Skip to content

arm64 support for windows #3007

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 14 commits into from
Sep 13, 2023
Merged

arm64 support for windows #3007

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
14 commits
Select commit Hold shift + click to select a range
d2a4c68
win arm fixes
es0m-msft Sep 3, 2023
416f987
Merge branch 'master' into last-working-old-file-format
es0m Sep 4, 2023
245f9e8
Merge pull request #1 from es0m/last-working-old-file-format
es0m Sep 4, 2023
8fb5cc9
PR comments; using portable initialization of uint32x2_t
es0m-msft Sep 4, 2023
1c1e74c
Merge branch 'last-working-old-file-format'
es0m-msft Sep 4, 2023
86ba3b0
added cross-platform compilation for msvc
es0m-msft Sep 5, 2023
61577f0
fixing check and order of generator platform check
es0m-msft Sep 5, 2023
98f6172
cmake : fix quoting issues
cebtenzzre Sep 5, 2023
b48994b
cmake : fix trailing whitespace
cebtenzzre Sep 5, 2023
47d5f35
cmake : fix whitespace
cebtenzzre Sep 7, 2023
cb86477
addressed PR comments
es0m-msft Sep 9, 2023
cfcdd52
switching to fast math for some more speedup, see https://learn.micro…
es0m-msft Sep 9, 2023
04b7bd2
removing fp:fast as per PR comment
es0m-msft Sep 11, 2023
e072111
fixed merge conflicts
es0m-msft Sep 12, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
17 changes: 14 additions & 3 deletions CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -461,6 +461,13 @@ endif()
# TODO: probably these flags need to be tweaked on some architectures
# feel free to update the Makefile for your architecture and send a pull request or issue
message(STATUS "CMAKE_SYSTEM_PROCESSOR: ${CMAKE_SYSTEM_PROCESSOR}")
if (MSVC)
string(TOLOWER "${CMAKE_GENERATOR_PLATFORM}" CMAKE_GENERATOR_PLATFORM_LWR)
message(STATUS "CMAKE_GENERATOR_PLATFORM: ${CMAKE_GENERATOR_PLATFORM}")
else ()
set(CMAKE_GENERATOR_PLATFORM_LWR "")
endif ()

if (NOT MSVC)
if (LLAMA_STATIC)
add_link_options(-static)
Expand All @@ -476,10 +483,14 @@ if (NOT MSVC)
endif()
endif()

if ((${CMAKE_SYSTEM_PROCESSOR} MATCHES "arm") OR (${CMAKE_SYSTEM_PROCESSOR} MATCHES "aarch64"))
if ((${CMAKE_SYSTEM_PROCESSOR} MATCHES "arm") OR (${CMAKE_SYSTEM_PROCESSOR} MATCHES "aarch64") OR ("${CMAKE_GENERATOR_PLATFORM_LWR}" MATCHES "arm64"))
message(STATUS "ARM detected")
if (MSVC)
# TODO: arm msvc?
add_compile_definitions(__ARM_NEON)
add_compile_definitions(__ARM_FEATURE_FMA)
add_compile_definitions(__ARM_FEATURE_DOTPROD)
# add_compile_definitions(__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) # MSVC doesn't support vdupq_n_f16, vld1q_f16, vst1q_f16
add_compile_definitions(__aarch64__) # MSVC defines _M_ARM64 instead
else()
if (${CMAKE_SYSTEM_PROCESSOR} MATCHES "armv6")
# Raspberry Pi 1, Zero
Expand All @@ -494,7 +505,7 @@ if ((${CMAKE_SYSTEM_PROCESSOR} MATCHES "arm") OR (${CMAKE_SYSTEM_PROCESSOR} MATC
add_compile_options(-mfp16-format=ieee -mno-unaligned-access)
endif()
endif()
elseif (${CMAKE_SYSTEM_PROCESSOR} MATCHES "^(x86_64|i686|AMD64)$")
elseif (${CMAKE_SYSTEM_PROCESSOR} MATCHES "^(x86_64|i686|AMD64)$" OR "${CMAKE_GENERATOR_PLATFORM_LWR}" MATCHES "^(x86_64|i686|amd64|x64)$" )
message(STATUS "x86 detected")
if (MSVC)
if (LLAMA_AVX512)
Expand Down
2 changes: 1 addition & 1 deletion ggml.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -283,7 +283,7 @@ typedef double ggml_float;
// 16-bit float
// on Arm, we use __fp16
// on x86, we use uint16_t
#ifdef __ARM_NEON
#if defined(__ARM_NEON) && !defined(_MSC_VER)

// if YCM cannot find <arm_neon.h>, make a symbolic link to it, for example:
//
Expand Down
2 changes: 1 addition & 1 deletion ggml.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -270,7 +270,7 @@ extern "C" {

#if defined(__ARM_NEON) && defined(__CUDACC__)
typedef half ggml_fp16_t;
#elif defined(__ARM_NEON)
#elif defined(__ARM_NEON) && !defined(_MSC_VER)
typedef __fp16 ggml_fp16_t;
#else
typedef uint16_t ggml_fp16_t;
Expand Down
5 changes: 4 additions & 1 deletion k_quants.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2609,7 +2609,10 @@ void ggml_vec_dot_q4_K_q8_K(const int n, float * restrict s, const void * restri

memcpy(utmp, x[i].scales, 12);

const uint32x2_t mins8 = {utmp[1] & kmask1, ((utmp[2] >> 4) & kmask2) | (((utmp[1] >> 6) & kmask3) << 4)};
uint32x2_t mins8 = { 0 };
mins8 = vset_lane_u32(utmp[1] & kmask1, mins8, 0);
mins8 = vset_lane_u32(((utmp[2] >> 4) & kmask2) | (((utmp[1] >> 6) & kmask3) << 4), mins8, 1);

utmp[1] = (utmp[2] & kmask2) | (((utmp[0] >> 6) & kmask3) << 4);
utmp[0] &= kmask1;

Expand Down