diff --git a/.gitmodules b/.gitmodules
index e69de29..49f1431 100644
--- a/.gitmodules
+++ b/.gitmodules
@@ -0,0 +1,3 @@
+[submodule "external/xsimd"]
+	path = external/xsimd
+	url = https://github.com/xtensor-stack/xsimd.git
diff --git a/CMakeLists.txt b/CMakeLists.txt
index aa74c79..fff8c15 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -1,6 +1,6 @@
 cmake_minimum_required(VERSION 3.18)
 include(tools/CompilerWarnings.cmake)
-include(tools/CPUFeatures.cmake)
+include(tools/SIMDFlags.cmake)
 set(CMAKE_VERBOSE_MAKEFILE ON)
 
 project(quicktex)
@@ -10,6 +10,8 @@ find_package(Python COMPONENTS Interpreter Development.Module)
 find_package(pybind11 CONFIG REQUIRED)
 find_package(OpenMP)
 
+add_subdirectory(external/xsimd)
+
 # Collect source files
 file(GLOB SOURCE_FILES
         "quicktex/*.cpp"
@@ -51,7 +53,7 @@ if (OpenMP_CXX_FOUND)
     target_link_libraries(_quicktex PUBLIC OpenMP::OpenMP_CXX)
 endif ()
 
-target_link_libraries(_quicktex PUBLIC hwy)
+target_link_libraries(_quicktex PUBLIC xsimd)
 
 # Set module features, like C/C++ standards
 target_compile_features(_quicktex PUBLIC cxx_std_17 c_std_11)
diff --git a/external/xsimd b/external/xsimd
new file mode 160000
index 0000000..1577b02
--- /dev/null
+++ b/external/xsimd
@@ -0,0 +1 @@
+Subproject commit 1577b02d549cca52aa5e943c16f2600950480289
diff --git a/quicktex/tests/_bindings.cpp b/quicktex/tests/_bindings.cpp
index 46134a2..e5b18fc 100644
--- a/quicktex/tests/_bindings.cpp
+++ b/quicktex/tests/_bindings.cpp
@@ -39,6 +39,6 @@ using namespace quicktex::tests;
 void InitCTests(py::module_ &quicktex) {
     py::module_ ctests = quicktex.def_submodule("_ctests", "Internal tests for C-level functions");
 
-    ctests.def("test_WidenSumS16", &TestWidenSumS16);
+//    ctests.def("test_WidenSumS16", &TestWidenSumS16);
 }
 }  // namespace quicktex::bindings
\ No newline at end of file
diff --git a/tools/CPUFeatures.cmake b/tools/CPUFeatures.cmake
deleted file mode 100644
index 2ee4425..0000000
--- a/tools/CPUFeatures.cmake
+++ /dev/null
@@ -1,70 +0,0 @@
-function(set_simd_flags target_name)
-    if (DEFINED ENV{QUICKTEX_HWY_MODE})
-        set(highway_mode $ENV{QUICKTEX_HWY_MODE})
-        message("Highway mode is ${highway_mode}")
-    else ()
-        message("Defaulting to AUTO highway mode")
-        set(highway_mode "AUTO")
-    endif ()
-
-    if ((CMAKE_OSX_ARCHITECTURES MATCHES ".*x86_64.*") OR (CMAKE_SYSTEM_PROCESSOR MATCHES "(x86)|(X86)|(amd64)|(AMD64)"))
-        set(X86 TRUE)
-        message("X86 Detected")
-    else ()
-        set(X86 FALSE)
-    endif ()
-
-    if ((CMAKE_OSX_ARCHITECTURES MATCHES ".*arm64.*") OR (CMAKE_SYSTEM_PROCESSOR MATCHES "(arm64)|(ARM64)|(aarch64)"))
-        set(ARM TRUE)
-        message("ARM Detected")
-    else ()
-        set(ARM FALSE)
-    endif ()
-
-    if (highway_mode STREQUAL "SCALAR")
-        # force Highway to use scalar ops. This should really only be used for testing
-        if (MSVC)
-            target_compile_options(${target_name} PUBLIC /DHWY_COMPILE_ONLY_SCALAR=1)
-        else ()
-            target_compile_options(${target_name} PUBLIC -DHWY_COMPILE_ONLY_SCALAR=1)
-        endif ()
-        return()
-    endif ()
-
-    # dynamic disbatch is not supported
-    if (MSVC)
-        target_compile_options(${target_name} PUBLIC /DHWY_COMPILE_ONLY_STATIC=1)
-    else ()
-        target_compile_options(${target_name} PUBLIC -DHWY_COMPILE_ONLY_STATIC=1)
-    endif ()
-
-    if (X86)
-        if (highway_mode STREQUAL "AUTO")
-            # setting -march=native on an M1 causes Clang to freak out
-            if (MSVC)
-                #MSVC has no -march=native equivalent. womp
-                message(WARNING "Compiling using MSVC without settig an explicit QUICKTEX_HWY_MODE defaults to serial operations. Please compile with Clang if you need vectorization")
-            elseif (!ARM)
-                target_compile_options(${target_name} PUBLIC -march=native)
-            endif ()
-        elseif (highway_mode STREQUAL "SSSE3")
-            if (MSVC)
-                message(SEND_ERROR "Compiling using SSSE3 is not supported with the MSVC compiler. Please use AVX or compile withClang")
-            else ()
-                target_compile_options(${target_name} PUBLIC -mssse3)
-            endif ()
-        elseif (highway_mode STREQUAL "SSE4")
-            if (MSVC)
-                message(SEND_ERROR "Compiling using SSE4 is not supported with the MSVC compiler. Please use AVX or compile with Clang")
-            else ()
-                target_compile_options(${target_name} PUBLIC -msse4)
-            endif ()
-        elseif (highway_mode STREQUAL "AVX2")
-            if (MSVC)
-                target_compile_options(${target_name} PUBLIC /arch:AVX2)
-            else ()
-                target_compile_options(${target_name} PUBLIC -mavx2)
-            endif ()
-        endif ()
-    endif ()
-endfunction()
\ No newline at end of file
diff --git a/tools/SIMDFlags.cmake b/tools/SIMDFlags.cmake
new file mode 100644
index 0000000..9ccd5e7
--- /dev/null
+++ b/tools/SIMDFlags.cmake
@@ -0,0 +1,64 @@
+function(set_simd_flags target_name)
+    if (DEFINED ENV{QUICKTEX_SIMD_MODE})
+        set(simd_mode $ENV{QUICKTEX_SIMD_MODE})
+        message("SIMD mode is ${simd_mode}")
+    else ()
+        message("Defaulting to AUTO SIMD mode. Resulting binary is not fit for distributing to other computers!")
+        set(simd_mode "AUTO")
+    endif ()
+
+    if ((CMAKE_OSX_ARCHITECTURES MATCHES "x86_64") OR (CMAKE_SYSTEM_PROCESSOR MATCHES "(x86)|(X86)|(amd64)|(AMD64)"))
+        set(X86 TRUE)
+        message("X86 Detected")
+    else ()
+        set(X86 FALSE)
+    endif ()
+
+    if ((CMAKE_OSX_ARCHITECTURES MATCHES "arm64") OR (CMAKE_SYSTEM_PROCESSOR MATCHES "(arm64)|(ARM64)|(aarch64)"))
+        set(ARM TRUE)
+        message("ARM Detected")
+    else ()
+        set(ARM FALSE)
+    endif ()
+
+    if (simd_mode STREQUAL "SCALAR")
+        # force xsimd to use scalar ops. This should really only be used for testing,
+        # since SSE2 and NEON are guranteed on 64-bit platforms
+        if (MSVC)
+            target_compile_options(${target_name} PUBLIC /DXSIMD_NO_SUPPORTED_ARCHITECTURE=1)
+        else ()
+            target_compile_options(${target_name} PUBLIC -DXSIMD_NO_SUPPORTED_ARCHITECTURE=1)
+        endif ()
+        return()
+    endif ()
+
+    if (X86)
+        if (simd_mode STREQUAL "AUTO")
+            if (MSVC)
+                #MSVC has no -march=native equivalent. womp
+            elseif (!ARM)
+                # setting -march=native on an M1 causes Clang to freak out,
+                # and arm64 is pretty samey instruction set wise (arm9 and SVE2 notwithstanding)
+                target_compile_options(${target_name} PUBLIC -march=native)
+            endif ()
+        elseif (simd_mode STREQUAL "SSSE3")
+            if (MSVC)
+                target_compile_options(${target_name} PUBLIC /DXSIMD_WITH_SSSE3)
+            else ()
+                target_compile_options(${target_name} PUBLIC -mssse3)
+            endif ()
+        elseif (simd_mode STREQUAL "SSE4")
+            if (MSVC)
+                target_compile_options(${target_name} PUBLIC /DXSIMD_WITH_SSE4_2 /d2archSSE42)
+            else ()
+                target_compile_options(${target_name} PUBLIC -msse4)
+            endif ()
+        elseif (simd_mode STREQUAL "AVX2")
+            if (MSVC)
+                target_compile_options(${target_name} PUBLIC /arch:AVX2)
+            else ()
+                target_compile_options(${target_name} PUBLIC -mavx2)
+            endif ()
+        endif ()
+    endif ()
+endfunction()
\ No newline at end of file