rocm cuda opencl openssl vulkan flexiblas wmma examples +amdgpu_targets_gfx908 +amdgpu_targets_gfx90a +amdgpu_targets_gfx942 +amdgpu_targets_gfx1030 +amdgpu_targets_gfx1100 +amdgpu_targets_gfx1101 +amdgpu_targets_gfx1200 +amdgpu_targets_gfx1201 amdgpu_targets_gfx803 amdgpu_targets_gfx900 amdgpu_targets_gfx906 amdgpu_targets_gfx940 amdgpu_targets_gfx941 amdgpu_targets_gfx1010 amdgpu_targets_gfx1011 amdgpu_targets_gfx1012 amdgpu_targets_gfx1031 amdgpu_targets_gfx1102 amdgpu_targets_gfx1103 amdgpu_targets_gfx1150 amdgpu_targets_gfx1151
View Download Browse License: MIT Overlay: booboo -
llama-cpp-0.2.0~amd64 ~arm64openblas +openmp blis rocm cuda opencl +openssl vulkan flexiblas examples +webui sycl cpu_flags_x86_avx cpu_flags_x86_avx2 cpu_flags_x86_avx512f cpu_flags_x86_avx512vbmi cpu_flags_x86_bmi2 cpu_flags_x86_f16c cpu_flags_x86_fma3 cpu_flags_x86_sse4_2 +amdgpu_targets_gfx908 +amdgpu_targets_gfx90a +amdgpu_targets_gfx942 +amdgpu_targets_gfx950 +amdgpu_targets_gfx1030 +amdgpu_targets_gfx1100 +amdgpu_targets_gfx1101 +amdgpu_targets_gfx1200 +amdgpu_targets_gfx1201 amdgpu_targets_gfx803 amdgpu_targets_gfx900 amdgpu_targets_gfx906 amdgpu_targets_gfx940 amdgpu_targets_gfx941 amdgpu_targets_gfx1010 amdgpu_targets_gfx1011 amdgpu_targets_gfx1012 amdgpu_targets_gfx1031 amdgpu_targets_gfx1102 amdgpu_targets_gfx1103 amdgpu_targets_gfx1150 amdgpu_targets_gfx1151
View Download Browse License: MIT Overlay: stuff -
llama-cpp-0.1.2~amd64 ~arm64openblas +openmp blis rocm cuda opencl +openssl vulkan flexiblas examples +webui sycl cpu_flags_x86_avx cpu_flags_x86_avx2 cpu_flags_x86_avx512f cpu_flags_x86_avx512vbmi cpu_flags_x86_bmi2 cpu_flags_x86_f16c cpu_flags_x86_fma3 cpu_flags_x86_sse4_2 +amdgpu_targets_gfx908 +amdgpu_targets_gfx90a +amdgpu_targets_gfx942 +amdgpu_targets_gfx950 +amdgpu_targets_gfx1030 +amdgpu_targets_gfx1100 +amdgpu_targets_gfx1101 +amdgpu_targets_gfx1200 +amdgpu_targets_gfx1201 amdgpu_targets_gfx803 amdgpu_targets_gfx900 amdgpu_targets_gfx906 amdgpu_targets_gfx940 amdgpu_targets_gfx941 amdgpu_targets_gfx1010 amdgpu_targets_gfx1011 amdgpu_targets_gfx1012 amdgpu_targets_gfx1031 amdgpu_targets_gfx1102 amdgpu_targets_gfx1103 amdgpu_targets_gfx1150 amdgpu_targets_gfx1151
View Download Browse License: MIT Overlay: stuff
ChangeLog
commit cc9dc3169653076492c0ab3e6e4a02bbd8d02f9f
Author: INODE64 <ffelix@inode64.com>
Date: Thu Aug 27 21:55:17 2026 +0200
sci-misc/llama-cpp: replace obsolete wmma USE with hip-graphs and vmm
GGML_HIP_ROCWMMA_FATTN no longer exists upstream (RDNA3/4 flash attention
uses the native WMMA path, AMD_WMMA_AVAILABLE), so drop the wmma flag and
the sci-libs/rocWMMA dependency. The bf16 WMMA patch is still required.
Expose two HIP build options instead:
- hip-graphs (default off): GGML_HIP_GRAPHS. With ROCm 10.0.0 on gfx1201
the HIP graph path drops token generation from ~48 to ~30 tok/s and makes
timings unstable.
- vmm (default on): GGML_HIP_NO_VMM=OFF, +1-2% pp/tg on ROCm 10.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
commit af0e3b067746cf9fd4e66cd7718bbe122efa503e
Author: INODE64 <ffelix@inode64.com>
Date: Thu Aug 27 18:31:56 2026 +0200
Support por llvm 13 y rocm 10.0.0
commit c950be4517242e6708b858d3791023ab5b4f7d47
Author: Fco Javier Felix <ffelix@inode64.com>
Date: Tue Jun 30 15:00:53 2026 +0200
sci-misc/llama-cpp: Sync with guru
commit b413c4f2de76d0d6be06b524828085ba91e59dc4
Author: Fco Javier Felix <ffelix@inode64.com>
Date: Fri Jun 5 20:22:15 2026 +0200
sci-misc/llama-cpp: drop 0_pre8838
commit 79c917e29b032c936001db7e0d7f2861488420fc
Author: INODE64 <ffelix@inode64.com>
Date: Tue Apr 21 12:01:15 2026 +0200
*/*: Massive updates
Author: INODE64 <ffelix@inode64.com>
Date: Thu Aug 27 21:55:17 2026 +0200
sci-misc/llama-cpp: replace obsolete wmma USE with hip-graphs and vmm
GGML_HIP_ROCWMMA_FATTN no longer exists upstream (RDNA3/4 flash attention
uses the native WMMA path, AMD_WMMA_AVAILABLE), so drop the wmma flag and
the sci-libs/rocWMMA dependency. The bf16 WMMA patch is still required.
Expose two HIP build options instead:
- hip-graphs (default off): GGML_HIP_GRAPHS. With ROCm 10.0.0 on gfx1201
the HIP graph path drops token generation from ~48 to ~30 tok/s and makes
timings unstable.
- vmm (default on): GGML_HIP_NO_VMM=OFF, +1-2% pp/tg on ROCm 10.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
commit af0e3b067746cf9fd4e66cd7718bbe122efa503e
Author: INODE64 <ffelix@inode64.com>
Date: Thu Aug 27 18:31:56 2026 +0200
Support por llvm 13 y rocm 10.0.0
commit c950be4517242e6708b858d3791023ab5b4f7d47
Author: Fco Javier Felix <ffelix@inode64.com>
Date: Tue Jun 30 15:00:53 2026 +0200
sci-misc/llama-cpp: Sync with guru
commit b413c4f2de76d0d6be06b524828085ba91e59dc4
Author: Fco Javier Felix <ffelix@inode64.com>
Date: Fri Jun 5 20:22:15 2026 +0200
sci-misc/llama-cpp: drop 0_pre8838
commit 79c917e29b032c936001db7e0d7f2861488420fc
Author: INODE64 <ffelix@inode64.com>
Date: Tue Apr 21 12:01:15 2026 +0200
*/*: Massive updates