SDL: Allow optimizing memcpy and memset where possible

From 7a069cc4b0b290d65ee3ceeacf08feb343f3467f Mon Sep 17 00:00:00 2001
From: Sam Lantinga <[EMAIL REDACTED]>
Date: Sat, 20 Jan 2024 08:12:11 -0800
Subject: [PATCH] Allow optimizing memcpy and memset where possible

Modern C runtimes have well optimized memset and memcpy, so use those instead of dispatching into SDL's versions. In addition, some compilers can analyze memset and memcpy calls and directly turn them into optimized assembly.
---
 VisualC-GDK/SDL/SDL.vcxproj             |   3 +
 VisualC-GDK/SDL/SDL.vcxproj.filters     |   3 +
 VisualC-WinRT/SDL-UWP.vcxproj           |   3 +
 VisualC-WinRT/SDL-UWP.vcxproj.filters   |   9 ++
 VisualC/SDL/SDL.vcxproj                 |   5 +-
 VisualC/SDL/SDL.vcxproj.filters         |  26 ++++-
 Xcode/SDL/SDL.xcodeproj/project.pbxproj |  12 ++
 docs/README-migration.md                |   3 +
 include/SDL3/SDL_stdinc.h               |  52 ++++++---
 src/dynapi/SDL_dynapi.c                 |   3 +
 src/stdlib/SDL_memcpy.c                 |  99 +++++++++++++++++
 src/stdlib/SDL_memmove.c                |  55 +++++++++
 src/stdlib/SDL_memset.c                 | 137 +++++++++++++++++++++++
 src/stdlib/SDL_mslibc.c                 |  34 ------
 src/stdlib/SDL_stdlib.c                 | 141 ------------------------
 src/stdlib/SDL_string.c                 |  23 ----
 16 files changed, 389 insertions(+), 219 deletions(-)
 create mode 100644 src/stdlib/SDL_memcpy.c
 create mode 100644 src/stdlib/SDL_memmove.c
 create mode 100644 src/stdlib/SDL_memset.c

diff --git a/VisualC-GDK/SDL/SDL.vcxproj b/VisualC-GDK/SDL/SDL.vcxproj
index 44cefc10a1c8..c064a77f9504 100644
--- a/VisualC-GDK/SDL/SDL.vcxproj
+++ b/VisualC-GDK/SDL/SDL.vcxproj
@@ -762,6 +762,9 @@
     <ClCompile Include="..\..\src\stdlib\SDL_getenv.c" />
     <ClCompile Include="..\..\src\stdlib\SDL_iconv.c" />
     <ClCompile Include="..\..\src\stdlib\SDL_malloc.c" />
+    <ClCompile Include="..\..\src\stdlib\SDL_memcpy.c" />
+    <ClCompile Include="..\..\src\stdlib\SDL_memmove.c" />
+    <ClCompile Include="..\..\src\stdlib\SDL_memset.c" />
     <ClCompile Include="..\..\src\stdlib\SDL_mslibc.c" />
     <MASM Condition="'$(Platform)'=='x64'" Include="..\..\src\stdlib\SDL_mslibc_x64.masm">
       <PrecompiledHeader>NotUsing</PrecompiledHeader>
diff --git a/VisualC-GDK/SDL/SDL.vcxproj.filters b/VisualC-GDK/SDL/SDL.vcxproj.filters
index 4414dd92f898..2767954b1f74 100644
--- a/VisualC-GDK/SDL/SDL.vcxproj.filters
+++ b/VisualC-GDK/SDL/SDL.vcxproj.filters
@@ -145,6 +145,9 @@
     <ClCompile Include="..\..\src\stdlib\SDL_getenv.c" />
     <ClCompile Include="..\..\src\stdlib\SDL_iconv.c" />
     <ClCompile Include="..\..\src\stdlib\SDL_malloc.c" />
+    <ClCompile Include="..\..\src\stdlib\SDL_memcpy.c" />
+    <ClCompile Include="..\..\src\stdlib\SDL_memmove.c" />
+    <ClCompile Include="..\..\src\stdlib\SDL_memset.c" />
     <ClCompile Include="..\..\src\stdlib\SDL_mslibc.c" />
     <ClCompile Include="..\..\src\stdlib\SDL_qsort.c" />
     <ClCompile Include="..\..\src\stdlib\SDL_stdlib.c" />
diff --git a/VisualC-WinRT/SDL-UWP.vcxproj b/VisualC-WinRT/SDL-UWP.vcxproj
index f4760087a144..37188a009c2c 100644
--- a/VisualC-WinRT/SDL-UWP.vcxproj
+++ b/VisualC-WinRT/SDL-UWP.vcxproj
@@ -424,6 +424,9 @@
     <ClCompile Include="..\src\stdlib\SDL_getenv.c" />
     <ClCompile Include="..\src\stdlib\SDL_iconv.c" />
     <ClCompile Include="..\src\stdlib\SDL_malloc.c" />
+    <ClCompile Include="..\src\stdlib\SDL_memcpy.c" />
+    <ClCompile Include="..\src\stdlib\SDL_memmove.c" />
+    <ClCompile Include="..\src\stdlib\SDL_memset.c" />
     <ClCompile Include="..\src\stdlib\SDL_mslibc.c" />
     <MASM Condition="'$(Platform)'=='x64'" Include="..\src\stdlib\SDL_mslibc_x64.masm" >
       <PrecompiledHeader>NotUsing</PrecompiledHeader>
diff --git a/VisualC-WinRT/SDL-UWP.vcxproj.filters b/VisualC-WinRT/SDL-UWP.vcxproj.filters
index 789ec30d3b35..784f5961baa0 100644
--- a/VisualC-WinRT/SDL-UWP.vcxproj.filters
+++ b/VisualC-WinRT/SDL-UWP.vcxproj.filters
@@ -714,6 +714,15 @@
     <ClCompile Include="..\src\stdlib\SDL_malloc.c">
       <Filter>Source Files</Filter>
     </ClCompile>
+    <ClCompile Include="..\src\stdlib\SDL_memcpy.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="..\src\stdlib\SDL_memmove.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="..\src\stdlib\SDL_memset.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
     <ClCompile Include="..\src\stdlib\SDL_qsort.c">
       <Filter>Source Files</Filter>
     </ClCompile>
diff --git a/VisualC/SDL/SDL.vcxproj b/VisualC/SDL/SDL.vcxproj
index b7840a4c411a..6a79e1f2e3f0 100644
--- a/VisualC/SDL/SDL.vcxproj
+++ b/VisualC/SDL/SDL.vcxproj
@@ -616,8 +616,11 @@
     <ClCompile Include="..\..\src\stdlib\SDL_getenv.c" />
     <ClCompile Include="..\..\src\stdlib\SDL_iconv.c" />
     <ClCompile Include="..\..\src\stdlib\SDL_malloc.c" />
+    <ClCompile Include="..\..\src\stdlib\SDL_memcpy.c" />
+    <ClCompile Include="..\..\src\stdlib\SDL_memmove.c" />
+    <ClCompile Include="..\..\src\stdlib\SDL_memset.c" />
     <ClCompile Include="..\..\src\stdlib\SDL_mslibc.c" />
-    <MASM Condition="'$(Platform)'=='x64'" Include="..\..\src\stdlib\SDL_mslibc_x64.masm" >
+    <MASM Condition="'$(Platform)'=='x64'" Include="..\..\src\stdlib\SDL_mslibc_x64.masm">
       <PrecompiledHeader>NotUsing</PrecompiledHeader>
     </MASM>
     <ClCompile Include="..\..\src\stdlib\SDL_qsort.c" />
diff --git a/VisualC/SDL/SDL.vcxproj.filters b/VisualC/SDL/SDL.vcxproj.filters
index 6517484f466f..6971749d59b5 100644
--- a/VisualC/SDL/SDL.vcxproj.filters
+++ b/VisualC/SDL/SDL.vcxproj.filters
@@ -822,6 +822,13 @@
       <Filter>render\direct3d12</Filter>
     </ClInclude>
     <ClInclude Include="..\..\src\hidapi\SDL_hidapi_c.h" />
+    <ClInclude Include="..\..\src\thread\generic\SDL_sysrwlock_c.h" />
+    <ClInclude Include="..\..\src\video\yuv2rgb\yuv_rgb_common.h" />
+    <ClInclude Include="..\..\src\video\yuv2rgb\yuv_rgb_internal.h" />
+    <ClInclude Include="..\..\src\video\yuv2rgb\yuv_rgb_lsx.h" />
+    <ClInclude Include="..\..\src\video\yuv2rgb\yuv_rgb_lsx_func.h" />
+    <ClInclude Include="..\..\src\video\yuv2rgb\yuv_rgb_sse.h" />
+    <ClInclude Include="..\..\src\video\yuv2rgb\yuv_rgb_std.h" />
   </ItemGroup>
   <ItemGroup>
     <ClCompile Include="..\..\src\audio\wasapi\SDL_wasapi.c" />
@@ -1197,9 +1204,6 @@
     <ClCompile Include="..\..\src\video\dummy\SDL_nullvideo.c">
       <Filter>video\dummy</Filter>
     </ClCompile>
-    <ClCompile Include="..\..\src\video\yuv2rgb\yuv_rgb.c">
-      <Filter>video\yuv2rgb</Filter>
-    </ClCompile>
     <ClCompile Include="..\..\src\video\windows\SDL_windowsclipboard.c">
       <Filter>video\windows</Filter>
     </ClCompile>
@@ -1281,6 +1285,15 @@
     <ClCompile Include="..\..\src\stdlib\SDL_malloc.c">
       <Filter>stdlib</Filter>
     </ClCompile>
+    <ClCompile Include="..\..\src\stdlib\SDL_memcpy.c">
+      <Filter>stdlib</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\src\stdlib\SDL_memmove.c">
+      <Filter>stdlib</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\src\stdlib\SDL_memset.c">
+      <Filter>stdlib</Filter>
+    </ClCompile>
     <ClCompile Include="..\..\src\stdlib\SDL_qsort.c">
       <Filter>stdlib</Filter>
     </ClCompile>
@@ -1381,9 +1394,10 @@
     <ClCompile Include="..\..\src\stdlib\SDL_mslibc.c">
       <Filter>stdlib</Filter>
     </ClCompile>
-    <MASM Include="..\..\src\stdlib\SDL_mslibc_x64.masm">
-      <Filter>stdlib</Filter>
-    </MASM>
+    <ClCompile Include="..\..\src\thread\generic\SDL_sysrwlock.c" />
+    <ClCompile Include="..\..\src\video\yuv2rgb\yuv_rgb_lsx.c" />
+    <ClCompile Include="..\..\src\video\yuv2rgb\yuv_rgb_sse.c" />
+    <ClCompile Include="..\..\src\video\yuv2rgb\yuv_rgb_std.c" />
   </ItemGroup>
   <ItemGroup>
     <ResourceCompile Include="..\..\src\core\windows\version.rc" />
diff --git a/Xcode/SDL/SDL.xcodeproj/project.pbxproj b/Xcode/SDL/SDL.xcodeproj/project.pbxproj
index a2af9d4da120..a846955a6bea 100644
--- a/Xcode/SDL/SDL.xcodeproj/project.pbxproj
+++ b/Xcode/SDL/SDL.xcodeproj/project.pbxproj
@@ -372,6 +372,9 @@
 		E4F7981C2AD8D85500669F54 /* SDL_dynapi_unsupported.h in Headers */ = {isa = PBXBuildFile; fileRef = E4F7981B2AD8D85500669F54 /* SDL_dynapi_unsupported.h */; };
 		E4F7981E2AD8D86A00669F54 /* SDL_render_unsupported.c in Sources */ = {isa = PBXBuildFile; fileRef = E4F7981D2AD8D86A00669F54 /* SDL_render_unsupported.c */; };
 		E4F798202AD8D87F00669F54 /* SDL_video_unsupported.c in Sources */ = {isa = PBXBuildFile; fileRef = E4F7981F2AD8D87F00669F54 /* SDL_video_unsupported.c */; };
+		F316ABD82B5C3185002EF551 /* SDL_memset.c in Sources */ = {isa = PBXBuildFile; fileRef = F316ABD62B5C3185002EF551 /* SDL_memset.c */; };
+		F316ABD92B5C3185002EF551 /* SDL_memcpy.c in Sources */ = {isa = PBXBuildFile; fileRef = F316ABD72B5C3185002EF551 /* SDL_memcpy.c */; };
+		F316ABDB2B5CA721002EF551 /* SDL_memmove.c in Sources */ = {isa = PBXBuildFile; fileRef = F316ABDA2B5CA721002EF551 /* SDL_memmove.c */; };
 		F31A92C828D4CB39003BFD6A /* SDL_offscreenopengles.h in Headers */ = {isa = PBXBuildFile; fileRef = F31A92C628D4CB39003BFD6A /* SDL_offscreenopengles.h */; };
 		F31A92D228D4CB39003BFD6A /* SDL_offscreenopengles.c in Sources */ = {isa = PBXBuildFile; fileRef = F31A92C728D4CB39003BFD6A /* SDL_offscreenopengles.c */; };
 		F32305FF28939F6400E66D30 /* SDL_hidapi_combined.c in Sources */ = {isa = PBXBuildFile; fileRef = F32305FE28939F6400E66D30 /* SDL_hidapi_combined.c */; };
@@ -878,6 +881,9 @@
 		E4F7981B2AD8D85500669F54 /* SDL_dynapi_unsupported.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = SDL_dynapi_unsupported.h; sourceTree = "<group>"; };
 		E4F7981D2AD8D86A00669F54 /* SDL_render_unsupported.c */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.c; path = SDL_render_unsupported.c; sourceTree = "<group>"; };
 		E4F7981F2AD8D87F00669F54 /* SDL_video_unsupported.c */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.c; path = SDL_video_unsupported.c; sourceTree = "<group>"; };
+		F316ABD62B5C3185002EF551 /* SDL_memset.c */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.c; path = SDL_memset.c; sourceTree = "<group>"; };
+		F316ABD72B5C3185002EF551 /* SDL_memcpy.c */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.c; path = SDL_memcpy.c; sourceTree = "<group>"; };
+		F316ABDA2B5CA721002EF551 /* SDL_memmove.c */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.c; path = SDL_memmove.c; sourceTree = "<group>"; };
 		F31A92C628D4CB39003BFD6A /* SDL_offscreenopengles.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = SDL_offscreenopengles.h; sourceTree = "<group>"; };
 		F31A92C728D4CB39003BFD6A /* SDL_offscreenopengles.c */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.c; path = SDL_offscreenopengles.c; sourceTree = "<group>"; };
 		F32305FE28939F6400E66D30 /* SDL_hidapi_combined.c */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.c; path = SDL_hidapi_combined.c; sourceTree = "<group>"; };
@@ -1941,6 +1947,9 @@
 				A7D8A8D423E2514000DCD162 /* SDL_getenv.c */,
 				A7D8A8D323E2514000DCD162 /* SDL_iconv.c */,
 				A7D8A8D923E2514000DCD162 /* SDL_malloc.c */,
+				F316ABD72B5C3185002EF551 /* SDL_memcpy.c */,
+				F316ABDA2B5CA721002EF551 /* SDL_memmove.c */,
+				F316ABD62B5C3185002EF551 /* SDL_memset.c */,
 				A7D8A8D723E2514000DCD162 /* SDL_qsort.c */,
 				A7D8A8D823E2514000DCD162 /* SDL_stdlib.c */,
 				A7D8A8D523E2514000DCD162 /* SDL_string.c */,
@@ -2559,6 +2568,7 @@
 				A7D8B9DD23E2514400DCD162 /* SDL_blendpoint.c in Sources */,
 				A7D8B4EE23E2514300DCD162 /* SDL_gamepad.c in Sources */,
 				E4A568B62AF763940062EEC4 /* SDL_sysmain_callbacks.c in Sources */,
+				F316ABD82B5C3185002EF551 /* SDL_memset.c in Sources */,
 				A7D8BA1323E2514400DCD162 /* SDL_render_sw.c in Sources */,
 				A7D8B42223E2514300DCD162 /* SDL_syssem.c in Sources */,
 				A7D8B53923E2514300DCD162 /* SDL_hidapi_xbox360.c in Sources */,
@@ -2610,6 +2620,7 @@
 				A7D8B76A23E2514300DCD162 /* SDL_wave.c in Sources */,
 				5616CA4C252BB2A6005D5928 /* SDL_url.c in Sources */,
 				A7D8BAD323E2514500DCD162 /* s_tan.c in Sources */,
+				F316ABDB2B5CA721002EF551 /* SDL_memmove.c in Sources */,
 				A7D8AA6523E2514000DCD162 /* SDL_hints.c in Sources */,
 				A7D8B53F23E2514300DCD162 /* SDL_hidapi_ps4.c in Sources */,
 				F362B91C2B3349E200D30B94 /* SDL_steam_virtual_gamepad.c in Sources */,
@@ -2623,6 +2634,7 @@
 				F395C1B12569C6A000942BFF /* SDL_mfijoystick.m in Sources */,
 				A7D8B99223E2514400DCD162 /* SDL_shaders_metal.metal in Sources */,
 				F3990DF52A787C10000D8759 /* SDL_sysurl.m in Sources */,
+				F316ABD92B5C3185002EF551 /* SDL_memcpy.c in Sources */,
 				A7D8B97A23E2514400DCD162 /* SDL_render.c in Sources */,
 				A7D8ABD323E2514100DCD162 /* SDL_stretch.c in Sources */,
 				A7D8BAFD23E2514500DCD162 /* s_floor.c in Sources */,
diff --git a/docs/README-migration.md b/docs/README-migration.md
index 6b3b3b85a66c..23ea2d1f2e88 100644
--- a/docs/README-migration.md
+++ b/docs/README-migration.md
@@ -1249,6 +1249,9 @@ M_PI is no longer defined in SDL_stdinc.h, you can use the new symbols SDL_PI_D
 The following functions have been renamed:
 * SDL_strtokr() => SDL_strtok_r()
 
+The following functions have been removed:
+* SDL_memcpy4()
+
 ## SDL_surface.h
 
 The userdata member of SDL_Surface has been replaced with a more general properties interface, which can be queried with SDL_GetSurfaceProperties()
diff --git a/include/SDL3/SDL_stdinc.h b/include/SDL3/SDL_stdinc.h
index 4f864873e126..4cc8480b2f5d 100644
--- a/include/SDL3/SDL_stdinc.h
+++ b/include/SDL3/SDL_stdinc.h
@@ -522,22 +522,45 @@ extern DECLSPEC int SDLCALL SDL_tolower(int x);
 extern DECLSPEC Uint16 SDLCALL SDL_crc16(Uint16 crc, const void *data, size_t len);
 extern DECLSPEC Uint32 SDLCALL SDL_crc32(Uint32 crc, const void *data, size_t len);
 
-extern DECLSPEC void *SDLCALL SDL_memset(SDL_OUT_BYTECAP(len) void *dst, int c, size_t len);
-extern DECLSPEC void *SDLCALL SDL_memset4(void *dst, Uint32 val, size_t dwords);
+extern DECLSPEC void *SDLCALL SDL_memcpy(SDL_OUT_BYTECAP(len) void *dst, SDL_IN_BYTECAP(len) const void *src, size_t len);
 
-#define SDL_zero(x) SDL_memset(&(x), 0, sizeof((x)))
-#define SDL_zerop(x) SDL_memset((x), 0, sizeof(*(x)))
-#define SDL_zeroa(x) SDL_memset((x), 0, sizeof((x)))
+/* Take advantage of compiler optimizations for memcpy */
+#ifndef SDL_SLOW_MEMCPY
+#ifdef SDL_memcpy
+#undef SDL_memcpy
+#endif
+#define SDL_memcpy  memcpy
+#endif
 
 #define SDL_copyp(dst, src)                                                                 \
     { SDL_COMPILE_TIME_ASSERT(SDL_copyp, sizeof (*(dst)) == sizeof (*(src))); }             \
-    SDL_memcpy((dst), (src), sizeof (*(src)))
+    SDL_memcpy((dst), (src), sizeof(*(src)))
 
+extern DECLSPEC void *SDLCALL SDL_memmove(SDL_OUT_BYTECAP(len) void *dst, SDL_IN_BYTECAP(len) const void *src, size_t len);
 
+/* Take advantage of compiler optimizations for memmove */
+#ifndef SDL_SLOW_MEMMOVE
+#ifdef SDL_memmove
+#undef SDL_memmove
+#endif
+#define SDL_memmove memmove
+#endif
 
-extern DECLSPEC void *SDLCALL SDL_memcpy(SDL_OUT_BYTECAP(len) void *dst, SDL_IN_BYTECAP(len) const void *src, size_t len);
+extern DECLSPEC void *SDLCALL SDL_memset(SDL_OUT_BYTECAP(len) void *dst, int c, size_t len);
+extern DECLSPEC void *SDLCALL SDL_memset4(void *dst, Uint32 val, size_t dwords);
+
+/* Take advantage of compiler optimizations for memset */
+#ifndef SDL_SLOW_MEMSET
+#ifdef SDL_memset
+#undef SDL_memset
+#endif
+#define SDL_memset  memset
+#endif
+
+#define SDL_zero(x) SDL_memset(&(x), 0, sizeof((x)))
+#define SDL_zerop(x) SDL_memset((x), 0, sizeof(*(x)))
+#define SDL_zeroa(x) SDL_memset((x), 0, sizeof((x)))
 
-extern DECLSPEC void *SDLCALL SDL_memmove(SDL_OUT_BYTECAP(len) void *dst, SDL_IN_BYTECAP(len) const void *src, size_t len);
 extern DECLSPEC int SDLCALL SDL_memcmp(const void *s1, const void *s2, size_t len);
 
 extern DECLSPEC size_t SDLCALL SDL_wcslen(const wchar_t *wstr);
@@ -730,9 +753,15 @@ size_t wcslcat(wchar_t *dst, const wchar_t *src, size_t size);
 #define SDL_calloc calloc
 #define SDL_realloc realloc
 #define SDL_free free
-#define SDL_memset memset
+#ifndef SDL_memcpy
 #define SDL_memcpy memcpy
+#endif
+#ifndef SDL_memmove
 #define SDL_memmove memmove
+#endif
+#ifndef SDL_memset
+#define SDL_memset memset
+#endif
 #define SDL_memcmp memcmp
 #define SDL_strlcpy strlcpy
 #define SDL_strlcat strlcat
@@ -759,11 +788,6 @@ size_t wcslcat(wchar_t *dst, const wchar_t *src, size_t size);
 #define SDL_vsnprintf vsnprintf
 #endif
 
-SDL_FORCE_INLINE void *SDL_memcpy4(SDL_OUT_BYTECAP(dwords*4) void *dst, SDL_IN_BYTECAP(dwords*4) const void *src, size_t dwords)
-{
-    return SDL_memcpy(dst, src, dwords * 4);
-}
-
 /**
  * If a * b would overflow, return -1. Otherwise store a * b via ret
  * and return 0.
diff --git a/src/dynapi/SDL_dynapi.c b/src/dynapi/SDL_dynapi.c
index 08c18b9100f0..5edc6aa943ce 100644
--- a/src/dynapi/SDL_dynapi.c
+++ b/src/dynapi/SDL_dynapi.c
@@ -26,6 +26,9 @@
 #if SDL_DYNAMIC_API
 
 #define SDL_DYNAMIC_API_ENVVAR "SDL3_DYNAMIC_API"
+#define SDL_SLOW_MEMCPY
+#define SDL_SLOW_MEMMOVE
+#define SDL_SLOW_MEMSET
 
 #ifdef HAVE_STDIO_H
 #include <stdio.h>
diff --git a/src/stdlib/SDL_memcpy.c b/src/stdlib/SDL_memcpy.c
new file mode 100644
index 000000000000..456b1d032ac4
--- /dev/null
+++ b/src/stdlib/SDL_memcpy.c
@@ -0,0 +1,99 @@
+/*
+  Simple DirectMedia Layer
+  Copyright (C) 1997-2024 Sam Lantinga <slouken@libsdl.org>
+
+  This software is provided 'as-is', without any express or implied
+  warranty.  In no event will the authors be held liable for any damages
+  arising from the use of this software.
+
+  Permission is granted to anyone to use this software for any purpose,
+  including commercial applications, and to alter it and redistribute it
+  freely, subject to the following restrictions:
+
+  1. The origin of this software must not be misrepresented; you must not
+     claim that you wrote the original software. If you use this software
+     in a product, an acknowledgment in the product documentation would be
+     appreciated but is not required.
+  2. Altered source versions must be plainly marked as such, and must not be
+     misrepresented as being the original software.
+  3. This notice may not be removed or altered from any source distribution.
+*/
+#include "SDL_internal.h"
+
+
+#ifdef SDL_memcpy
+#undef SDL_memcpy
+#endif
+#if SDL_DYNAMIC_API
+#define SDL_memcpy SDL_memcpy_REAL
+#endif
+void *SDL_memcpy(SDL_OUT_BYTECAP(len) void *dst, SDL_IN_BYTECAP(len) const void *src, size_t len)
+{
+#ifdef __GNUC__
+    /* Presumably this is well tuned for speed.
+       On my machine this is twice as fast as the C code below.
+     */
+    return __builtin_memcpy(dst, src, len);
+#elif defined(HAVE_MEMCPY)
+    return memcpy(dst, src, len);
+#elif defined(HAVE_BCOPY)
+    bcopy(src, dst, len);
+    return dst;
+#else
+    /* GCC 4.9.0 with -O3 will generate movaps instructions with the loop
+       using Uint32* pointers, so we need to make sure the pointers are
+       aligned before we loop using them.
+     */
+    if (((uintptr_t)src & 0x3) || ((uintptr_t)dst & 0x3)) {
+        /* Do an unaligned byte copy */
+        Uint8 *srcp1 = (Uint8 *)src;
+        Uint8 *dstp1 = (Uint8 *)dst;
+
+        while (len--) {
+            *dstp1++ = *srcp1++;
+        }
+    } else {
+        size_t left = (len % 4);
+        Uint32 *srcp4, *dstp4;
+        Uint8 *srcp1, *dstp1;
+
+        srcp4 = (Uint32 *)src;
+        dstp4 = (Uint32 *)dst;
+        len /= 4;
+        while (len--) {
+            *dstp4++ = *srcp4++;
+        }
+
+        srcp1 = (Uint8 *)srcp4;
+        dstp1 = (Uint8 *)dstp4;
+        switch (left) {
+        case 3:
+            *dstp1++ = *srcp1++;
+        case 2:
+            *dstp1++ = *srcp1++;
+        case 1:
+            *dstp1++ = *srcp1++;
+        }
+    }
+    return dst;
+#endif /* __GNUC__ */
+}
+
+/* The optimizer on Visual Studio 2005 and later generates memcpy() and memset() calls.
+   Always provide it for the SDL3 DLL, but skip it when building static lib w/ static runtime. */
+#if defined(_MSC_VER) && (_MSC_VER >= 1400) && (!defined(_MT) || defined(DLL_EXPORT))
+/* NOLINTNEXTLINE(readability-redundant-declaration) */
+extern void *memcpy(void *dst, const void *src, size_t len);
+#ifndef __INTEL_LLVM_COMPILER
+#pragma intrinsic(memcpy)
+#endif
+
+#ifndef __clang__
+#pragma function(memcpy)
+#endif
+/* NOLINTNEXTLINE(readability-inconsistent-declaration-parameter-name) */
+void *memcpy(void *dst, const void *src, size_t len)
+{
+    return SDL_memcpy(dst, src, len);
+}
+#endif /* (_MSC_VER >= 1400) && (!defined(_MT) || defined(DLL_EXPORT)) */
diff --git a/src/stdlib/SDL_memmove.c b/src/stdlib/SDL_memmove.c
new file mode 100644
index 000000000000..1daed2ffb4a2
--- /dev/null
+++ b/src/stdlib/SDL_memmove.c
@@ -0,0 +1,55 @@
+/*
+  Simple DirectMedia Layer
+  Copyright (C) 1997-2024 Sam Lantinga <slouken@libsdl.org>
+
+  This software is provided 'as-is', without any express or implied
+  warranty.  In no event will the authors be held liable for any damages
+  arising from the use of this software.
+
+  Permission is granted to anyone to use this software for any purpose,
+  including commercial applications, and to alter it and redistribute it
+  freely, subject to the following restrictions:
+
+  1. The origin of this software must not be misrepresented; you must not
+     claim that you wrote the original software. If you use this software
+     in a product, an acknowledgment in the product documentation would be
+     appreciated but is not required.
+  2. Altered source versions must be plainly marked as such, and must not be
+     misrepresented as being the original software.
+  3. This notice may not be removed or altered from any source distribution.
+*/
+#include "SDL_internal.h"
+
+
+#ifdef SDL_memmove
+#undef SDL_memmove
+#endif
+#if SDL_DYNAMIC_API
+#define SDL_memmove SDL_memmove_REAL
+#endif
+void *SDL_memmove(SDL_OUT_BYTECAP(len) void *dst, SDL_IN_BYTECAP(len) const void *src, size_t len)
+{
+#ifdef __GNUC__
+    /* Presumably this is well tuned for speed. */
+    return __builtin_memmove(dst, src, len);
+#elif defined(HAVE_MEMMOVE)
+    return memmove(dst, src, len);
+#else
+    char *srcp = (char *)src;
+    char *dstp = (char *)dst;
+
+    if (src < dst) {
+        srcp += len - 1;
+        dstp += len - 1;
+        while (len--) {
+            *dstp-- = *srcp--;
+        }
+    } else {
+        while (len--) {
+            *dstp++ = *srcp++;
+        }
+    }
+    return dst;
+#endif /* HAVE_MEMMOVE */
+}
+
diff --git a/src/stdlib/SDL_memset.c b/src/stdlib/SDL_memset.c
new file mode 100644
index 000000000000..30d2d6a57cb1
--- /dev/null
+++ b/src/stdlib/SDL_memset.c
@@ -0,0 +1,137 @@
+/*
+  Simple DirectMedia Layer
+  Copyright (C) 1997-2024 Sam Lantinga <slouken@libsdl.org>
+
+  This software is provided 'as-is', without any express or implied
+  warranty.  In no event will the authors be held liable for any damages
+  arising from the use of this software.
+
+  Permission is granted to anyone to use this software for any purpose,
+  including commercial applications, and to alter it and redistribute it
+  freely, subject to the following restrictions:
+
+  1. The origin of this software must not be misrepresented; you must not
+     claim that you wrote the original software. If you use this software
+     in a product, an acknowledgment in the product documentation would be
+     appreciated but is not required.
+  2. Altered source versions must be plainly marked as such, and must not be
+     misrepresented as being the original software.
+  3. This notice may not be removed or altered from any source distribution.
+*/
+#include "SDL_internal.h"
+
+
+#ifdef SDL_memset
+#undef SDL_memset
+#endif
+#if SDL_DYNAMIC_API
+#define SDL_memset SDL_memset_REAL
+#endif
+void *SDL_memset(SDL_OUT_BYTECAP(len) void *dst, int c, size_t len)
+{
+#ifdef __GNUC__
+    return __builtin_memset(dst, c, len);
+#elif defined(HAVE_MEMSET)
+    return memset(dst, c, len);
+#else
+    size_t left;
+    Uint32 *dstp4;
+    Uint8 *dstp1 = (Uint8 *)dst;
+    Uint8 value1;
+    Uint32 value4;
+
+    /* The value used in memset() is a byte, passed as an int */
+    c &= 0xff;
+
+    /* The destination pointer needs to be aligned on a 4-byte boundary to
+     * execute a 32-bit set. Set first bytes manually if needed until it is
+     * aligned. */
+    value1 = (Uint8)c;
+    while ((uintptr_t)dstp1 & 0x3) {
+        if (len--) {
+            *dstp1++ = value1;
+        } else {
+            return dst;
+        }
+    }
+
+    value4 = ((Uint32)c | ((Uint32)c << 8) | ((Uint32)c << 16) | ((Uint32)c << 24));
+    dstp4 = (Uint32 *)dstp1;
+    left = (len % 4);
+    len /= 4;
+    while (len--) {
+        *dstp4++ = value4;
+    }
+
+    dstp1 = (Uint8 *)dstp4;
+    switch (left) {
+    case 3:
+        *dstp1++ = value1;
+    case 2:
+        *dstp1++ = value1;
+    case 1:
+        *dstp1++ = value1;
+    }
+
+    return dst;
+#endif /* HAVE_MEMSET */
+}
+
+/* Note that memset() is a byte assignment and this is a 32-bit assignment, so they're not directly equivalent. */
+void *SDL_memset4(void *dst, Uint32 val, size_t dwords)
+{
+#if defined(__APPLE__) && defined(HAVE_STRING_H)
+    memset_pattern4(dst, &val, dwords * 4);
+#elif defined(__GNUC__) && defined(__i386__)
+    int u0, u1, u2;
+    __asm__ __volatile__(
+        "cld \n\t"
+        "rep ; stosl \n\t"
+        : "=&D"(u0), "=&a"(u1), "=&c"(u2)
+        : "0"(dst), "1"(val), "2"(SDL_static_cast(Uint32, dwords))
+        : "memory");
+#else
+    size_t _n = (dwords + 3) / 4;
+    Uint32 *_p = SDL_static_cast(Uint32 *, dst);
+    Uint32 _val = (val);
+    if (dwords == 0) {
+        return dst;
+    }
+    switch (dwords % 4) {
+    case 0:
+        do {
+            *_p++ = _val;
+            SDL_FALLTHROUGH;
+        case 3:
+            *_p++ = _val;
+            SDL_FALLTHROUGH;
+        case 2:
+            *_p++ = _val;
+            SDL_FALLTHROUGH;
+        case 1:
+            *_p++ = _val;
+        } while (--_n);
+    }
+#endif
+    return dst;
+}
+
+/* The optimizer on Visual Studio 2005 and later generates memcpy() and memset() calls.
+   Always provide it for the SDL3 DLL, but skip it when building static lib w/ static runtime. */
+#if defined(_MSC_VER) && (_MSC_VER >= 1400) && (!defined(_MT) || defined(DLL_EXPORT))
+/* NOLINTNEXTLINE(readability-redundant-declaration) */
+extern void *memset(void *dst, int c, size_t len);
+#ifndef __INTEL_LLVM_COMPILER
+#pragma intrinsic(memset)
+#endif
+
+#ifndef __clang__
+#pragma function(memset)
+#endif
+/* NOLINTNEXTLINE(readability-inconsistent-declaration-parameter-name) */
+void *memset(void *dst, int c, size_t len)
+{
+    return SDL_memset(dst, c, len);
+}
+#endif /* (_MSC_VER >= 1400) && (!defined(_MT) || defined(DLL_EXPORT)) */
+
diff --git a/src/stdlib/SDL_mslibc.c b/src/stdlib/SDL_mslibc.c
index eac162dce0f0..43fec0974221 100644
--- a/src/stdlib/SDL_mslibc.c
+++ b/src/stdlib/SDL_mslibc.c
@@ -33,40 +33,6 @@
 __declspec(selectany) int _fltused = 1;
 #endif
 
-/* The optimizer on Visual Studio 2005 and later generates memcpy() and memset() calls.
-   Always provide it for the SDL3 DLL, but skip it when building static lib w/ static runtime. */
-#if (_MSC_VER >= 1400) && (!defined(_MT) || defined(DLL_EXPORT))
-/* NOLINTNEXTLINE(readability-redundant-declaration) */
-extern void *memcpy(void *dst, const void *src, size_t len);
-#ifndef __INTEL_LLVM_COMPILER
-#pragma intrinsic(memcpy)
-#endif
-
-#ifndef __clang__
-#pragma function(memcpy)
-#endif
-/* NOLINTNEXTLINE(readability-inconsistent-declaration-parameter-name) */
-void *memcpy(void *dst, const void *src, size_t len)
-{
-    return SDL_memcpy(dst, src, len);
-}
-
-/* NOLINTNEXTLINE(readability-redundant-declaration) */
-extern void *memset(void *dst, int c, size_t len);
-#ifndef __INTEL_LLVM_COMPILER
-#pragma intrinsic(memset)
-#endif
-
-#ifndef __clang__
-#pragma function(memset)
-#endif
-/* NOLINTNEXTLINE(readability-inconsistent-declaration-parameter-name) */
-void *memset(void *dst, int c, size_t len)
-{
-    return SDL_memset(dst, c, len);
-}
-#endif /* (_MSC_VER >= 1400) && (!defined(_MT) || defined(DLL_EXPORT)) */
-
 #ifdef _M_IX86
 
 /* Float to long */
diff --git a/src/stdlib/SDL_stdlib.c b/src/stdlib/SDL_stdlib.c
index 97cab4c5267b..edce6e4aee0e 100644
--- a/src/stdlib/SDL_stdlib.c
+++ b/src/stdlib/SDL_stdlib.c
@@ -510,147 +510,6 @@ int SDL_toupper(int x) { return ((x) >= 'a') && ((x) <= 'z') ? ('A' + ((x) - 'a'
 int SDL_tolower(int x) { return ((x) >= 'A') && ((x) <= 'Z') ? ('a' + ((x) - 'A')) : (x); }
 #endif
 
-/* This file contains a portable memcpy manipulation function for SDL */
-
-void *SDL_memcpy(SDL_OUT_BYTECAP(len) void *dst, SDL_IN_BYTECAP(len) const void *src, size_t len)
-{
-#ifdef __GNUC__
-    /* Presumably this is well tuned for speed.
-       On my machine this is twice as fast as the C code below.
-     */
-    return __builtin_memcpy(dst, src, len);
-#elif defined(HAVE_MEMCPY)
-    return memcpy(dst, src, len);
-#elif defined(HAVE_BCOPY)
-    bcopy(src, dst, len);
-    return dst;
-#else
-    /* GCC 4.9.0 with -O3 will generate movaps instructions with the loop
-       using Uint32* pointers, so we need to make sure the pointers are
-       aligned before we loop using them.
-     */
-    if (((uintptr_t)src & 0x3) || ((uintptr_t)dst & 0x3)) {
-        /* Do an unaligned byte copy */
-        Uint8 *srcp1 = (Uint8 *)src;
-        Uint8 *dstp1 = (Uint8 *)dst;
-
-        while (len--) {
-            *dstp1++ = *srcp1++;
-        }
-    } else {
-        size_t left = (len % 4);
-        Uint32 *srcp4, *dstp4;
-        Uint8 *srcp1, *dstp1;
-
-        srcp4 = (Uint32 *)src;
-        dstp4 = (Uint32 *)dst;
-        len /= 4;
-        while (len--) {
-            *dstp4++ = *srcp4++;
-        }
-
-        srcp1 = (Uint8 *)srcp4;
-        dstp1 = (Uint8 *)dstp4;
-        switch (left) {
-        case 3:
-            *dstp1++ = *srcp1++;
-        case 2:
-            *dstp1++ = *srcp1++;
-        case 1:
-            *dstp1++ = *srcp1++;
-        }
-    }
-    return dst;
-#endif /* __GNUC__ */
-}
-
-void *SDL_memset(SDL_OUT_BYTECAP(len) void *dst, int c, size_t len)
-{
-#ifdef HAVE_MEMSET
-    return memset(dst, c, len);
-#else
-    size_t left;
-    Uint32 *dstp4;
-    Uint8 *dstp1 = (Uint8 *)dst;
-    Uint8 value1;
-    Uint32 value4;
-
-    /* The value used in memset() is a byte, passed as an int */
-    c &= 0xff;
-
-    /* The destination pointer needs to be aligned on a 4-byte boundary to
-     * execute a 32-bit set. Set first bytes manually if needed until it is
-     * aligned. */
-    value1 = (Uint8)c;
-    while ((uintptr_t)dstp1 & 0x3) {
-        if (len--) {
-            *dstp1++ = value1;
-        } else {
-            return dst;
-        }
-    }
-
-    value4 = ((Uint32)c | ((Uint32)c << 8) | ((Uint32)c << 16) | ((Uint32)c << 24));
-    dstp4 = (Uint32 *)dstp1;
-    left = (len % 4);
-    len /= 4;
-    while (len--) {
-        *dstp4++ = value4;
-    }
-
-    dstp1 = (Uint8 *)dstp4;
-    switch (left) {
-    case 3:
-        *dstp1++ = value1;
-    case 2:
-        *dstp1++ = value1;
-    case 1:
-        *dstp1++ = value1;
-    }
-
-    return dst;
-#endif /* HAVE_MEMSET */
-}
-
-/* Note that memset() is a byte assignment and this is a 32-bit assignment, so the

(Patch may be truncated, please check the link at the top of this post.)