vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

ggml: aarch64: SVE kernels for q8_0_q8_0, q4_0_q8_0 vector dot (#7433)

Browse source 
* Add SVE support for q4_0_q8_0 q8_0_q8_0

* remove ifdef
This commit is contained in:
Masaya, Kato 2024-05-25 17:42:31 +09:00 • committed by GitHub
parent 9791f40258
commit faa0e6979a
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
7 changed files with 85 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

4
CMakeLists.txt
View file

@ -72,6 +72,7 @@ else()
set(INS_ENB ON)
endif()
option(LLAMA_SVE "llama: enable SVE" OFF)
option(LLAMA_AVX "llama: enable AVX" ${INS_ENB})
option(LLAMA_AVX2 "llama: enable AVX2" ${INS_ENB})
option(LLAMA_AVX512 "llama: enable AVX512" OFF)
@ -1040,6 +1041,9 @@ if (CMAKE_OSX_ARCHITECTURES STREQUAL "arm64" OR CMAKE_GENERATOR_PLATFORM_LWR STR
# Raspberry Pi 3, 4, Zero 2 (32-bit)
list(APPEND ARCH_FLAGS -mno-unaligned-access)
endif()
if (LLAMA_SVE)
list(APPEND ARCH_FLAGS -march=armv8.6-a+sve)
endif()
endif()
elseif (CMAKE_OSX_ARCHITECTURES STREQUAL "x86_64" OR CMAKE_GENERATOR_PLATFORM_LWR MATCHES "^(x86_64|i686|amd64|x64|win32)$" OR
(NOT CMAKE_OSX_ARCHITECTURES AND NOT CMAKE_GENERATOR_PLATFORM_LWR AND