Drop SSE2 backend

Author: sergcpp

.gitea/workflows/release.yml

@@ -580,30 +580,30 @@ jobs:
   #       with:
   #         name: macos-x86_64-cpu-output
   #         path: macos-x86_64-cpu-output/
-  test-linux-x86_64-cpu-sse2:
-    runs-on: [ linux, x86_64 ]
-    needs:
-      - build-linux-x86_64-rel
-    steps:
-      - name: Checkout Workflows
-        uses: actions/checkout@v4
-        with:
-          sparse-checkout: .gitea
-      - name: Re-initialize repository
-        uses: ./.gitea/actions/checkout
-      - name: Run Tests
-        uses: ./.gitea/actions/test-sde
-        with:
-          bin-dir: "linux-x86_64"
-          out-dir: "linux-x86_64-cpu-sse2-output"
-          test-args: "--arch SSE2 --nogpu -j4"
-          sde-args: "-p4p"
-      - name: Upload artifacts
-        uses: actions/upload-artifact@v3
-        if: always()
-        with:
-          name: linux-x86_64-cpu-sse2-output
-          path: linux-x86_64-cpu-sse2-output/
+  # test-linux-x86_64-cpu-sse2:
+  #   runs-on: [ linux, x86_64 ]
+  #   needs:
+  #     - build-linux-x86_64-rel
+  #   steps:
+  #     - name: Checkout Workflows
+  #       uses: actions/checkout@v4
+  #       with:
+  #         sparse-checkout: .gitea
+  #     - name: Re-initialize repository
+  #       uses: ./.gitea/actions/checkout
+  #     - name: Run Tests
+  #       uses: ./.gitea/actions/test-sde
+  #       with:
+  #         bin-dir: "linux-x86_64"
+  #         out-dir: "linux-x86_64-cpu-sse2-output"
+  #         test-args: "--arch SSE2 --nogpu -j4"
+  #         sde-args: "-p4p"
+  #     - name: Upload artifacts
+  #       uses: actions/upload-artifact@v3
+  #       if: always()
+  #       with:
+  #         name: linux-x86_64-cpu-sse2-output
+  #         path: linux-x86_64-cpu-sse2-output/
   test-linux-x86_64-cpu-sse41:
     runs-on: [ linux, x86_64 ]
     needs:
@@ -820,7 +820,7 @@ jobs:
   #         path: windows-arm64-gpu-perf-output/
   analyze-output:
     runs-on: [ linux ]
-    needs: [ test-windows-x86_64-cpu, test-windows-x86_64-gpu-amd, test-windows-x86_64-gpu-uhd, test-windows-x86_64-gpu-arc, test-linux-x86_64-cpu, test-linux-x86_64-gpu-nv, test-linux-x86_64-cpu-sse2, test-linux-x86_64-cpu-sse41, test-windows-x86_64-cpu-avx, test-windows-x86_64-cpu-avx2, test-windows-x86_64-cpu-avx512, test-linux-x86_64-cpu-tsan, test-windows-x86_64-cpu-asan, test-windows-x86_64-gpu-amd-asan ]
+    needs: [ test-windows-x86_64-cpu, test-windows-x86_64-gpu-amd, test-windows-x86_64-gpu-uhd, test-windows-x86_64-gpu-arc, test-linux-x86_64-cpu, test-linux-x86_64-gpu-nv, test-linux-x86_64-cpu-sse41, test-windows-x86_64-cpu-avx, test-windows-x86_64-cpu-avx2, test-windows-x86_64-cpu-avx512, test-linux-x86_64-cpu-tsan, test-windows-x86_64-cpu-asan, test-windows-x86_64-gpu-amd-asan ]
     if: always()
     steps:
       - name: Checkout Workflows
@@ -879,10 +879,10 @@ jobs:
       #   with:
       #     name: macos-x86_64-cpu-output
       #     path: macos-x86_64-cpu-output/
-      - uses: actions/download-artifact@v3
-        with:
-          name: linux-x86_64-cpu-sse2-output
-          path: linux-x86_64-cpu-sse2-output/
+      # - uses: actions/download-artifact@v3
+      #   with:
+      #     name: linux-x86_64-cpu-sse2-output
+      #     path: linux-x86_64-cpu-sse2-output/
       - uses: actions/download-artifact@v3
         with:
           name: linux-x86_64-cpu-sse41-output
@@ -921,5 +921,5 @@ jobs:
       #     path: windows-arm64-gpu-perf-output/
       - name: Analyze Output
         run: |
-          python scripts/analyze_output.py windows-x86_64-cpu-output/test_Ray_output.txt windows-x86_64-gpu-amd-output/test_Ray_output.txt windows-x86_64-gpu-uhd-output/test_Ray_output.txt windows-x86_64-gpu-arc-output/test_Ray_output.txt linux-x86_64-cpu-output/test_Ray_output.txt linux-x86_64-gpu-nv-output/test_Ray_output.txt linux-x86_64-cpu-sse2-output/test_Ray_output.txt linux-x86_64-cpu-sse41-output/test_Ray_output.txt windows-x86_64-cpu-avx-output/test_Ray_output.txt windows-x86_64-cpu-avx2-output/test_Ray_output.txt windows-x86_64-cpu-avx512-output/test_Ray_output.txt linux-x86_64-cpu-tsan-output/test_Ray_output.txt windows-x86_64-cpu-asan-output/test_Ray_output.txt windows-x86_64-gpu-amd-asan-output/test_Ray_output.txt
+          python scripts/analyze_output.py windows-x86_64-cpu-output/test_Ray_output.txt windows-x86_64-gpu-amd-output/test_Ray_output.txt windows-x86_64-gpu-uhd-output/test_Ray_output.txt windows-x86_64-gpu-arc-output/test_Ray_output.txt linux-x86_64-cpu-output/test_Ray_output.txt linux-x86_64-gpu-nv-output/test_Ray_output.txt linux-x86_64-cpu-sse41-output/test_Ray_output.txt windows-x86_64-cpu-avx-output/test_Ray_output.txt windows-x86_64-cpu-avx2-output/test_Ray_output.txt windows-x86_64-cpu-avx512-output/test_Ray_output.txt linux-x86_64-cpu-tsan-output/test_Ray_output.txt windows-x86_64-cpu-asan-output/test_Ray_output.txt windows-x86_64-gpu-amd-asan-output/test_Ray_output.txt
         shell: bash

CMakeLists.txt

@@ -134,8 +134,6 @@ endif()
 
 if (ENABLE_SIMD_IMPL)
     set(INTERNAL_SOURCE_FILES ${INTERNAL_SOURCE_FILES}
-                              internal/RendererSSE2.h
-                              internal/RendererSSE2.cpp
                               internal/RendererSSE41.h
                               internal/RendererSSE41.cpp
                               internal/RendererNEON.h
@@ -148,7 +146,6 @@ if (ENABLE_SIMD_IMPL)
                               internal/RendererAVX512.cpp)
 
     set_source_files_properties(
-        internal/RendererSSE2.cpp
         internal/RendererSSE41.cpp
         internal/RendererNEON.cpp
         internal/RendererAVX.cpp
@@ -184,7 +181,6 @@ endif(ENABLE_DX_IMPL AND WIN32)
 if(NOT CMAKE_GENERATOR_PLATFORM MATCHES "ARM64")
     if(MSVC)
         if(NOT CMAKE_CL_64)
-            set_source_files_properties(internal/RendererSSE2.cpp PROPERTIES COMPILE_FLAGS /arch:SSE2)
             set_source_files_properties(internal/TextureUtilsSSE2.cpp PROPERTIES COMPILE_FLAGS /arch:SSE2)
             set_source_files_properties(internal/RendererSSE41.cpp PROPERTIES COMPILE_FLAGS /arch:SSE2)
         endif()
@@ -198,7 +194,6 @@ if(NOT CMAKE_GENERATOR_PLATFORM MATCHES "ARM64")
     else()
         set_source_files_properties(internal/simd/detect.cpp PROPERTIES COMPILE_FLAGS -mxsave)
         set_source_files_properties(internal/TextureUtilsSSE2.cpp PROPERTIES COMPILE_FLAGS -msse2)
-        set_source_files_properties(internal/RendererSSE2.cpp PROPERTIES COMPILE_FLAGS -msse2)
         set_source_files_properties(internal/RendererSSE41.cpp PROPERTIES COMPILE_FLAGS -msse4.1)
         set_source_files_properties(internal/RendererAVX.cpp PROPERTIES COMPILE_FLAGS -mavx)
         set_source_files_properties(internal/RendererAVX2.cpp PROPERTIES COMPILE_FLAGS "-mavx2 -mfma")

Ray.cpp

@@ -15,12 +15,11 @@
 #include "internal/RendererAVX.h"
 #include "internal/RendererAVX2.h"
 #include "internal/RendererAVX512.h"
-#include "internal/RendererSSE2.h"
 #include "internal/RendererSSE41.h"
 #elif defined(__arm__) || defined(__aarch64__) || defined(_M_ARM) || defined(_M_ARM64)
 #include "internal/RendererNEON.h"
 #elif defined(__i386__) || defined(__x86_64__)
-#include "internal/RendererSSE2.h"
+#include "internal/RendererSSE41.h"
 #endif
 #else // ENABLE_SIMD_IMPL
 #pragma message("Ray: Compiling without SIMD support")
@@ -93,10 +92,6 @@ Ray::RendererBase *Ray::CreateRenderer(const settings_t &s, ILog *log,
         log->Info("Ray: Creating SSE41 renderer %ix%i", s.w, s.h);
         return Sse41::CreateRenderer(s, log);
     }
-    if ((enabled_types & eRendererType::SIMD_SSE2) && features.sse2_supported) {
-        log->Info("Ray: Creating SSE2 renderer %ix%i", s.w, s.h);
-        return Sse2::CreateRenderer(s, log);
-    }
 #endif
 #ifdef ENABLE_REF_IMPL
     if (enabled_types & eRendererType::Reference) {

Ray.h

@@ -16,7 +16,7 @@ extern LogStdout g_stdout_log;
 
 /// Default renderer flags used to choose backend, by default tries to create gpu renderer first
 const Bitmask<eRendererType> DefaultEnabledRenderTypes =
-    Bitmask<eRendererType>{eRendererType::Reference} | eRendererType::SIMD_SSE2 | eRendererType::SIMD_AVX |
+    Bitmask<eRendererType>{eRendererType::Reference} | eRendererType::SIMD_SSE41 | eRendererType::SIMD_AVX |
     eRendererType::SIMD_AVX2 | eRendererType::SIMD_NEON | eRendererType::Vulkan | eRendererType::DirectX12;
 
 /** @brief Creates renderer

RendererBase.cpp

@@ -7,8 +7,6 @@ std::string_view RendererTypeName(const eRendererType rt) {
     switch (rt) {
     case eRendererType::Reference:
         return "REF";
-    case eRendererType::SIMD_SSE2:
-        return "SSE2";
     case eRendererType::SIMD_SSE41:
         return "SSE41";
     case eRendererType::SIMD_AVX:
@@ -31,8 +29,6 @@ std::string_view RendererTypeName(const eRendererType rt) {
 eRendererType RendererTypeFromName(std::string_view name) {
     if (name == "REF") {
         return eRendererType::Reference;
-    } else if (name == "SSE2") {
-        return eRendererType::SIMD_SSE2;
     } else if (name == "SSE41") {
         return eRendererType::SIMD_SSE41;
     } else if (name == "AVX") {
@@ -54,7 +50,6 @@ eRendererType RendererTypeFromName(std::string_view name) {
 bool RendererSupportsMultithreading(const eRendererType rt) {
     switch (rt) {
     case eRendererType::Reference:
-    case eRendererType::SIMD_SSE2:
     case eRendererType::SIMD_SSE41:
     case eRendererType::SIMD_AVX:
     case eRendererType::SIMD_AVX2:

RendererBase.h

@@ -23,7 +23,6 @@ enum class eRendererType : uint32_t {
     // Reference CPU renderer, slightly vectorized, the easiest to modify and debug
     Reference,
     // SIMD CPU renderers, heavily vectorized in SPMD fashion
-    SIMD_SSE2,
     SIMD_SSE41,
     SIMD_AVX,
     SIMD_AVX2,
@@ -35,7 +34,7 @@ enum class eRendererType : uint32_t {
 };
 
 // All CPU renderers
-const Bitmask<eRendererType> RendererCPU = Bitmask<eRendererType>{eRendererType::Reference} | eRendererType::SIMD_SSE2 |
+const Bitmask<eRendererType> RendererCPU = Bitmask<eRendererType>{eRendererType::Reference} |
                                            eRendererType::SIMD_SSE41 | eRendererType::SIMD_NEON |
                                            eRendererType::SIMD_AVX | eRendererType::SIMD_AVX2;
 // All GPU renderers

internal/RendererSSE2.cpp

@@ -121,7 +121,7 @@ bool g_stop_on_fail = false;
 std::atomic_bool g_tests_success{true};
 std::atomic_bool g_log_contains_errors{false};
 bool g_catch_flt_exceptions = false;
-bool g_determine_sample_count = false;
+bool g_determine_sample_count = true;
 bool g_minimal_output = false;
 bool g_nohwrt = false;
 bool g_nodx = false;
@@ -227,8 +227,8 @@ int main(int argc, char *argv[]) {
     }
 #endif
 
-    static const char *ArchListFull[] = {"REF", "SSE2", "SSE41", "AVX", "AVX2", "AVX512", "NEON", "VK", "DX", nullptr};
-    static const char *ArchListFullNoGPU[] = {"REF", "SSE2", "SSE41", "AVX", "AVX2", "AVX512", "NEON", nullptr};
+    static const char *ArchListFull[] = {"REF", "SSE41", "AVX", "AVX2", "AVX512", "NEON", "VK", "DX", nullptr};
+    static const char *ArchListFullNoGPU[] = {"REF", "SSE41", "AVX", "AVX2", "AVX512", "NEON", nullptr};
     static const char *ArchListDefault[] = {"AVX2", "NEON", "VK", "DX", nullptr};
 #ifndef __APPLE__
     static const char *ArchListDefaultNoGPU[] = {"AVX2", "NEON", nullptr};