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

ggml : fix AVX build + update to new Q8_0 format

Browse source
This commit is contained in:
Georgi Gerganov 2023-04-22 11:08:12 +03:00
parent 955ef9a5d5
commit 872c365a91
3 changed files with 20 additions and 8 deletions
Download patch file Download diff file Expand all files Collapse all files

10
Makefile
View file

@ -74,13 +74,17 @@ endif
# feel free to update the Makefile for your architecture and send a pull request or issue
ifeq ($(UNAME_M),$(filter $(UNAME_M),x86_64 i686))
# Use all CPU extensions that are available:
CFLAGS += -march=native -mtune=native
CFLAGS += -march=native -mtune=native
CXXFLAGS += -march=native -mtune=native
# Usage AVX-only
#CFLAGS += -mfma -mf16c -mavx
#CXXFLAGS += -mfma -mf16c -mavx
endif
ifneq ($(filter ppc64%,$(UNAME_M)),)
POWER9_M := $(shell grep "POWER9" /proc/cpuinfo)
ifneq (,$(findstring POWER9,$(POWER9_M)))
CFLAGS += -mcpu=power9
CFLAGS += -mcpu=power9
CXXFLAGS += -mcpu=power9
endif
# Require c++23's std::byteswap for big-endian support.
@ -114,7 +118,7 @@ ifdef LLAMA_GPROF
CXXFLAGS += -pg
endif
ifneq ($(filter aarch64%,$(UNAME_M)),)
CFLAGS += -mcpu=native
CFLAGS += -mcpu=native
CXXFLAGS += -mcpu=native
endif
ifneq ($(filter armv6%,$(UNAME_M)),)