llama.cpp
Numa
#1556
Merged
Overview
Commits
20
Changes
View On GitHub
Commits
  • detect NUMA systems and pin work threads to nodes (linux)
    zrm committed 3 years ago
  • disable mmap prefetch/readahead for NUMA systems
    zrm committed 3 years ago
  • avoid sending finalize op to thread pool if it does nothing
    zrm committed 3 years ago
  • silence robot
    zrm committed 3 years ago
  • fix args
    zrm committed 3 years ago
  • make --numa a param
    zrm committed 3 years ago
  • recommendation that n_nodes evenly divide n_threads did not warrant such aggressive enforcement
    zrm committed 3 years ago
  • lower synchronization overhead
    zrm committed 3 years ago
  • statically allocate
    zrm committed 3 years ago
  • move numa state to g_state
    zrm committed 3 years ago
  • add description for --numa
    zrm committed 3 years ago
  • Merge branch 'master' into HEAD
    ggerganov committed 3 years ago
  • ggml : minor style changes
    ggerganov committed 3 years ago
  • ggml : minor style + try fix sanitizer build
    ggerganov committed 3 years ago
  • Merge branch 'master' into HEAD
    ggerganov committed 3 years ago
  • llama : allow to initialize backend with NUMA support
    ggerganov committed 3 years ago
  • llama : avoid ggml include in llama-util.h
    ggerganov committed 3 years ago
  • ggml : style / formatting
    ggerganov committed 3 years ago
  • ggml : fix handling of ops with n_threads > n_tasks > 1
    ggerganov committed 3 years ago
  • server : utilize numa parameter
    ggerganov committed 3 years ago
FAQ Terms Privacy Refunds Impressum
Loading