Merge c0rp3n/fastnoise-simd to torss/FastNoiseSIMD

.gitignore

@@ -262,3 +262,23 @@ paket-files/
 /TEST/FastNoiseSIMD Preview
 /cellModel.skp
 /cellModel.skb
+
+# CMake
+CMakeLists.txt.user
+CMakeCache.txt
+CMakeFiles
+CMakeScripts
+Testing
+Makefile
+cmake_install.cmake
+install_manifest.txt
+compile_commands.json
+CTestTestfile.cmake
+_deps
+[Bb]uild
+
+# VSCode
+.vscode/
+
+# FastNoiseSIMD
+FastNoiseSIMD_config.h

.travis.yml

@@ -0,0 +1,30 @@
+sudo: required
+
+# Enable C++ support
+language: cpp
+
+# Use Linux by default
+os: linux
+
+# Use Ubuntu 18.04 for GCC 7.4.0
+dist: bionic
+
+# Compiler selection
+compiler: gcc
+
+install:
+  - DEPS_DIR="${TRAVIS_BUILD_DIR}/deps"
+  - mkdir -p ${DEPS_DIR} && cd ${DEPS_DIR}
+  - |
+    CMAKE_URL="https://cmake.org/files/v3.16/cmake-3.16.2-Linux-x86_64.tar.gz"
+    mkdir cmake && travis_retry wget --no-check-certificate --quiet -O - ${CMAKE_URL} | tar --strip-components=1 -xz -C cmake
+    export PATH=${DEPS_DIR}/cmake/bin:${PATH}
+  - ../install-catch2.sh
+
+# Build steps
+script:
+  - cd ${TRAVIS_BUILD_DIR}
+  - mkdir build
+  - cd build
+  - cmake -DFN_COMPILE_AVX2=ON -DFN_COMPILE_AVX512=OFF -DBUILD_TESTING=ON .. && make
+  - ctest

CMakeLists.txt

@@ -1,91 +1,153 @@
-# This is currently very rudimentary and likely needs various improvements.
+cmake_minimum_required(VERSION 3.10)
 
-cmake_minimum_required(VERSION 3.7) # Requiring 3.7 might be incorrect.
-project(FastNoiseSIMD)
+# detect if FastNoiseSIMD is being bundled,
+# disable testsuite in that case
+if(NOT DEFINED PROJECT_NAME)
+  set(NOT_SUBPROJECT ON)
+endif()
+
+project(FastNoiseSIMD
+    VERSION 0.7.0
+    DESCRIPTION "C++ SIMD Noise Library"
+    LANGUAGES CXX
+)
 
-set(CMAKE_CXX_STANDARD 11) # Requiring C++11 might be incorrect.
-set(CMAKE_CXX_STANDARD_REQUIRED ON)
-set(CMAKE_CXX_EXTENSIONS OFF)
+include(GNUInstallDirs)
+include(CMakePackageConfigHelpers)
 
-set(directory ${PROJECT_NAME})
-set(library_name ${PROJECT_NAME})
-set(file_prefix ${PROJECT_NAME})
+if(BUILD_TESTING)
+    # Catch2 needed for testing
+    find_package(Catch2 CONFIG REQUIRED)
+endif()
 
-set_property(DIRECTORY ${PROJECT_SOURCE_DIR} PROPERTY VS_STARTUP_PROJECT ${library_name})
+# Do stuff depending on the compiler
+if(CMAKE_CXX_COMPILER_ID MATCHES "GNU" OR CMAKE_CXX_COMPILER_ID MATCHES "Clang")
+    set(CMAKE_CXX_FLAGS "-W -Wall -Wextra -Wpedantic -Wunused-value -Wold-style-cast")
+    set(CMAKE_CXX_FLAGS_DEBUG "-g -O0")
+    set(CMAKE_CXX_FLAGS_RELEASE "-O3")
+elseif(CMAKE_CXX_COMPILER_ID MATCHES "MSVC")
+    set(CMAKE_CXX_FLAGS "/W4")
+    set(CMAKE_CXX_FLAGS_DEBUG "/O0 /ZI")
+    set(CMAKE_CXX_FLAGS_RELEASE "/O2 /Ob2")
+endif()
+
+set_property(DIRECTORY ${PROJECT_SOURCE_DIR} PROPERTY VS_STARTUP_PROJECT FastNoiseSIMD)
 
 if(CMAKE_SYSTEM_PROCESSOR MATCHES "arm" OR CMAKE_SYSTEM_PROCESSOR MATCHES "aarch64")
-	option(FN_COMPILE_NEON
-		"Only on arm or aarch64."
-		ON)
+    option(FN_COMPILE_NEON
+        "Only on arm or aarch64."
+        ON)
+endif()
+
+option(FN_COMPILE_SSE2 "" ON)
+
+option(FN_COMPILE_SSE41 "" ON)
+
+option(
+    FN_COMPILE_AVX2 
+    "This does not break support for pre AVX CPUs, AVX code is only run if \
+    support is detected." 
+    OFF
+)
+
+option(FN_COMPILE_AVX512 "Only the latest compilers will support this." OFF)
+
+option(
+    FN_USE_FMA 
+    "Using FMA instructions with AVX(51)2/NEON provides a small performance \
+    increase but can cause minute variations in noise output compared to other \
+    SIMD levels due to higher calculation precision."
+    ON
+)
+
+option(
+    FN_ALIGNED_SETS
+    "Using aligned sets of memory for float arrays allows faster storing of \
+    SIMD data."
+    ON
+)
+
+configure_file(
+    ${CMAKE_CURRENT_LIST_DIR}/cmake/FastNoiseSIMD_config.h.in
+    ${CMAKE_CURRENT_LIST_DIR}/include/FastNoiseSIMD/FastNoiseSIMD_config.h
+    @ONLY
+)
+
+add_library(FastNoiseSIMD STATIC
+    src/FastNoiseSIMD.cpp
+    src/FastNoiseSIMD_avx2.cpp
+    src/FastNoiseSIMD_avx512.cpp
+    src/FastNoiseSIMD_internal.cpp
+    src/FastNoiseSIMD_neon.cpp
+    src/FastNoiseSIMD_sse2.cpp
+    src/FastNoiseSIMD_sse41.cpp
+)
+
+set_target_properties(FastNoiseSIMD
+    PROPERTIES 
+        CXX_STANDARD 11
+        CXX_STANDARD_REQUIRED ON
+        CMAKE_DEBUG_POSTFIX "d"
+)
+
+target_include_directories(FastNoiseSIMD PUBLIC
+    $<BUILD_INTERFACE:${CMAKE_CURRENT_LIST_DIR}/include>
+    $<INSTALL_INTERFACE:${CMAKE_INSTALL_INCLUDEDIR}>
+)
+
+set(FN_CXX_FLAGS)
+if(${FN_COMPILE_AVX512})
+    if(${MSVC})
+        list(APPEND FN_CXX_FLAGS "/arch:AVX512")
+    else()
+        list(APPEND FN_CXX_FLAGS "-march=skylake-avx512")
+    endif()
+elseif(${FN_COMPILE_AVX2})
+    if(${MSVC})
+        list(APPEND FN_CXX_FLAGS "/arch:AVX2")
+    else()
+        list(APPEND FN_CXX_FLAGS "-march=core-avx2")
+    endif()
 endif()
 
-option(FN_COMPILE_SSE2
-	""
-	ON)
-option(FN_COMPILE_SSE41
-	""
-	ON)
-option(FN_COMPILE_AVX2
-	"This does not break support for pre AVX CPUs, AVX code is only run if support is detected."
-	OFF)
-option(FN_COMPILE_AVX512
-	"Only the latest compilers will support this."
-	OFF)
-option(FN_USE_FMA
-	"Using FMA instructions with AVX(51)2/NEON provides a small performance increase but can cause \
-minute variations in noise output compared to other SIMD levels due to higher calculation precision."
-	ON)
-option(FN_ALIGNED_SETS
-	"Using aligned sets of memory for float arrays allows faster storing of SIMD data."
-	ON)
-
-configure_file(${directory}/${file_prefix}.cmake.h include/${file_prefix}.h)
-
-set(flags)
-if(${MSVC})
-	if(${FN_COMPILE_AVX2} OR ${FN_COMPILE_AVX512})
-		list(APPEND flags "/arch:AVX2")
-	endif()
-elseif()
-	if(${FN_COMPILE_AVX2} OR ${FN_COMPILE_AVX512})
-		list(APPEND flags "-march=core-avx2")
-	endif()
+target_compile_options(FastNoiseSIMD PRIVATE "${FN_CXX_FLAGS}")
+
+if(BUILD_TESTING)
+    include(test/tests.cmake)
 endif()
 
-add_library(${library_name} STATIC
-	${directory}/${file_prefix}.cpp
-	${directory}/${file_prefix}_avx2.cpp
-	${directory}/${file_prefix}_avx512.cpp
-	${directory}/${file_prefix}_internal.cpp
-	${directory}/${file_prefix}_neon.cpp
-	${directory}/${file_prefix}_sse2.cpp
-	${directory}/${file_prefix}_sse41.cpp
-	)
-
-target_include_directories(${library_name} PRIVATE
-	${CMAKE_CURRENT_BINARY_DIR}/include
-	${directory}/
-	)
-
-set(public_headers
-	${CMAKE_CURRENT_BINARY_DIR}/include/${file_prefix}.h
-	# ${directory}/${file_prefix}_internal.h
-	)
-set_target_properties(${library_name} PROPERTIES
-	PUBLIC_HEADER "${public_headers}"
-	)
-
-target_compile_options(${library_name} PRIVATE
-	"${flags}"
-	)
-
-install(TARGETS ${library_name}
-	CONFIGURATIONS Debug
-	ARCHIVE DESTINATION lib/Debug
-	PUBLIC_HEADER DESTINATION include
-	)
-install(TARGETS ${library_name}
-	CONFIGURATIONS Release
-	ARCHIVE DESTINATION lib/Release
-	PUBLIC_HEADER DESTINATION include
-	)
+# Only perform the installation steps when not being used as
+# a subproject via `add_subdirectory`, or the destinations will break
+if(NOT_SUBPROJECT)
+    set(FN_CMAKE_CONFIG_DESTINATION "${CMAKE_INSTALL_LIBDIR}/cmake/FastNoiseSIMD")
+
+    configure_package_config_file(
+        ${CMAKE_CURRENT_LIST_DIR}/cmake/FastNoiseSIMDConfig.cmake.in
+        ${CMAKE_CURRENT_BINARY_DIR}/FastNoiseSIMDConfig.cmake
+        INSTALL_DESTINATION ${FN_CMAKE_CONFIG_DESTINATION}
+    )
+
+    # create and install an export set for FastNoiseSIMD target as FastNoiseSIMD
+    install(
+        TARGETS FastNoiseSIMD EXPORT FastNoiseSIMDTargets
+        DESTINATION ${CMAKE_INSTALL_LIBDIR}
+    )
+
+
+    install(EXPORT FastNoiseSIMDTargets DESTINATION ${FN_CMAKE_CONFIG_DESTINATION})
+
+    write_basic_package_version_file(
+        "${CMAKE_CURRENT_BINARY_DIR}/FastNoiseSIMDConfigVersion.cmake"
+        COMPATIBILITY SameMajorVersion
+    )
+
+    install(TARGETS FastNoiseSIMD LIBRARY DESTINATION "${CMAKE_INSTALL_LIBDIR}")
+    install(DIRECTORY "include/" DESTINATION "${CMAKE_INSTALL_INCLUDEDIR}")
+
+    install(
+        FILES
+            "${CMAKE_CURRENT_BINARY_DIR}/FastNoiseSIMDConfig.cmake"
+            "${CMAKE_CURRENT_BINARY_DIR}/FastNoiseSIMDConfigVersion.cmake"
+        DESTINATION ${FN_CMAKE_CONFIG_DESTINATION}
+    )
+endif()

README.md

@@ -1,7 +1,20 @@
-# FastNoise SIMD
-FastNoise SIMD is the SIMD implementation of my noise library [FastNoise](https://github.com/Auburns/FastNoise). It aims to provide faster performance through the use of intrinsic(SIMD) CPU functions. Vectorisation of the code allows noise functions to process data in sets of 4/8/16 increasing performance by 700% in some cases (Simplex).
-
-After releasing FastNoise I got in contact with the author of [FastNoise SIMD](https://github.com/jackmott/FastNoise-SIMD) (naming is coincidence) and was inspired to work with SIMD functions myself. Through his code and discussions with him I created my implementation with even more optimisation thanks to the removal of lookup tables. 
+<h1 align="center">
+    FastNoise SIMD
+</h1>
+<p align="center">
+    <a href="https://travis-ci.org/open-terra/terra">
+        <img src="https://img.shields.io/travis/c0rp3n/fastnoise-simd/master.svg?label=Travis&style=flat-square&logo=travis" alt="Travis Build Status">
+    </a>
+    <a href="https://ci.appveyor.com/project/C0RP3N/terra">
+        <img src="https://img.shields.io/appveyor/ci/C0RP3N/fastnoise-simd/master.svg?label=AppVeyor&style=flat-square&logo=appveyor" alt="Appveyor Build Status">
+    </a>
+    <br>
+    <strong>FastNoise SIMD is the SIMD implementation of <a href="https://github.com/Auburns/FastNoise">FastNoise</a>.</strong>
+</p>
+
+FastNoise SIMD aims to provide faster performance through the use of intrinsic(SIMD) CPU functions. Vectorisation of the code allows noise functions to process data in sets of 4/8/16 increasing performance by 700% in some cases (Simplex).
+
+Inspired by [FastNoise SIMD](https://github.com/jackmott/FastNoise-SIMD) (naming is coincidence). FastNoise SIMD was created with even more optimisation thanks to the removal of lookup tables.
 
 Runtime detection of highest supported instruction set ensures the fastest possible performance with only 1 compile needed. If no support is found it will fallback to standard types (float/int).
 

appveyor.yml

@@ -0,0 +1,25 @@
+version: 1.0.{build}
+image: Visual Studio 2019
+configuration: Debug
+clone_depth: 50
+platform: x64
+
+cache:
+  - c:\tools\vcpkg\installed
+
+install:
+  - vcpkg install Catch2:x64-windows
+  - vcpkg integrate install
+
+before_build:
+  - mkdir build
+  - cd build
+  - cmake -DFN_COMPILE_AVX2=ON -DFN_COMPILE_AVX512=OFF -DBUILD_TESTING=ON -DCMAKE_TOOLCHAIN_FILE=c:/tools/vcpkg/scripts/buildsystems/vcpkg.cmake -DVCPKG_TARGET_TRIPLET=x64-windows -G "Visual Studio 16 2019" ..
+
+build:
+  project: build/FastNoiseSIMD.sln
+  parallel: true
+  verbosity: minimal
+
+test_script:
+  - ctest

cmake/FastNoiseSIMD.pc.in

@@ -0,0 +1,7 @@
+includedir=@CMAKE_INSTALL_FULL_INCLUDEDIR@
+
+Name: FastNoiseSIMD
+Description: 
+URL: https://github.com/c0rp3n/fastnoise-simd
+Version: @FastNoiseSIMD_VERSION@
+Cflags: -I${includedir}

cmake/FastNoiseSIMDConfig.cmake.in

@@ -0,0 +1,6 @@
+@PACKAGE_INIT@
+
+# Provide path for scripts
+list(APPEND CMAKE_MODULE_PATH "${CMAKE_CURRENT_LIST_DIR}")
+
+include(${CMAKE_CURRENT_LIST_DIR}/FastNoiseSIMDTargets.cmake)

cmake/FastNoiseSIMD_config.h.in

@@ -0,0 +1,33 @@
+#ifndef FASTNOISE_SIMD_CONFIG_H
+#define FASTNOISE_SIMD_CONFIG_H
+
+#pragma once
+
+#if defined(__arm__) || defined(__aarch64__)
+#define FN_ARM
+//#define FN_IOS
+#cmakedefine FN_COMPILE_NEON
+#else
+
+// Comment out lines to not compile for certain instruction sets
+#cmakedefine FN_COMPILE_SSE2
+#cmakedefine FN_COMPILE_SSE41
+
+// To compile AVX2 set C++ code generation to use /arch:AVX(2) on FastNoiseSIMD_avx2.cpp
+// Note: This does not break support for pre AVX CPUs, AVX code is only run if support is detected
+#cmakedefine FN_COMPILE_AVX2
+
+// Only the latest compilers will support this
+#cmakedefine FN_COMPILE_AVX512
+
+// Using FMA instructions with AVX(51)2/NEON provides a small performance increase but can cause 
+// minute variations in noise output compared to other SIMD levels due to higher calculation precision
+// Intel compiler will always generate FMA instructions, use /Qfma- or -no-fma to disable
+#cmakedefine FN_USE_FMA
+#endif
+
+// Using aligned sets of memory for float arrays allows faster storing of SIMD data
+// Comment out to allow unaligned float arrays to be used as sets
+#cmakedefine FN_ALIGNED_SETS
+
+#endif