LibSoftGPU: Use `FloatVector4` pixel format for `Image`

This increases memory usage for textures 4-fold, but allows us to
completely remove calls to `(un)pack_color` in a very hot loop.

Userland/Libraries/LibSoftGPU/Image.cpp

@@ -9,10 +9,123 @@
 
 namespace SoftGPU {
 
+static constexpr FloatVector4 unpack_color(void const* ptr, GPU::ImageFormat format)
+{
+    constexpr auto one_over_255 = 1.0f / 255;
+    switch (format) {
+    case GPU::ImageFormat::RGB888: {
+        auto rgb = reinterpret_cast<u8 const*>(ptr);
+        return {
+            rgb[0] * one_over_255,
+            rgb[1] * one_over_255,
+            rgb[2] * one_over_255,
+            1.0f,
+        };
+    }
+    case GPU::ImageFormat::BGR888: {
+        auto bgr = reinterpret_cast<u8 const*>(ptr);
+        return {
+            bgr[2] * one_over_255,
+            bgr[1] * one_over_255,
+            bgr[0] * one_over_255,
+            1.0f,
+        };
+    }
+    case GPU::ImageFormat::RGBA8888: {
+        auto rgba = *reinterpret_cast<u32 const*>(ptr);
+        return {
+            (rgba & 0xff) * one_over_255,
+            ((rgba >> 8) & 0xff) * one_over_255,
+            ((rgba >> 16) & 0xff) * one_over_255,
+            ((rgba >> 24) & 0xff) * one_over_255,
+        };
+    }
+    case GPU::ImageFormat::BGRA8888: {
+        auto bgra = *reinterpret_cast<u32 const*>(ptr);
+        return {
+            ((bgra >> 16) & 0xff) * one_over_255,
+            ((bgra >> 8) & 0xff) * one_over_255,
+            (bgra & 0xff) * one_over_255,
+            ((bgra >> 24) & 0xff) * one_over_255,
+        };
+    }
+    case GPU::ImageFormat::RGB565: {
+        auto rgb = *reinterpret_cast<u16 const*>(ptr);
+        return {
+            ((rgb >> 11) & 0x1f) / 31.f,
+            ((rgb >> 5) & 0x3f) / 63.f,
+            (rgb & 0x1f) / 31.f,
+            1.0f
+        };
+    }
+    case GPU::ImageFormat::L8: {
+        auto luminance = *reinterpret_cast<u8 const*>(ptr);
+        auto clamped_luminance = luminance * one_over_255;
+        return {
+            clamped_luminance,
+            clamped_luminance,
+            clamped_luminance,
+            1.0f,
+        };
+    }
+    case GPU::ImageFormat::L8A8: {
+        auto luminance_and_alpha = reinterpret_cast<u8 const*>(ptr);
+        auto clamped_luminance = luminance_and_alpha[0] * one_over_255;
+        return {
+            clamped_luminance,
+            clamped_luminance,
+            clamped_luminance,
+            luminance_and_alpha[1] * one_over_255,
+        };
+    }
+    default:
+        VERIFY_NOT_REACHED();
+    }
+}
+
+static constexpr void pack_color(FloatVector4 const& color, void* ptr, GPU::ImageFormat format)
+{
+    auto r = static_cast<u8>(clamp(color.x(), 0.0f, 1.0f) * 255);
+    auto g = static_cast<u8>(clamp(color.y(), 0.0f, 1.0f) * 255);
+    auto b = static_cast<u8>(clamp(color.z(), 0.0f, 1.0f) * 255);
+    auto a = static_cast<u8>(clamp(color.w(), 0.0f, 1.0f) * 255);
+
+    switch (format) {
+    case GPU::ImageFormat::RGB888:
+        reinterpret_cast<u8*>(ptr)[0] = r;
+        reinterpret_cast<u8*>(ptr)[1] = g;
+        reinterpret_cast<u8*>(ptr)[2] = b;
+        return;
+    case GPU::ImageFormat::BGR888:
+        reinterpret_cast<u8*>(ptr)[2] = b;
+        reinterpret_cast<u8*>(ptr)[1] = g;
+        reinterpret_cast<u8*>(ptr)[0] = r;
+        return;
+    case GPU::ImageFormat::RGBA8888:
+        *reinterpret_cast<u32*>(ptr) = r | (g << 8) | (b << 16) | (a << 24);
+        return;
+    case GPU::ImageFormat::BGRA8888:
+        *reinterpret_cast<u32*>(ptr) = b | (g << 8) | (r << 16) | (a << 24);
+        return;
+    case GPU::ImageFormat::RGB565:
+        *reinterpret_cast<u16*>(ptr) = (r & 0x1f) | ((g & 0x3f) << 5) | ((b & 0x1f) << 11);
+        return;
+    case GPU::ImageFormat::L8:
+        *reinterpret_cast<u8*>(ptr) = r;
+        return;
+    case GPU::ImageFormat::L8A8:
+        reinterpret_cast<u8*>(ptr)[0] = r;
+        reinterpret_cast<u8*>(ptr)[1] = a;
+        return;
+    default:
+        VERIFY_NOT_REACHED();
+    }
+}
+
 Image::Image(void* const ownership_token, unsigned width, unsigned height, unsigned depth, unsigned max_levels, unsigned layers)
     : GPU::Image(ownership_token)
     , m_num_layers(layers)
-    , m_mipmap_buffers(FixedArray<RefPtr<Typed3DBuffer<GPU::ColorType>>>::must_create_but_fixme_should_propagate_errors(layers * max_levels))
+    , m_mipmap_buffers(FixedArray<RefPtr<Typed3DBuffer<FloatVector4>>>::must_create_but_fixme_should_propagate_errors(layers * max_levels))
 {
     VERIFY(width > 0);
     VERIFY(height > 0);
@@ -27,7 +140,7 @@ Image::Image(void* const ownership_token, unsigned width, unsigned height, unsig
     unsigned level;
     for (level = 0; level < max_levels; ++level) {
         for (unsigned layer = 0; layer < layers; ++layer)
-            m_mipmap_buffers[layer * layers + level] = MUST(Typed3DBuffer<GPU::ColorType>::try_create(width, height, depth));
+            m_mipmap_buffers[layer * layers + level] = MUST(Typed3DBuffer<FloatVector4>::try_create(width, height, depth));
 
         if (width <= 1 && height <= 1 && depth <= 1)
             break;

Userland/Libraries/LibSoftGPU/Image.h

@@ -8,132 +8,15 @@
 #pragma once
 
 #include <AK/FixedArray.h>
-#include <AK/RefCounted.h>
 #include <AK/RefPtr.h>
-#include <LibGPU/Enums.h>
 #include <LibGPU/Image.h>
 #include <LibGPU/ImageDataLayout.h>
-#include <LibGPU/ImageFormat.h>
 #include <LibGfx/Vector3.h>
 #include <LibGfx/Vector4.h>
 #include <LibSoftGPU/Buffer/Typed3DBuffer.h>
-#include <LibSoftGPU/Config.h>
 
 namespace SoftGPU {
 
-inline static FloatVector4 unpack_color(void const* ptr, GPU::ImageFormat format)
-{
-    constexpr auto one_over_255 = 1.0f / 255;
-    switch (format) {
-    case GPU::ImageFormat::RGB888: {
-        auto rgb = reinterpret_cast<u8 const*>(ptr);
-        return {
-            rgb[0] * one_over_255,
-            rgb[1] * one_over_255,
-            rgb[2] * one_over_255,
-            1.0f,
-        };
-    }
-    case GPU::ImageFormat::BGR888: {
-        auto bgr = reinterpret_cast<u8 const*>(ptr);
-        return {
-            bgr[2] * one_over_255,
-            bgr[1] * one_over_255,
-            bgr[0] * one_over_255,
-            1.0f,
-        };
-    }
-    case GPU::ImageFormat::RGBA8888: {
-        auto rgba = *reinterpret_cast<u32 const*>(ptr);
-        return {
-            (rgba & 0xff) * one_over_255,
-            ((rgba >> 8) & 0xff) * one_over_255,
-            ((rgba >> 16) & 0xff) * one_over_255,
-            ((rgba >> 24) & 0xff) * one_over_255,
-        };
-    }
-    case GPU::ImageFormat::BGRA8888: {
-        auto bgra = *reinterpret_cast<u32 const*>(ptr);
-        return {
-            ((bgra >> 16) & 0xff) * one_over_255,
-            ((bgra >> 8) & 0xff) * one_over_255,
-            (bgra & 0xff) * one_over_255,
-            ((bgra >> 24) & 0xff) * one_over_255,
-        };
-    }
-    case GPU::ImageFormat::RGB565: {
-        auto rgb = *reinterpret_cast<u16 const*>(ptr);
-        return {
-            ((rgb >> 11) & 0x1f) / 31.f,
-            ((rgb >> 5) & 0x3f) / 63.f,
-            (rgb & 0x1f) / 31.f,
-            1.0f
-        };
-    }
-    case GPU::ImageFormat::L8: {
-        auto luminance = *reinterpret_cast<u8 const*>(ptr);
-        auto clamped_luminance = luminance * one_over_255;
-        return {
-            clamped_luminance,
-            clamped_luminance,
-            clamped_luminance,
-            1.0f,
-        };
-    }
-    case GPU::ImageFormat::L8A8: {
-        auto luminance_and_alpha = reinterpret_cast<u8 const*>(ptr);
-        auto clamped_luminance = luminance_and_alpha[0] * one_over_255;
-        return {
-            clamped_luminance,
-            clamped_luminance,
-            clamped_luminance,
-            luminance_and_alpha[1] * one_over_255,
-        };
-    }
-    default:
-        VERIFY_NOT_REACHED();
-    }
-}
-
-inline static void pack_color(FloatVector4 const& color, void* ptr, GPU::ImageFormat format)
-{
-    auto r = static_cast<u8>(clamp(color.x(), 0.0f, 1.0f) * 255);
-    auto g = static_cast<u8>(clamp(color.y(), 0.0f, 1.0f) * 255);
-    auto b = static_cast<u8>(clamp(color.z(), 0.0f, 1.0f) * 255);
-    auto a = static_cast<u8>(clamp(color.w(), 0.0f, 1.0f) * 255);
-
-    switch (format) {
-    case GPU::ImageFormat::RGB888:
-        reinterpret_cast<u8*>(ptr)[0] = r;
-        reinterpret_cast<u8*>(ptr)[1] = g;
-        reinterpret_cast<u8*>(ptr)[2] = b;
-        return;
-    case GPU::ImageFormat::BGR888:
-        reinterpret_cast<u8*>(ptr)[2] = b;
-        reinterpret_cast<u8*>(ptr)[1] = g;
-        reinterpret_cast<u8*>(ptr)[0] = r;
-        return;
-    case GPU::ImageFormat::RGBA8888:
-        *reinterpret_cast<u32*>(ptr) = r | (g << 8) | (b << 16) | (a << 24);
-        return;
-    case GPU::ImageFormat::BGRA8888:
-        *reinterpret_cast<u32*>(ptr) = b | (g << 8) | (r << 16) | (a << 24);
-        return;
-    case GPU::ImageFormat::RGB565:
-        *reinterpret_cast<u16*>(ptr) = (r & 0x1f) | ((g & 0x3f) << 5) | ((b & 0x1f) << 11);
-        return;
-    case GPU::ImageFormat::L8:
-        *reinterpret_cast<u8*>(ptr) = r;
-        return;
-    case GPU::ImageFormat::L8A8:
-        reinterpret_cast<u8*>(ptr)[0] = r;
-        reinterpret_cast<u8*>(ptr)[1] = a;
-        return;
-    default:
-        VERIFY_NOT_REACHED();
-    }
-}
-
 class Image final : public GPU::Image {
 public:
     Image(void* const ownership_token, unsigned width, unsigned height, unsigned depth, unsigned max_levels, unsigned layers);
@@ -149,12 +32,12 @@ public:
 
     FloatVector4 texel(unsigned layer, unsigned level, int x, int y, int z) const
     {
-        return unpack_color(texel_pointer(layer, level, x, y, z), GPU::ImageFormat::BGRA8888);
+        return *texel_pointer(layer, level, x, y, z);
     }
 
     void set_texel(unsigned layer, unsigned level, int x, int y, int z, FloatVector4 const& color)
     {
-        pack_color(color, texel_pointer(layer, level, x, y, z), GPU::ImageFormat::BGRA8888);
+        *texel_pointer(layer, level, x, y, z) = color;
     }
 
     virtual void write_texels(unsigned layer, unsigned level, Vector3<unsigned> const& offset, Vector3<unsigned> const& size, void const* data, GPU::ImageDataLayout const& layout) override;
@@ -162,12 +45,12 @@ public:
     virtual void copy_texels(GPU::Image const& source, unsigned source_layer, unsigned source_level, Vector3<unsigned> const& source_offset, Vector3<unsigned> const& size, unsigned destination_layer, unsigned destination_level, Vector3<unsigned> const& destination_offset) override;
 
 private:
-    void const* texel_pointer(unsigned layer, unsigned level, int x, int y, int z) const
+    FloatVector4 const* texel_pointer(unsigned layer, unsigned level, int x, int y, int z) const
     {
         return m_mipmap_buffers[layer * m_num_layers + level]->buffer_pointer(x, y, z);
     }
 
-    void* texel_pointer(unsigned layer, unsigned level, int x, int y, int z)
+    FloatVector4* texel_pointer(unsigned layer, unsigned level, int x, int y, int z)
     {
         return m_mipmap_buffers[layer * m_num_layers + level]->buffer_pointer(x, y, z);
     }
@@ -176,7 +59,7 @@ private:
     unsigned m_num_levels { 0 };
     unsigned m_num_layers { 0 };
 
-    FixedArray<RefPtr<Typed3DBuffer<GPU::ColorType>>> m_mipmap_buffers;
+    FixedArray<RefPtr<Typed3DBuffer<FloatVector4>>> m_mipmap_buffers;
 
     bool m_width_is_power_of_two { false };
     bool m_height_is_power_of_two { false };