llama.cpp
fix: use `vm_allocate` to allocate CPU backend buffer on macOS
#9875
Merged
Overview
Commits
15
Changes
View On GitHub
Commits
  • fix: use `vm_allocate` to allocate CPU backend buffer on macOS
    giladgd committed 1 year ago
  • fix: switch to `posix_memalign` to keep existing `free()` usages work
    giladgd committed 1 year ago
  • feat: move `GGML_ALIGNED_MALLOC` to `ggml-backend-impl.h`, add support for `vm_allocate` on macOS
    giladgd committed 1 year ago
  • style: formatting
    giladgd committed 1 year ago
  • fix: move const outside of `#ifndef`
    giladgd committed 1 year ago
  • style: formatting
    giladgd committed 1 year ago
  • fix: unused var
    giladgd committed 1 year ago
  • fix: transform `GGML_ALIGNED_MALLOC` and `GGML_ALIGNED_FREE` into functions and add them to `ggml-impl.h`
    giladgd committed 1 year ago
  • fix: unused var
    giladgd committed 1 year ago
  • fix: page align to `GGUF_DEFAULT_ALIGNMENT`
    giladgd committed 1 year ago
  • fix: page align to `TENSOR_ALIGNMENT`
    giladgd committed 1 year ago
  • fix: convert `TENSOR_ALIGNMENT` to a macro
    giladgd committed 1 year ago
  • fix: increase page size to `32` on iOS
    giladgd committed 1 year ago
  • fix: iOS page size
    giladgd committed 1 year ago
  • fix: `hbw_posix_memalign` alignment
    giladgd committed 1 year ago
FAQ Terms Privacy Refunds Impressum
Loading