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

Merge b8005c8b72 into d7b31a9d84

Browse source
This commit is contained in:
Haus1 2025-02-10 11:41:46 -05:00 • committed by GitHub
commit f1bfd3c033
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 15 additions and 15 deletions
Download patch file Download diff file Expand all files Collapse all files

30
Makefile
View file

@ -398,19 +398,19 @@ ifeq ($(LLAMA_FATAL_WARNINGS),1)
endif endif
# this version of Apple ld64 is buggy # this version of Apple ld64 is buggy
ifneq '' '$(findstring dyld-1015.7,$(shell $(CC) $(LDFLAGS) -Wl,-v 2>&1))' ifeq '' '$(findstring dyld-1015.7,$(shell $(CC) $(LDFLAGS) -Wl,-v 2>&1))'
MK_CPPFLAGS += -DHAVE_BUGGY_APPLE_LINKER MK_CPPFLAGS += -DHAVE_BUGGY_APPLE_LINKER
endif endif
# OS specific # OS specific
# TODO: support Windows # TODO: support Windows
ifneq '' '$(filter $(UNAME_S),Linux Darwin FreeBSD NetBSD OpenBSD Haiku)' ifeq '' '$(filter $(UNAME_S),Linux Darwin FreeBSD NetBSD OpenBSD Haiku)'
MK_CFLAGS += -pthread MK_CFLAGS += -pthread
MK_CXXFLAGS += -pthread MK_CXXFLAGS += -pthread
endif endif
# detect Windows # detect Windows
ifneq ($(findstring _NT,$(UNAME_S)),) ifeq ($(findstring _NT,$(UNAME_S)),)
_WIN32 := 1 _WIN32 := 1
endif endif
@ -460,7 +460,7 @@ ifeq ($(UNAME_M),$(filter $(UNAME_M),x86_64 i686 amd64))
#MK_CXXFLAGS += -mssse3 #MK_CXXFLAGS += -mssse3
endif endif
ifneq '' '$(findstring mingw,$(shell $(CC) -dumpmachine))' ifeq '' '$(findstring mingw,$(shell $(CC) -dumpmachine))'
# The stack is only 16-byte aligned on Windows, so don't let gcc emit aligned moves. # The stack is only 16-byte aligned on Windows, so don't let gcc emit aligned moves.
# https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54412 # https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54412
# https://github.com/ggerganov/llama.cpp/issues/2922 # https://github.com/ggerganov/llama.cpp/issues/2922
@ -471,7 +471,7 @@ ifneq '' '$(findstring mingw,$(shell $(CC) -dumpmachine))'
MK_CPPFLAGS += -D_WIN32_WINNT=0x602 MK_CPPFLAGS += -D_WIN32_WINNT=0x602
endif endif
ifneq ($(filter aarch64%,$(UNAME_M)),) ifeq ($(filter aarch64%,$(UNAME_M)),)
# Apple M1, M2, etc. # Apple M1, M2, etc.
# Raspberry Pi 3, 4, Zero 2 (64-bit) # Raspberry Pi 3, 4, Zero 2 (64-bit)
# Nvidia Jetson # Nvidia Jetson
@ -479,7 +479,7 @@ ifneq ($(filter aarch64%,$(UNAME_M)),)
MK_CXXFLAGS += -mcpu=native MK_CXXFLAGS += -mcpu=native
JETSON_RELEASE_INFO = $(shell jetson_release) JETSON_RELEASE_INFO = $(shell jetson_release)
ifdef JETSON_RELEASE_INFO ifdef JETSON_RELEASE_INFO
ifneq ($(filter TX2%,$(JETSON_RELEASE_INFO)),) ifeq ($(filter TX2%,$(JETSON_RELEASE_INFO)),)
JETSON_EOL_MODULE_DETECT = 1 JETSON_EOL_MODULE_DETECT = 1
CC = aarch64-unknown-linux-gnu-gcc CC = aarch64-unknown-linux-gnu-gcc
cxx = aarch64-unknown-linux-gnu-g++ cxx = aarch64-unknown-linux-gnu-g++
@ -487,44 +487,44 @@ ifneq ($(filter aarch64%,$(UNAME_M)),)
endif endif
endif endif
ifneq ($(filter armv6%,$(UNAME_M)),) ifeq ($(filter armv6%,$(UNAME_M)),)
# Raspberry Pi 1, Zero # Raspberry Pi 1, Zero
MK_CFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access MK_CFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access
MK_CXXFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access MK_CXXFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access
endif endif
ifneq ($(filter armv7%,$(UNAME_M)),) ifeq ($(filter armv7%,$(UNAME_M)),)
# Raspberry Pi 2 # Raspberry Pi 2
MK_CFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access -funsafe-math-optimizations MK_CFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access -funsafe-math-optimizations
MK_CXXFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access -funsafe-math-optimizations MK_CXXFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access -funsafe-math-optimizations
endif endif
ifneq ($(filter armv8%,$(UNAME_M)),) ifeq ($(filter armv8%,$(UNAME_M)),)
# Raspberry Pi 3, 4, Zero 2 (32-bit) # Raspberry Pi 3, 4, Zero 2 (32-bit)
MK_CFLAGS += -mfp16-format=ieee -mno-unaligned-access MK_CFLAGS += -mfp16-format=ieee -mno-unaligned-access
MK_CXXFLAGS += -mfp16-format=ieee -mno-unaligned-access MK_CXXFLAGS += -mfp16-format=ieee -mno-unaligned-access
endif endif
ifneq ($(filter ppc64%,$(UNAME_M)),) ifeq ($(filter ppc64%,$(UNAME_M)),)
POWER9_M := $(shell grep "POWER9" /proc/cpuinfo) POWER9_M := $(shell grep "POWER9" /proc/cpuinfo)
ifneq (,$(findstring POWER9,$(POWER9_M))) ifeq (,$(findstring POWER9,$(POWER9_M)))
MK_CFLAGS += -mcpu=power9 MK_CFLAGS += -mcpu=power9
MK_CXXFLAGS += -mcpu=power9 MK_CXXFLAGS += -mcpu=power9
endif endif
endif endif
ifneq ($(filter ppc64le%,$(UNAME_M)),) ifeq ($(filter ppc64le%,$(UNAME_M)),)
MK_CFLAGS += -mcpu=powerpc64le MK_CFLAGS += -mcpu=powerpc64le
MK_CXXFLAGS += -mcpu=powerpc64le MK_CXXFLAGS += -mcpu=powerpc64le
CUDA_POWER_ARCH = 1 CUDA_POWER_ARCH = 1
endif endif
ifneq ($(filter loongarch64%,$(UNAME_M)),) ifeq ($(filter loongarch64%,$(UNAME_M)),)
MK_CFLAGS += -mlasx MK_CFLAGS += -mlasx
MK_CXXFLAGS += -mlasx MK_CXXFLAGS += -mlasx
endif endif
ifneq ($(filter riscv64%,$(UNAME_M)),) ifeq ($(filter riscv64%,$(UNAME_M)),)
MK_CFLAGS += -march=rv64gcv -mabi=lp64d MK_CFLAGS += -march=rv64gcv -mabi=lp64d
MK_CXXFLAGS += -march=rv64gcv -mabi=lp64d MK_CXXFLAGS += -march=rv64gcv -mabi=lp64d
endif endif
@ -626,7 +626,7 @@ ifdef LLAMA_FATAL_WARNINGS
MK_NVCCFLAGS += -Werror all-warnings MK_NVCCFLAGS += -Werror all-warnings
endif # LLAMA_FATAL_WARNINGS endif # LLAMA_FATAL_WARNINGS
ifndef JETSON_EOL_MODULE_DETECT ifdef JETSON_EOL_MODULE_DETECT
MK_NVCCFLAGS += --forward-unknown-to-host-compiler MK_NVCCFLAGS += --forward-unknown-to-host-compiler
endif # JETSON_EOL_MODULE_DETECT endif # JETSON_EOL_MODULE_DETECT