Skip to content

Commit

Permalink
Merge pull request #126 from meneguzzi/sse2neon
Browse files Browse the repository at this point in the history 
Added sse2neon for arm compilation.
  • Loading branch information
@georgmartius
georgmartius committed Jan 7, 2024
2 parents 05829db + 406e61d commit cbe3d5c
Show file tree
Hide file tree
Showing 5 changed files with 9,219 additions and 0 deletions.
4 changes: 4 additions & 0 deletions .gitignore
View file
Original file line number Diff line number Diff line change
Expand Up @@ -9,3 +9,7 @@ CMakeFiles
Makefile
cmake_install.cmake
install_manifest.txt

# Don't include the build directory and compiled Mac dynamic libraries.
build
*.dylib
6 changes: 6 additions & 0 deletions CMakeLists.txt
View file
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,8 @@ set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} "${CMAKE_SOURCE_DIR}/CMakeModules/")

include (FindSSE)
include (GNUInstallDirs)
# include (ExternalProject) # There must be a nicer way to include this
# include (sse2neon)
find_package(OpenMP)

set(MAJOR_VERSION 1)
Expand All @@ -30,6 +32,10 @@ if(NOT WIN32)
add_definitions(-fPIC)
endif()

if(APPLE)
add_definitions(-march=armv8-a+fp+simd+crypto+crc)

This comment has been minimized.

Sign in to view

Copy link
@ddennedy

ddennedy Jan 8, 2024 •
edited
Loading

Contributor

@meneguzzi This broke compilation of universal (CMAKE_OSX_ARCHITECTURES='arm64;x86_64') on my M1 build machine for daily Shotcut builds. See https://github.com/mltframework/shotcut/actions/runs/7447186625

The error is error: unknown target CPU 'armv8-a+fp+simd+crypto+crc'
From cmake output -- The C compiler identification is AppleClang 14.0.0.14000029

This comment has been minimized.

Sign in to view

Copy link
@meneguzzi

meneguzzi Jan 8, 2024

Contributor

Weird, I tested it months ago. Next time my pull request should include tests as well. My apologies for this. Until next week I won't have bandwidth to fix this. If you have time to try and fix it, I'd appreciate it.

This comment has been minimized.

Sign in to view

Copy link
@meneguzzi

meneguzzi Jan 8, 2024

Contributor

I guess an easy fix is to add a check on the architecture, do you happen to know preprocessor constant applies only to Apple Silicon?

This comment has been minimized.

Sign in to view

Copy link
@ddennedy

ddennedy Jan 8, 2024

Contributor

I was thinking to change if(APPLE) to if (APPLE AND NOT CMAKE_OSX_ARCHITECTURES), but I have not yet tested it. Where does the strange (to me) architecture string you are using come from? Maybe also add a comment about that with a link. I will try it and report back.

This comment has been minimized.

Sign in to view

Copy link
@meneguzzi

meneguzzi Jan 8, 2024

Contributor

I found it in the CMake Documentation here:

https://cmake.org/cmake/help/latest/variable/APPLE.html

I did not find anything authoritative and "CMakey" to zero in on whether it's an ARM or Intel (or Motorola for real old timers).

This comment has been minimized.

Sign in to view

Copy link
@ddennedy

ddennedy Jan 8, 2024

Contributor

"APPLE" is not the confusing part, "-march=armv8-a+fp+simd+crypto+crc" is.

This comment has been minimized.

Sign in to view

Copy link
@ddennedy

ddennedy Jan 10, 2024

Contributor

I made PR #128 to address this. I think your arch string came from the sse2neon readme. In that project, DLTcollab/sse2neon#502 suggests there is no obvious path to use sse2neon in a mulit-arch universal build. I tried "armv8-a+fp+simd+crypto+crc;x86_64", but it did not work with Apple clang (14). -DCMAKE_OSX_ARCHITECTURES=arm64;x86_64 works now, but I am not certain that it benefits from sse2neon.
endif(APPLE)

### ORC is not used in any active code at the moment ###
# I tried it with 0.4.14
# 0.4.10 did not work (not all opcode implemented)
Expand Down
2 changes: 2 additions & 0 deletions README.md
View file
Original file line number Diff line number Diff line change
@@ -1,5 +1,7 @@
# VidStab

[![C/C++ CI](https://github.com/georgmartius/vid.stab/actions/workflows/c-cpp.yml/badge.svg)](https://github.com/georgmartius/vid.stab/actions/workflows/c-cpp.yml)

Vidstab is a video stabilization library which can be plugged-in with Ffmpeg and Transcode.

**Why is it needed**
Expand Down
7 changes: 7 additions & 0 deletions src/motiondetect_opt.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -37,6 +37,13 @@
#define SSE2_CMP_SUM_ROWS 8
#endif

#ifdef __ARM_NEON__
#include "sse2neon.h"
#define USE_SSE2
#define USE_SSE2_CMP_HOR
#define SSE2_CMP_SUM_ROWS 8
#endif

#ifdef USE_SSE2
/**
\see contrastSubImg using SSE2 optimization, Planar (1 byte per channel) only
Expand Down
Loading

1 comment on commit cbe3d5c

@georgmartius
Copy link
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks, I merged the new PR.

Please sign in to comment.