From 9cdf5c6c3126d022cea24eb4cb905779738ef157 Mon Sep 17 00:00:00 2001
From: ReinUsesLisp <reinuseslisp@airmail.cc>
Date: Wed, 21 Aug 2019 02:17:14 -0300
Subject: [PATCH] renderer_opengl: Use block linear swizzling for CPU
 framebuffers

---
 src/video_core/morton.cpp                     | 116 +-----------------
 src/video_core/morton.h                       |   3 -
 .../renderer_opengl/renderer_opengl.cpp       |  66 +++++-----
 3 files changed, 34 insertions(+), 151 deletions(-)

diff --git a/src/video_core/morton.cpp b/src/video_core/morton.cpp
index 3e91cbc83d..084f85e670 100644
--- a/src/video_core/morton.cpp
+++ b/src/video_core/morton.cpp
@@ -25,8 +25,8 @@ static void MortonCopy(u32 stride, u32 block_height, u32 height, u32 block_depth
 
     // With the BCn formats (DXT and DXN), each 4x4 tile is swizzled instead of just individual
     // pixel values.
-    const u32 tile_size_x{GetDefaultBlockWidth(format)};
-    const u32 tile_size_y{GetDefaultBlockHeight(format)};
+    constexpr u32 tile_size_x{GetDefaultBlockWidth(format)};
+    constexpr u32 tile_size_y{GetDefaultBlockHeight(format)};
 
     if constexpr (morton_to_linear) {
         Tegra::Texture::UnswizzleTexture(buffer, addr, tile_size_x, tile_size_y, bytes_per_pixel,
@@ -186,99 +186,6 @@ static MortonCopyFn GetSwizzleFunction(MortonSwizzleMode mode, Surface::PixelFor
     return morton_to_linear_fns[static_cast<std::size_t>(format)];
 }
 
-static u32 MortonInterleave128(u32 x, u32 y) {
-    // 128x128 Z-Order coordinate from 2D coordinates
-    static constexpr u32 xlut[] = {
-        0x0000, 0x0001, 0x0002, 0x0003, 0x0008, 0x0009, 0x000a, 0x000b, 0x0040, 0x0041, 0x0042,
-        0x0043, 0x0048, 0x0049, 0x004a, 0x004b, 0x0800, 0x0801, 0x0802, 0x0803, 0x0808, 0x0809,
-        0x080a, 0x080b, 0x0840, 0x0841, 0x0842, 0x0843, 0x0848, 0x0849, 0x084a, 0x084b, 0x1000,
-        0x1001, 0x1002, 0x1003, 0x1008, 0x1009, 0x100a, 0x100b, 0x1040, 0x1041, 0x1042, 0x1043,
-        0x1048, 0x1049, 0x104a, 0x104b, 0x1800, 0x1801, 0x1802, 0x1803, 0x1808, 0x1809, 0x180a,
-        0x180b, 0x1840, 0x1841, 0x1842, 0x1843, 0x1848, 0x1849, 0x184a, 0x184b, 0x2000, 0x2001,
-        0x2002, 0x2003, 0x2008, 0x2009, 0x200a, 0x200b, 0x2040, 0x2041, 0x2042, 0x2043, 0x2048,
-        0x2049, 0x204a, 0x204b, 0x2800, 0x2801, 0x2802, 0x2803, 0x2808, 0x2809, 0x280a, 0x280b,
-        0x2840, 0x2841, 0x2842, 0x2843, 0x2848, 0x2849, 0x284a, 0x284b, 0x3000, 0x3001, 0x3002,
-        0x3003, 0x3008, 0x3009, 0x300a, 0x300b, 0x3040, 0x3041, 0x3042, 0x3043, 0x3048, 0x3049,
-        0x304a, 0x304b, 0x3800, 0x3801, 0x3802, 0x3803, 0x3808, 0x3809, 0x380a, 0x380b, 0x3840,
-        0x3841, 0x3842, 0x3843, 0x3848, 0x3849, 0x384a, 0x384b, 0x0000, 0x0001, 0x0002, 0x0003,
-        0x0008, 0x0009, 0x000a, 0x000b, 0x0040, 0x0041, 0x0042, 0x0043, 0x0048, 0x0049, 0x004a,
-        0x004b, 0x0800, 0x0801, 0x0802, 0x0803, 0x0808, 0x0809, 0x080a, 0x080b, 0x0840, 0x0841,
-        0x0842, 0x0843, 0x0848, 0x0849, 0x084a, 0x084b, 0x1000, 0x1001, 0x1002, 0x1003, 0x1008,
-        0x1009, 0x100a, 0x100b, 0x1040, 0x1041, 0x1042, 0x1043, 0x1048, 0x1049, 0x104a, 0x104b,
-        0x1800, 0x1801, 0x1802, 0x1803, 0x1808, 0x1809, 0x180a, 0x180b, 0x1840, 0x1841, 0x1842,
-        0x1843, 0x1848, 0x1849, 0x184a, 0x184b, 0x2000, 0x2001, 0x2002, 0x2003, 0x2008, 0x2009,
-        0x200a, 0x200b, 0x2040, 0x2041, 0x2042, 0x2043, 0x2048, 0x2049, 0x204a, 0x204b, 0x2800,
-        0x2801, 0x2802, 0x2803, 0x2808, 0x2809, 0x280a, 0x280b, 0x2840, 0x2841, 0x2842, 0x2843,
-        0x2848, 0x2849, 0x284a, 0x284b, 0x3000, 0x3001, 0x3002, 0x3003, 0x3008, 0x3009, 0x300a,
-        0x300b, 0x3040, 0x3041, 0x3042, 0x3043, 0x3048, 0x3049, 0x304a, 0x304b, 0x3800, 0x3801,
-        0x3802, 0x3803, 0x3808, 0x3809, 0x380a, 0x380b, 0x3840, 0x3841, 0x3842, 0x3843, 0x3848,
-        0x3849, 0x384a, 0x384b, 0x0000, 0x0001, 0x0002, 0x0003, 0x0008, 0x0009, 0x000a, 0x000b,
-        0x0040, 0x0041, 0x0042, 0x0043, 0x0048, 0x0049, 0x004a, 0x004b, 0x0800, 0x0801, 0x0802,
-        0x0803, 0x0808, 0x0809, 0x080a, 0x080b, 0x0840, 0x0841, 0x0842, 0x0843, 0x0848, 0x0849,
-        0x084a, 0x084b, 0x1000, 0x1001, 0x1002, 0x1003, 0x1008, 0x1009, 0x100a, 0x100b, 0x1040,
-        0x1041, 0x1042, 0x1043, 0x1048, 0x1049, 0x104a, 0x104b, 0x1800, 0x1801, 0x1802, 0x1803,
-        0x1808, 0x1809, 0x180a, 0x180b, 0x1840, 0x1841, 0x1842, 0x1843, 0x1848, 0x1849, 0x184a,
-        0x184b, 0x2000, 0x2001, 0x2002, 0x2003, 0x2008, 0x2009, 0x200a, 0x200b, 0x2040, 0x2041,
-        0x2042, 0x2043, 0x2048, 0x2049, 0x204a, 0x204b, 0x2800, 0x2801, 0x2802, 0x2803, 0x2808,
-        0x2809, 0x280a, 0x280b, 0x2840, 0x2841, 0x2842, 0x2843, 0x2848, 0x2849, 0x284a, 0x284b,
-        0x3000, 0x3001, 0x3002, 0x3003, 0x3008, 0x3009, 0x300a, 0x300b, 0x3040, 0x3041, 0x3042,
-        0x3043, 0x3048, 0x3049, 0x304a, 0x304b, 0x3800, 0x3801, 0x3802, 0x3803, 0x3808, 0x3809,
-        0x380a, 0x380b, 0x3840, 0x3841, 0x3842, 0x3843, 0x3848, 0x3849, 0x384a, 0x384b,
-    };
-    static constexpr u32 ylut[] = {
-        0x0000, 0x0004, 0x0010, 0x0014, 0x0020, 0x0024, 0x0030, 0x0034, 0x0080, 0x0084, 0x0090,
-        0x0094, 0x00a0, 0x00a4, 0x00b0, 0x00b4, 0x0100, 0x0104, 0x0110, 0x0114, 0x0120, 0x0124,
-        0x0130, 0x0134, 0x0180, 0x0184, 0x0190, 0x0194, 0x01a0, 0x01a4, 0x01b0, 0x01b4, 0x0200,
-        0x0204, 0x0210, 0x0214, 0x0220, 0x0224, 0x0230, 0x0234, 0x0280, 0x0284, 0x0290, 0x0294,
-        0x02a0, 0x02a4, 0x02b0, 0x02b4, 0x0300, 0x0304, 0x0310, 0x0314, 0x0320, 0x0324, 0x0330,
-        0x0334, 0x0380, 0x0384, 0x0390, 0x0394, 0x03a0, 0x03a4, 0x03b0, 0x03b4, 0x0400, 0x0404,
-        0x0410, 0x0414, 0x0420, 0x0424, 0x0430, 0x0434, 0x0480, 0x0484, 0x0490, 0x0494, 0x04a0,
-        0x04a4, 0x04b0, 0x04b4, 0x0500, 0x0504, 0x0510, 0x0514, 0x0520, 0x0524, 0x0530, 0x0534,
-        0x0580, 0x0584, 0x0590, 0x0594, 0x05a0, 0x05a4, 0x05b0, 0x05b4, 0x0600, 0x0604, 0x0610,
-        0x0614, 0x0620, 0x0624, 0x0630, 0x0634, 0x0680, 0x0684, 0x0690, 0x0694, 0x06a0, 0x06a4,
-        0x06b0, 0x06b4, 0x0700, 0x0704, 0x0710, 0x0714, 0x0720, 0x0724, 0x0730, 0x0734, 0x0780,
-        0x0784, 0x0790, 0x0794, 0x07a0, 0x07a4, 0x07b0, 0x07b4, 0x0000, 0x0004, 0x0010, 0x0014,
-        0x0020, 0x0024, 0x0030, 0x0034, 0x0080, 0x0084, 0x0090, 0x0094, 0x00a0, 0x00a4, 0x00b0,
-        0x00b4, 0x0100, 0x0104, 0x0110, 0x0114, 0x0120, 0x0124, 0x0130, 0x0134, 0x0180, 0x0184,
-        0x0190, 0x0194, 0x01a0, 0x01a4, 0x01b0, 0x01b4, 0x0200, 0x0204, 0x0210, 0x0214, 0x0220,
-        0x0224, 0x0230, 0x0234, 0x0280, 0x0284, 0x0290, 0x0294, 0x02a0, 0x02a4, 0x02b0, 0x02b4,
-        0x0300, 0x0304, 0x0310, 0x0314, 0x0320, 0x0324, 0x0330, 0x0334, 0x0380, 0x0384, 0x0390,
-        0x0394, 0x03a0, 0x03a4, 0x03b0, 0x03b4, 0x0400, 0x0404, 0x0410, 0x0414, 0x0420, 0x0424,
-        0x0430, 0x0434, 0x0480, 0x0484, 0x0490, 0x0494, 0x04a0, 0x04a4, 0x04b0, 0x04b4, 0x0500,
-        0x0504, 0x0510, 0x0514, 0x0520, 0x0524, 0x0530, 0x0534, 0x0580, 0x0584, 0x0590, 0x0594,
-        0x05a0, 0x05a4, 0x05b0, 0x05b4, 0x0600, 0x0604, 0x0610, 0x0614, 0x0620, 0x0624, 0x0630,
-        0x0634, 0x0680, 0x0684, 0x0690, 0x0694, 0x06a0, 0x06a4, 0x06b0, 0x06b4, 0x0700, 0x0704,
-        0x0710, 0x0714, 0x0720, 0x0724, 0x0730, 0x0734, 0x0780, 0x0784, 0x0790, 0x0794, 0x07a0,
-        0x07a4, 0x07b0, 0x07b4, 0x0000, 0x0004, 0x0010, 0x0014, 0x0020, 0x0024, 0x0030, 0x0034,
-        0x0080, 0x0084, 0x0090, 0x0094, 0x00a0, 0x00a4, 0x00b0, 0x00b4, 0x0100, 0x0104, 0x0110,
-        0x0114, 0x0120, 0x0124, 0x0130, 0x0134, 0x0180, 0x0184, 0x0190, 0x0194, 0x01a0, 0x01a4,
-        0x01b0, 0x01b4, 0x0200, 0x0204, 0x0210, 0x0214, 0x0220, 0x0224, 0x0230, 0x0234, 0x0280,
-        0x0284, 0x0290, 0x0294, 0x02a0, 0x02a4, 0x02b0, 0x02b4, 0x0300, 0x0304, 0x0310, 0x0314,
-        0x0320, 0x0324, 0x0330, 0x0334, 0x0380, 0x0384, 0x0390, 0x0394, 0x03a0, 0x03a4, 0x03b0,
-        0x03b4, 0x0400, 0x0404, 0x0410, 0x0414, 0x0420, 0x0424, 0x0430, 0x0434, 0x0480, 0x0484,
-        0x0490, 0x0494, 0x04a0, 0x04a4, 0x04b0, 0x04b4, 0x0500, 0x0504, 0x0510, 0x0514, 0x0520,
-        0x0524, 0x0530, 0x0534, 0x0580, 0x0584, 0x0590, 0x0594, 0x05a0, 0x05a4, 0x05b0, 0x05b4,
-        0x0600, 0x0604, 0x0610, 0x0614, 0x0620, 0x0624, 0x0630, 0x0634, 0x0680, 0x0684, 0x0690,
-        0x0694, 0x06a0, 0x06a4, 0x06b0, 0x06b4, 0x0700, 0x0704, 0x0710, 0x0714, 0x0720, 0x0724,
-        0x0730, 0x0734, 0x0780, 0x0784, 0x0790, 0x0794, 0x07a0, 0x07a4, 0x07b0, 0x07b4,
-    };
-    return xlut[x % 128] + ylut[y % 128];
-}
-
-static u32 GetMortonOffset128(u32 x, u32 y, u32 bytes_per_pixel) {
-    // Calculates the offset of the position of the pixel in Morton order
-    // Framebuffer images are split into 128x128 tiles.
-
-    constexpr u32 block_height = 128;
-    const u32 coarse_x = x & ~127;
-
-    const u32 i = MortonInterleave128(x, y);
-
-    const u32 offset = coarse_x * block_height;
-
-    return (i + offset) * bytes_per_pixel;
-}
-
 void MortonSwizzle(MortonSwizzleMode mode, Surface::PixelFormat format, u32 stride,
                    u32 block_height, u32 height, u32 block_depth, u32 depth, u32 tile_width_spacing,
                    u8* buffer, u8* addr) {
@@ -286,23 +193,4 @@ void MortonSwizzle(MortonSwizzleMode mode, Surface::PixelFormat format, u32 stri
                                      tile_width_spacing, buffer, addr);
 }
 
-void MortonCopyPixels128(MortonSwizzleMode mode, u32 width, u32 height, u32 bytes_per_pixel,
-                         u32 linear_bytes_per_pixel, u8* morton_data, u8* linear_data) {
-    const bool morton_to_linear = mode == MortonSwizzleMode::MortonToLinear;
-    u8* data_ptrs[2];
-    for (u32 y = 0; y < height; ++y) {
-        for (u32 x = 0; x < width; ++x) {
-            const u32 coarse_y = y & ~127;
-            const u32 morton_offset =
-                GetMortonOffset128(x, y, bytes_per_pixel) + coarse_y * width * bytes_per_pixel;
-            const u32 linear_pixel_index = (x + y * width) * linear_bytes_per_pixel;
-
-            data_ptrs[morton_to_linear ? 1 : 0] = morton_data + morton_offset;
-            data_ptrs[morton_to_linear ? 0 : 1] = &linear_data[linear_pixel_index];
-
-            std::memcpy(data_ptrs[0], data_ptrs[1], bytes_per_pixel);
-        }
-    }
-}
-
 } // namespace VideoCore
diff --git a/src/video_core/morton.h b/src/video_core/morton.h
index ee5b45555e..b714a7e3fa 100644
--- a/src/video_core/morton.h
+++ b/src/video_core/morton.h
@@ -15,7 +15,4 @@ void MortonSwizzle(MortonSwizzleMode mode, VideoCore::Surface::PixelFormat forma
                    u32 block_height, u32 height, u32 block_depth, u32 depth, u32 tile_width_spacing,
                    u8* buffer, u8* addr);
 
-void MortonCopyPixels128(MortonSwizzleMode mode, u32 width, u32 height, u32 bytes_per_pixel,
-                         u32 linear_bytes_per_pixel, u8* morton_data, u8* linear_data);
-
 } // namespace VideoCore
diff --git a/src/video_core/renderer_opengl/renderer_opengl.cpp b/src/video_core/renderer_opengl/renderer_opengl.cpp
index f26adc3880..8c44b330ea 100644
--- a/src/video_core/renderer_opengl/renderer_opengl.cpp
+++ b/src/video_core/renderer_opengl/renderer_opengl.cpp
@@ -147,45 +147,43 @@ void RendererOpenGL::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
  * Loads framebuffer from emulated memory into the active OpenGL texture.
  */
 void RendererOpenGL::LoadFBToScreenInfo(const Tegra::FramebufferConfig& framebuffer) {
-    const auto pixel_format{
-        VideoCore::Surface::PixelFormatFromGPUPixelFormat(framebuffer.pixel_format)};
-    const u32 bytes_per_pixel{VideoCore::Surface::GetBytesPerPixel(pixel_format)};
-    const u64 size_in_bytes{framebuffer.stride * framebuffer.height * bytes_per_pixel};
-    const VAddr framebuffer_addr{framebuffer.address + framebuffer.offset};
-
     // Framebuffer orientation handling
     framebuffer_transform_flags = framebuffer.transform_flags;
     framebuffer_crop_rect = framebuffer.crop_rect;
 
-    // Ensure no bad interactions with GL_UNPACK_ALIGNMENT, which by default
-    // only allows rows to have a memory alignement of 4.
-    ASSERT(framebuffer.stride % 4 == 0);
-
-    if (!rasterizer->AccelerateDisplay(framebuffer, framebuffer_addr, framebuffer.stride)) {
-        // Reset the screen info's display texture to its own permanent texture
-        screen_info.display_texture = screen_info.texture.resource.handle;
-
-        rasterizer->FlushRegion(ToCacheAddr(Memory::GetPointer(framebuffer_addr)), size_in_bytes);
-
-        constexpr u32 linear_bpp = 4;
-        VideoCore::MortonCopyPixels128(VideoCore::MortonSwizzleMode::MortonToLinear,
-                                       framebuffer.width, framebuffer.height, bytes_per_pixel,
-                                       linear_bpp, Memory::GetPointer(framebuffer_addr),
-                                       gl_framebuffer_data.data());
-
-        glPixelStorei(GL_UNPACK_ROW_LENGTH, static_cast<GLint>(framebuffer.stride));
-
-        // Update existing texture
-        // TODO: Test what happens on hardware when you change the framebuffer dimensions so that
-        //       they differ from the LCD resolution.
-        // TODO: Applications could theoretically crash yuzu here by specifying too large
-        //       framebuffer sizes. We should make sure that this cannot happen.
-        glTextureSubImage2D(screen_info.texture.resource.handle, 0, 0, 0, framebuffer.width,
-                            framebuffer.height, screen_info.texture.gl_format,
-                            screen_info.texture.gl_type, gl_framebuffer_data.data());
-
-        glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
+    const VAddr framebuffer_addr{framebuffer.address + framebuffer.offset};
+    if (rasterizer->AccelerateDisplay(framebuffer, framebuffer_addr, framebuffer.stride)) {
+        return;
     }
+
+    // Reset the screen info's display texture to its own permanent texture
+    screen_info.display_texture = screen_info.texture.resource.handle;
+
+    const auto pixel_format{
+        VideoCore::Surface::PixelFormatFromGPUPixelFormat(framebuffer.pixel_format)};
+    const u32 bytes_per_pixel{VideoCore::Surface::GetBytesPerPixel(pixel_format)};
+    const u64 size_in_bytes{framebuffer.stride * framebuffer.height * bytes_per_pixel};
+    const auto host_ptr{Memory::GetPointer(framebuffer_addr)};
+    rasterizer->FlushRegion(ToCacheAddr(host_ptr), size_in_bytes);
+
+    // TODO(Rodrigo): Read this from HLE
+    constexpr u32 block_height_log2 = 4;
+    VideoCore::MortonSwizzle(VideoCore::MortonSwizzleMode::MortonToLinear, pixel_format,
+                             framebuffer.stride, block_height_log2, framebuffer.height, 0, 1, 1,
+                             gl_framebuffer_data.data(), host_ptr);
+
+    glPixelStorei(GL_UNPACK_ROW_LENGTH, static_cast<GLint>(framebuffer.stride));
+
+    // Update existing texture
+    // TODO: Test what happens on hardware when you change the framebuffer dimensions so that
+    //       they differ from the LCD resolution.
+    // TODO: Applications could theoretically crash yuzu here by specifying too large
+    //       framebuffer sizes. We should make sure that this cannot happen.
+    glTextureSubImage2D(screen_info.texture.resource.handle, 0, 0, 0, framebuffer.width,
+                        framebuffer.height, screen_info.texture.gl_format,
+                        screen_info.texture.gl_type, gl_framebuffer_data.data());
+
+    glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
 }
 
 /**