diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h index 180be4ff46..a15ca9e43d 100644 --- a/src/video_core/engines/maxwell_3d.h +++ b/src/video_core/engines/maxwell_3d.h @@ -413,7 +413,7 @@ public: struct { u32 address_high; u32 address_low; - u32 format; + Tegra::DepthFormat format; u32 block_dimensions; u32 layer_stride; diff --git a/src/video_core/gpu.h b/src/video_core/gpu.h index d0a4ac2671..cc5ca656e4 100644 --- a/src/video_core/gpu.h +++ b/src/video_core/gpu.h @@ -24,6 +24,15 @@ enum class RenderTargetFormat : u32 { R11G11B10_FLOAT = 0xE0, }; +enum class DepthFormat : u32 { + Z32_FLOAT = 0xA, + Z16_UNORM = 0x13, + S8_Z24_UNORM = 0x14, + Z24_X8_UNORM = 0x15, + Z24_S8_UNORM = 0x16, + Z24_C8_UNORM = 0x18, +}; + /// Returns the number of bytes per pixel of each rendertarget format. u32 RenderTargetBytesPerPixel(RenderTargetFormat format); diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp index 62ee45a368..b80f4336d7 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp @@ -307,7 +307,9 @@ void RasterizerOpenGL::DrawArrays() { // TODO(bunnei): Implement these const bool has_stencil = false; const bool using_color_fb = true; - const bool using_depth_fb = false; + + const bool using_depth_fb = regs.zeta.Address() != 0; + const MathUtil::Rectangle viewport_rect{regs.viewport_transform[0].GetRect()}; const bool write_color_fb = diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp index ae48378f31..35ad4f1615 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp @@ -84,22 +84,18 @@ static constexpr std::array tex_form true}, // DXT45 {GL_COMPRESSED_RED_RGTC1, GL_RED, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm, true}, // DXN1 {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_4X4 + + // DepthStencil formats + {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8, ComponentType::UNorm, + false}, // Z24S8 }}; static const FormatTuple& GetFormatTuple(PixelFormat pixel_format, ComponentType component_type) { - const SurfaceType type = SurfaceParams::GetFormatType(pixel_format); - if (type == SurfaceType::ColorTexture) { - ASSERT(static_cast(pixel_format) < tex_format_tuples.size()); - auto& format = tex_format_tuples[static_cast(pixel_format)]; - ASSERT(component_type == format.component_type); - return format; - } else if (type == SurfaceType::Depth || type == SurfaceType::DepthStencil) { - // TODO(Subv): Implement depth formats - ASSERT_MSG(false, "Unimplemented"); - } + ASSERT(static_cast(pixel_format) < tex_format_tuples.size()); + auto& format = tex_format_tuples[static_cast(pixel_format)]; + ASSERT(component_type == format.component_type); - UNREACHABLE(); - return {}; + return format; } VAddr SurfaceParams::GetCpuAddr() const { @@ -149,11 +145,17 @@ void MortonCopy(u32 stride, u32 block_height, u32 height, u8* gl_buffer, Tegra:: const auto& gpu = Core::System::GetInstance().GPU(); if (morton_to_gl) { - auto data = Tegra::Texture::UnswizzleTexture( - *gpu.memory_manager->GpuToCpuAddress(addr), - SurfaceParams::TextureFormatFromPixelFormat(format), stride, height, block_height); - - std::memcpy(gl_buffer, data.data(), data.size()); + if (SurfaceParams::GetFormatType(format) == SurfaceType::ColorTexture) { + auto data = Tegra::Texture::UnswizzleTexture( + *gpu.memory_manager->GpuToCpuAddress(addr), + SurfaceParams::TextureFormatFromPixelFormat(format), stride, height, block_height); + std::memcpy(gl_buffer, data.data(), data.size()); + } else { + auto data = Tegra::Texture::UnswizzleDepthTexture( + *gpu.memory_manager->GpuToCpuAddress(addr), + SurfaceParams::DepthFormatFromPixelFormat(format), stride, height, block_height); + std::memcpy(gl_buffer, data.data(), data.size()); + } } else { // TODO(bunnei): Assumes the default rendering GOB size of 16 (128 lines). We should // check the configuration for this and perform more generic un/swizzle @@ -174,7 +176,7 @@ static constexpr std::array, MortonCopy, MortonCopy, MortonCopy, MortonCopy, MortonCopy, - MortonCopy, + MortonCopy, MortonCopy, }; static constexpr std::array, + MortonCopy, }; // Allocate an uninitialized texture of appropriate size and format for the surface @@ -397,9 +400,15 @@ SurfaceSurfaceRect_Tuple RasterizerCacheOpenGL::GetFramebufferSurfaces( // get color and depth surfaces const SurfaceParams color_params{SurfaceParams::CreateForFramebuffer(regs.rt[0])}; - const SurfaceParams depth_params{color_params}; + SurfaceParams depth_params{color_params}; - ASSERT_MSG(!using_depth_fb, "depth buffer is unimplemented"); + if (using_depth_fb) { + depth_params.addr = regs.zeta.Address(); + depth_params.pixel_format = SurfaceParams::PixelFormatFromDepthFormat(regs.zeta.format); + depth_params.component_type = SurfaceParams::ComponentTypeFromDepthFormat(regs.zeta.format); + depth_params.type = SurfaceParams::GetFormatType(depth_params.pixel_format); + depth_params.size_in_bytes = depth_params.SizeInBytes(); + } MathUtil::Rectangle color_rect{}; Surface color_surface; diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.h b/src/video_core/renderer_opengl/gl_rasterizer_cache.h index 99be250b44..eea432b0b0 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer_cache.h +++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.h @@ -37,7 +37,14 @@ struct SurfaceParams { DXN1 = 11, // This is also known as BC4 ASTC_2D_4X4 = 12, - Max, + MaxColorFormat, + + // DepthStencil formats + Z24S8 = 13, + + MaxDepthStencilFormat, + + Max = MaxDepthStencilFormat, Invalid = 255, }; @@ -84,6 +91,7 @@ struct SurfaceParams { 4, // DXT45 4, // DXN1 4, // ASTC_2D_4X4 + 1, // Z24S8 }}; ASSERT(static_cast(format) < compression_factor_table.size()); @@ -108,6 +116,7 @@ struct SurfaceParams { 128, // DXT45 64, // DXN1 32, // ASTC_2D_4X4 + 32, // Z24S8 }}; ASSERT(static_cast(format) < bpp_table.size()); @@ -117,6 +126,16 @@ struct SurfaceParams { return GetFormatBpp(pixel_format); } + static PixelFormat PixelFormatFromDepthFormat(Tegra::DepthFormat format) { + switch (format) { + case Tegra::DepthFormat::Z24_S8_UNORM: + return PixelFormat::Z24S8; + default: + NGLOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast(format)); + UNREACHABLE(); + } + } + static PixelFormat PixelFormatFromRenderTargetFormat(Tegra::RenderTargetFormat format) { switch (format) { case Tegra::RenderTargetFormat::RGBA8_UNORM: @@ -205,6 +224,15 @@ struct SurfaceParams { } } + static Tegra::DepthFormat DepthFormatFromPixelFormat(PixelFormat format) { + switch (format) { + case PixelFormat::Z24S8: + return Tegra::DepthFormat::Z24_S8_UNORM; + default: + UNREACHABLE(); + } + } + static ComponentType ComponentTypeFromTexture(Tegra::Texture::ComponentType type) { // TODO(Subv): Implement more component types switch (type) { @@ -244,11 +272,26 @@ struct SurfaceParams { } } + static ComponentType ComponentTypeFromDepthFormat(Tegra::DepthFormat format) { + switch (format) { + case Tegra::DepthFormat::Z24_S8_UNORM: + return ComponentType::UNorm; + default: + NGLOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast(format)); + UNREACHABLE(); + } + } + static SurfaceType GetFormatType(PixelFormat pixel_format) { - if (static_cast(pixel_format) < MaxPixelFormat) { + if (static_cast(pixel_format) < static_cast(PixelFormat::MaxColorFormat)) { return SurfaceType::ColorTexture; } + if (static_cast(pixel_format) < + static_cast(PixelFormat::MaxDepthStencilFormat)) { + return SurfaceType::DepthStencil; + } + // TODO(Subv): Implement the other formats ASSERT(false); diff --git a/src/video_core/textures/decoders.cpp b/src/video_core/textures/decoders.cpp index eaf15da328..680f22ddb8 100644 --- a/src/video_core/textures/decoders.cpp +++ b/src/video_core/textures/decoders.cpp @@ -5,6 +5,7 @@ #include #include "common/assert.h" #include "core/memory.h" +#include "video_core/gpu.h" #include "video_core/textures/decoders.h" #include "video_core/textures/texture.h" @@ -73,6 +74,16 @@ u32 BytesPerPixel(TextureFormat format) { } } +static u32 DepthBytesPerPixel(DepthFormat format) { + switch (format) { + case DepthFormat::Z24_S8_UNORM: + return 4; + default: + UNIMPLEMENTED_MSG("Format not implemented"); + break; + } +} + std::vector UnswizzleTexture(VAddr address, TextureFormat format, u32 width, u32 height, u32 block_height) { u8* data = Memory::GetPointer(address); @@ -110,6 +121,26 @@ std::vector UnswizzleTexture(VAddr address, TextureFormat format, u32 width, return unswizzled_data; } +std::vector UnswizzleDepthTexture(VAddr address, DepthFormat format, u32 width, u32 height, + u32 block_height) { + u8* data = Memory::GetPointer(address); + u32 bytes_per_pixel = DepthBytesPerPixel(format); + + std::vector unswizzled_data(width * height * bytes_per_pixel); + + switch (format) { + case DepthFormat::Z24_S8_UNORM: + CopySwizzledData(width, height, bytes_per_pixel, bytes_per_pixel, data, + unswizzled_data.data(), true, block_height); + break; + default: + UNIMPLEMENTED_MSG("Format not implemented"); + break; + } + + return unswizzled_data; +} + std::vector DecodeTexture(const std::vector& texture_data, TextureFormat format, u32 width, u32 height) { std::vector rgba_data; diff --git a/src/video_core/textures/decoders.h b/src/video_core/textures/decoders.h index 2562c4b060..2b088c077e 100644 --- a/src/video_core/textures/decoders.h +++ b/src/video_core/textures/decoders.h @@ -17,6 +17,12 @@ namespace Texture { std::vector UnswizzleTexture(VAddr address, TextureFormat format, u32 width, u32 height, u32 block_height = TICEntry::DefaultBlockHeight); +/** + * Unswizzles a swizzled depth texture without changing its format. + */ +std::vector UnswizzleDepthTexture(VAddr address, DepthFormat format, u32 width, u32 height, + u32 block_height = TICEntry::DefaultBlockHeight); + /// Copies texture data from a buffer and performs swizzling/unswizzling as necessary. void CopySwizzledData(u32 width, u32 height, u32 bytes_per_pixel, u32 out_bytes_per_pixel, u8* swizzled_data, u8* unswizzled_data, bool unswizzle, u32 block_height);