LibGL: Implement Texture State Management

Some very primitive Texture State management. Data can now be
uploaded to textures.

Userland/Libraries/LibGL/Tex/Texture.cpp

@@ -0,0 +1,135 @@
+/*
+ * Copyright (c) 2021, Jesse Buhagiar <jooster669@gmail.com>
+ *
+ * SPDX-License-Identifier: BSD-2-Clause
+ */
+
+#include <AK/Format.h>
+#include <LibGL/GL/gl.h>
+#include <LibGL/Tex/Texture.h>
+#include <string.h>
+
+namespace GL {
+
+void Texture::upload_texture_data(GLenum, GLint lod, GLint internal_format, GLsizei width, GLsizei height, GLint, GLenum format, GLenum, const GLvoid* pixels)
+{
+    // NOTE: Some target, format, and internal formats are currently unsupported.
+    // Considering we control this library, and `gl.h` itself, we don't need to add any
+    // checks here to see if we support them; the program will simply fail to compile..
+
+    // Somebody passed us in nullptr...
+    // Apparently this allocates memory on the GPU (according to Khronos docs..)?
+    if (pixels == nullptr) {
+        dbgln("LibGL: pixels == nullptr when uploading texture data.");
+        VERIFY_NOT_REACHED();
+    }
+
+    m_internal_format = internal_format;
+
+    // Get reference to the mip
+    auto& mip = m_mipmaps[lod];
+    const u8* pixel_byte_array = reinterpret_cast<const u8*>(pixels);
+
+    // Copy pixel data to storage
+
+    // Pixels are already 32-bits wide
+    if (format == GL_RGBA || format == GL_BGRA) {
+        mip.pixel_data().resize(width * height * sizeof(u32));
+        memcpy(mip.pixel_data().data(), pixels, width * height * sizeof(u32));
+    } else {
+        mip.pixel_data().resize(width * height * 3);
+        // Copy RGB or BGR pixel data
+        for (auto i = 0; i < width * height * 3; i += 3) {
+            u32 b0 = pixel_byte_array[i];     // B or R
+            u32 b1 = pixel_byte_array[i + 1]; // G
+            u32 b2 = pixel_byte_array[i + 2]; // R or B
+
+            u32 pixel = ((0xffu << 24) | (b0 << 16) | (b1 << 8) | b2);
+            mip.pixel_data().append(pixel);
+        }
+    }
+
+    // Now we need to swizzle the texture data from `format` to `internal_format`
+    switch (format) {
+    case GL_BGR: {
+        if (internal_format == GL_RGB) {
+            swizzle(mip.pixel_data(), [](u32 pixel) -> u32 {
+                u8 r = pixel & 0xff;
+                u8 g = (pixel >> 8) & 0xff;
+                u8 b = (pixel >> 16) & 0xff;
+
+                return (0xff << 24) | (r << 16) | (g << 8) | b;
+            });
+        } else if (internal_format == GL_RGBA) {
+            swizzle(mip.pixel_data(), [](u32 pixel) -> u32 {
+                u8 r = pixel & 0xff;
+                u8 g = (pixel >> 8) & 0xff;
+                u8 b = (pixel >> 16) & 0xff;
+
+                return (r << 24) | (g << 16) | (b << 8) | 0xff;
+            });
+        }
+    } break;
+    case GL_BGRA: {
+        if (internal_format == GL_RGB) {
+            swizzle(mip.pixel_data(), [](u32 pixel) -> u32 {
+                u8 r = (pixel >> 8) & 0xff;
+                u8 g = (pixel >> 16) & 0xff;
+                u8 b = (pixel >> 24) & 0xff;
+
+                return (0xff << 24) | (r << 16) | (g << 8) | b;
+            });
+        } else if (internal_format == GL_RGBA) {
+            swizzle(mip.pixel_data(), [](u32 pixel) -> u32 {
+                u8 a = pixel & 0xff;
+                u8 r = (pixel >> 8) & 0xff;
+                u8 g = (pixel >> 16) & 0xff;
+                u8 b = (pixel >> 24) & 0xff;
+
+                return (r << 24) | (g << 16) | (b << 8) | a;
+            });
+        }
+    } break;
+    case GL_RGB: {
+        if (internal_format == GL_RGBA) {
+            swizzle(mip.pixel_data(), [](u32 pixel) -> u32 {
+                u8 r = pixel & 0xff;
+                u8 g = (pixel >> 8) & 0xff;
+                u8 b = (pixel >> 16) & 0xff;
+
+                return (r << 24) | (g << 16) | (b << 8) | 0xff;
+            });
+        }
+    } break;
+    case GL_RGBA:
+        break;
+    default:
+        // Let's crash for now so we can implement format by format
+        VERIFY_NOT_REACHED();
+    }
+
+    mip.set_width(width);
+    mip.set_height(height);
+}
+
+FloatVector4 Texture::sample_texel(const FloatVector2& uv) const
+{
+    auto& mip = m_mipmaps.at(0);
+
+    // FIXME: Remove this to prevent a crash when we have proper texture binding
+    if (mip.width() == 0 || mip.height() == 0)
+        return { 1.0f, 1.0f, 1.0f, 1.0f };
+
+    u32 u = static_cast<u32>(uv.x() * mip.width());
+    u32 v = static_cast<u32>(uv.y() * mip.height());
+
+    u32 pixel = mip.pixel_data().at(v * mip.width() + u);
+
+    float b0 = ((pixel)&0xff) / 255.0f;
+    float b1 = ((pixel >> 8) & 0xff) / 255.0f;
+    float b2 = ((pixel >> 16) & 0xff) / 255.0f;
+
+    return { b0, b1, b2, 1.0f };
+}
+
+}

Userland/Libraries/LibGL/Tex/Texture.h

@@ -0,0 +1,78 @@
+/*
+ * Copyright (c) 2021, Jesse Buhagiar <jooster669@gmail.com>
+ *
+ * SPDX-License-Identifier: BSD-2-Clause
+ */
+
+#pragma once
+
+#include <AK/Array.h>
+#include <AK/RefCounted.h>
+#include <AK/Vector.h>
+#include <LibGL/GL/gl.h>
+#include <LibGfx/Vector2.h>
+#include <LibGfx/Vector4.h>
+
+namespace GL {
+
+class Texture final : public RefCounted<Texture> {
+public:
+    // FIXME: These shouldn't really belong here, they're context specific.
+    static constexpr u16 MAX_TEXTURE_SIZE = 2048;
+    static constexpr u8 LOG2_MAX_TEXTURE_SIZE = 11;
+
+    class MipMap {
+    public:
+        MipMap() = default;
+        ~MipMap() = default;
+
+        void set_width(GLsizei width) { m_width = width; }
+        void set_height(GLsizei height) { m_height = height; }
+        GLsizei width() const { return m_width; }
+        GLsizei height() const { return m_height; }
+
+        Vector<u32>& pixel_data() { return m_pixel_data; }
+        const Vector<u32>& pixel_data() const { return m_pixel_data; }
+
+    private:
+        GLsizei m_width;
+        GLsizei m_height;
+        Vector<u32> m_pixel_data;
+    };
+
+    // To quote the Khronos documentation:
+    // "You could say that a texture object contains a sampler object, which you access through the texture interface."
+    // FIXME: Better name?
+    struct TextureSamplerParamaters {
+        GLint m_min_filter { GL_NEAREST_MIPMAP_LINEAR };
+        GLint m_mag_filter { GL_LINEAR };
+        GLint m_wrap_s_mode { GL_REPEAT };
+        GLint m_wrap_t_mode { GL_REPEAT };
+    };
+
+public:
+    Texture() = default;
+    ~Texture() { }
+
+    void upload_texture_data(GLenum target, GLint lod, GLint internal_format, GLsizei width, GLsizei height, GLint border, GLenum format, GLenum type, const GLvoid* pixels);
+    void replace_sub_texture_data(GLint lod, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, GLenum format, GLenum type, const GLvoid* data);
+    FloatVector4 sample_texel(const FloatVector2& uv) const;
+
+    GLenum internal_format() const { return m_internal_format; }
+
+private:
+    template<typename TCallback>
+    void swizzle(Vector<u32>& pixels, TCallback&& callback)
+    {
+        for (auto& pixel : pixels)
+            pixel = callback(pixel);
+    }
+
+private:
+    // FIXME: Mipmaps are currently unused, but we have the plumbing for it at least
+    Array<MipMap, LOG2_MAX_TEXTURE_SIZE> m_mipmaps;
+    GLenum m_internal_format;
+    TextureSamplerParamaters m_sampler_params;
+};
+
+}