[video_core/vic] remove handrolled vector implementation that also uses software prefetching(!!!!); don't try to outsmart compiler - also remove template spam (#2856)

Also removes sse2neon :)
Software prefetching SUCKS and it's evil don't do it
Signed-off-by: lizzie <lizzie@eden-emu.dev>

Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/2856
Reviewed-by: crueter <crueter@eden-emu.dev>
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
Reviewed-by: Caio Oliveira <caiooliveirafarias0@gmail.com>
Co-authored-by: lizzie <lizzie@eden-emu.dev>
Co-committed-by: lizzie <lizzie@eden-emu.dev>

.ci/license-header.sh

@@ -4,7 +4,7 @@
 # SPDX-License-Identifier: GPL-3.0-or-later
 
 # specify full path if dupes may exist
-EXCLUDE_FILES="CPM.cmake CPMUtil.cmake GetSCMRev.cmake sse2neon.h renderdoc_app.h tools/cpm tools/shellcheck.sh tools/update-cpm.sh externals/stb externals/glad externals/getopt externals/gamemode externals/FidelityFX-FSR externals/demangle externals/bc_decoder"
+EXCLUDE_FILES="CPM.cmake CPMUtil.cmake GetSCMRev.cmake renderdoc_app.h tools/cpm tools/shellcheck.sh tools/update-cpm.sh externals/stb externals/glad externals/getopt externals/gamemode externals/FidelityFX-FSR externals/demangle externals/bc_decoder"
 
 # license header constants, please change when needed :))))
 YEAR=2025

.patch/sse2neon/0001-Add-support-for-clang-cl-on-Windows-633.patch

@@ -1,129 +0,0 @@
-From d765ebed3598ddfd7167fc546474626ac5ef9498 Mon Sep 17 00:00:00 2001
-From: Anthony Roberts <anthony.roberts@linaro.org>
-Date: Fri, 2 Aug 2024 16:55:57 +0100
-Subject: [PATCH] Add support for clang-cl on Windows (#633)
-
-This commit adds support for clang-cl (clang, pretending to be MSVC) to
-SSE2NEON on Windows ARM64 platforms. This change is part of some Blender
-work, as using clang-cl provides a ~20-40% speedup compared to MSVC.
-
-Compiled with the following command line (via a VS2022 Native ARM64 Tools
-CMD window):
-  msbuild sse2neon.vcxproj /p:Configuration=Release /p:CLToolExe=clang-cl.exe
-                           /p:CLToolPath="C:\Program Files\LLVM\bin\"
-
-Known failures in test suite:
-  Test mm_cvttpd_epi32
-  Test rdtsc
-
-Co-authored-by: Anthony Roberts <anthony.roberts@linaro.org>
----
- sse2neon.h | 22 +++++++++++-----------
- 1 file changed, 11 insertions(+), 11 deletions(-)
-
-diff --git a/sse2neon.h b/sse2neon.h
-index 56254b5..76cf8e3 100644
---- a/sse2neon.h
-+++ b/sse2neon.h
-@@ -180,7 +180,7 @@
-     }
- 
- /* Compiler barrier */
--#if defined(_MSC_VER)
-+#if defined(_MSC_VER) && !defined(__clang__)
- #define SSE2NEON_BARRIER() _ReadWriteBarrier()
- #else
- #define SSE2NEON_BARRIER()                     \
-@@ -856,7 +856,7 @@ FORCE_INLINE uint64x2_t _sse2neon_vmull_p64(uint64x1_t _a, uint64x1_t _b)
- {
-     poly64_t a = vget_lane_p64(vreinterpret_p64_u64(_a), 0);
-     poly64_t b = vget_lane_p64(vreinterpret_p64_u64(_b), 0);
--#if defined(_MSC_VER)
-+#if defined(_MSC_VER) && !defined(__clang__)
-     __n64 a1 = {a}, b1 = {b};
-     return vreinterpretq_u64_p128(vmull_p64(a1, b1));
- #else
-@@ -1767,7 +1767,7 @@ FORCE_INLINE void _mm_free(void *addr)
- FORCE_INLINE uint64_t _sse2neon_get_fpcr(void)
- {
-     uint64_t value;
--#if defined(_MSC_VER)
-+#if defined(_MSC_VER) && !defined(__clang__)
-     value = _ReadStatusReg(ARM64_FPCR);
- #else
-     __asm__ __volatile__("mrs %0, FPCR" : "=r"(value)); /* read */
-@@ -1777,7 +1777,7 @@ FORCE_INLINE uint64_t _sse2neon_get_fpcr(void)
- 
- FORCE_INLINE void _sse2neon_set_fpcr(uint64_t value)
- {
--#if defined(_MSC_VER)
-+#if defined(_MSC_VER) && !defined(__clang__)
-     _WriteStatusReg(ARM64_FPCR, value);
- #else
-     __asm__ __volatile__("msr FPCR, %0" ::"r"(value));  /* write */
-@@ -2246,7 +2246,7 @@ FORCE_INLINE __m128 _mm_or_ps(__m128 a, __m128 b)
- FORCE_INLINE void _mm_prefetch(char const *p, int i)
- {
-     (void) i;
--#if defined(_MSC_VER)
-+#if defined(_MSC_VER) && !defined(__clang__)
-     switch (i) {
-     case _MM_HINT_NTA:
-         __prefetch2(p, 1);
-@@ -4817,7 +4817,7 @@ FORCE_INLINE __m128i _mm_packus_epi16(const __m128i a, const __m128i b)
- // https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm_pause
- FORCE_INLINE void _mm_pause(void)
- {
--#if defined(_MSC_VER)
-+#if defined(_MSC_VER) && !defined(__clang__)
-     __isb(_ARM64_BARRIER_SY);
- #else
-     __asm__ __volatile__("isb\n");
-@@ -5713,7 +5713,7 @@ FORCE_INLINE __m128d _mm_undefined_pd(void)
- #pragma GCC diagnostic ignored "-Wuninitialized"
- #endif
-     __m128d a;
--#if defined(_MSC_VER)
-+#if defined(_MSC_VER) && !defined(__clang__)
-     a = _mm_setzero_pd();
- #endif
-     return a;
-@@ -8127,7 +8127,7 @@ FORCE_INLINE int _sse2neon_sido_negative(int res, int lb, int imm8, int bound)
- 
- FORCE_INLINE int _sse2neon_clz(unsigned int x)
- {
--#ifdef _MSC_VER
-+#if defined(_MSC_VER) && !defined(__clang__)
-     unsigned long cnt = 0;
-     if (_BitScanReverse(&cnt, x))
-         return 31 - cnt;
-@@ -8139,7 +8139,7 @@ FORCE_INLINE int _sse2neon_clz(unsigned int x)
- 
- FORCE_INLINE int _sse2neon_ctz(unsigned int x)
- {
--#ifdef _MSC_VER
-+#if defined(_MSC_VER) && !defined(__clang__)
-     unsigned long cnt = 0;
-     if (_BitScanForward(&cnt, x))
-         return cnt;
-@@ -9055,7 +9055,7 @@ FORCE_INLINE __m128i _mm_aeskeygenassist_si128(__m128i a, const int rcon)
-     // AESE does ShiftRows and SubBytes on A
-     uint8x16_t u8 = vaeseq_u8(vreinterpretq_u8_m128i(a), vdupq_n_u8(0));
- 
--#ifndef _MSC_VER
-+#if !defined(_MSC_VER) || defined(__clang__)
-     uint8x16_t dest = {
-         // Undo ShiftRows step from AESE and extract X1 and X3
-         u8[0x4], u8[0x1], u8[0xE], u8[0xB],  // SubBytes(X1)
-@@ -9242,7 +9242,7 @@ FORCE_INLINE uint64_t _rdtsc(void)
-      * bits wide and it is attributed with the flag 'cap_user_time_short'
-      * is true.
-      */
--#if defined(_MSC_VER)
-+#if defined(_MSC_VER) && !defined(__clang__)
-     val = _ReadStatusReg(ARM64_SYSREG(3, 3, 14, 0, 2));
- #else
-     __asm__ __volatile__("mrs %0, cntvct_el0" : "=r"(val));
--- 
-2.48.1
-

CMakeLists.txt

@@ -613,9 +613,6 @@ find_package(VulkanUtilityLibraries)
 find_package(SimpleIni)
 find_package(SPIRV-Tools)
 find_package(sirit)
-if (ARCHITECTURE_arm64)
-    find_package(sse2neon)
-endif()
 
 if (ARCHITECTURE_x86 OR ARCHITECTURE_x86_64)
     find_package(xbyak)

externals/CMakeLists.txt

@@ -399,10 +399,3 @@ if (ANDROID)
 
     add_library(oboe::oboe ALIAS oboe)
 endif()
-
-# sse2neon
-if (ARCHITECTURE_arm64 AND NOT TARGET sse2neon)
-    AddJsonPackage(sse2neon)
-    add_library(sse2neon INTERFACE)
-    target_include_directories(sse2neon INTERFACE ${sse2neon_SOURCE_DIR})
-endif()

externals/cpmfile.json

@@ -213,14 +213,5 @@
         "key": "steamdeck",
         "bundled": true,
         "skip_updates": "true"
-    },
-    "sse2neon": {
-        "repo": "DLTcollab/sse2neon",
-        "sha": "66267b52fd",
-        "hash": "3aed8676e1b8c428acb076464663e3968a721457b08710a7c5f8df2fbdaa5601053c1606169a55e987e7a58dd17e3cc3b7fbf953aa891c5ac5f8ce2941862e4b",
-        "download_only": "true",
-        "patches": [
-            "0001-Add-support-for-clang-cl-on-Windows-633.patch"
-        ]
     }
 }

src/video_core/CMakeLists.txt

@@ -398,8 +398,4 @@ if (ANDROID AND ARCHITECTURE_arm64)
     target_link_libraries(video_core PRIVATE adrenotools)
 endif()
 
-if (ARCHITECTURE_arm64)
-    target_link_libraries(video_core PRIVATE sse2neon)
-endif()
-
 create_target_directory_groups(video_core)

src/video_core/host1x/vic.h

@@ -1,3 +1,6 @@
+// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
 // SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
@@ -610,24 +613,12 @@ public:
 
 private:
     void Execute();
-
     void Blend(const ConfigStruct& config, const SlotStruct& slot);
-
+    void ReadProgressiveY8__V8U8_N420(const SlotStruct& slot, std::span<const PlaneOffsets> offsets, std::shared_ptr<const FFmpeg::Frame> frame, bool planar, bool interlaced);
-    template <bool Planar, bool Interlaced = false>
+    void ReadInterlacedY8__V8U8_N420(const SlotStruct& slot, std::span<const PlaneOffsets> offsets, std::shared_ptr<const FFmpeg::Frame> frame, bool planar, bool top_field);
-    void ReadProgressiveY8__V8U8_N420(const SlotStruct& slot, std::span<const PlaneOffsets> offsets,
+    void ReadY8__V8U8_N420(const SlotStruct& slot, std::span<const PlaneOffsets> offsets, std::shared_ptr<const FFmpeg::Frame> frame, bool planar);
-                                      std::shared_ptr<const FFmpeg::Frame> frame);
-    template <bool Planar, bool TopField>
-    void ReadInterlacedY8__V8U8_N420(const SlotStruct& slot, std::span<const PlaneOffsets> offsets,
-                                     std::shared_ptr<const FFmpeg::Frame> frame);
-
-    template <bool Planar>
-    void ReadY8__V8U8_N420(const SlotStruct& slot, std::span<const PlaneOffsets> offsets,
-                           std::shared_ptr<const FFmpeg::Frame> frame);
-
     void WriteY8__V8U8_N420(const OutputSurfaceConfig& output_surface_config);
-
+    void WriteABGR(const OutputSurfaceConfig& output_surface_config, VideoPixelFormat format);
-    template <VideoPixelFormat Format>
-    void WriteABGR(const OutputSurfaceConfig& output_surface_config);
 
     s32 id;
     s32 nvdec_id{-1};
@@ -636,8 +627,6 @@ private:
     VicRegisters regs{};
     FrameQueue& frame_queue;
 
-    const bool has_sse41{false};
-
     Common::ScratchBuffer<Pixel> output_surface;
     Common::ScratchBuffer<Pixel> slot_surface;
     Common::ScratchBuffer<u8> luma_scratch;