summaryrefslogtreecommitdiffstats
path: root/src
diff options
context:
space:
mode:
Diffstat (limited to 'src')
-rw-r--r--src/audio_core/time_stretch.cpp6
-rw-r--r--src/video_core/CMakeLists.txt2
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer.cpp10
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer.h1
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer_cache.cpp243
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer_cache.h87
-rw-r--r--src/video_core/renderer_opengl/gl_shader_cache.cpp5
-rw-r--r--src/video_core/renderer_opengl/maxwell_to_gl.h27
-rw-r--r--src/video_core/renderer_opengl/utils.cpp38
-rw-r--r--src/video_core/renderer_opengl/utils.h15
-rw-r--r--src/video_core/textures/decoders.h6
-rw-r--r--src/video_core/utils.h26
12 files changed, 321 insertions, 145 deletions
diff --git a/src/audio_core/time_stretch.cpp b/src/audio_core/time_stretch.cpp
index cee8b12dd..2fe0b3aef 100644
--- a/src/audio_core/time_stretch.cpp
+++ b/src/audio_core/time_stretch.cpp
@@ -32,10 +32,10 @@ std::size_t TimeStretcher::Process(const s16* in, std::size_t num_in, s16* out,
// We were given actual_samples number of samples, and num_samples were requested from us.
double current_ratio = static_cast<double>(num_in) / static_cast<double>(num_out);
- const double max_latency = 1.0; // seconds
+ const double max_latency = 0.25; // seconds
const double max_backlog = m_sample_rate * max_latency;
const double backlog_fullness = m_sound_touch.numSamples() / max_backlog;
- if (backlog_fullness > 5.0) {
+ if (backlog_fullness > 4.0) {
// Too many samples in backlog: Don't push anymore on
num_in = 0;
}
@@ -49,7 +49,7 @@ std::size_t TimeStretcher::Process(const s16* in, std::size_t num_in, s16* out,
// This low-pass filter smoothes out variance in the calculated stretch ratio.
// The time-scale determines how responsive this filter is.
- constexpr double lpf_time_scale = 2.0; // seconds
+ constexpr double lpf_time_scale = 0.712; // seconds
const double lpf_gain = 1.0 - std::exp(-time_delta / lpf_time_scale);
m_stretch_ratio += lpf_gain * (current_ratio - m_stretch_ratio);
diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt
index 09ecc5bad..c5f7128ec 100644
--- a/src/video_core/CMakeLists.txt
+++ b/src/video_core/CMakeLists.txt
@@ -51,6 +51,8 @@ add_library(video_core STATIC
renderer_opengl/maxwell_to_gl.h
renderer_opengl/renderer_opengl.cpp
renderer_opengl/renderer_opengl.h
+ renderer_opengl/utils.cpp
+ renderer_opengl/utils.h
textures/astc.cpp
textures/astc.h
textures/decoders.cpp
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index cb180b93c..7bb5544fc 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -731,11 +731,15 @@ void RasterizerOpenGL::SamplerInfo::SyncWithConfig(const Tegra::Texture::TSCEntr
if (mag_filter != config.mag_filter) {
mag_filter = config.mag_filter;
- glSamplerParameteri(s, GL_TEXTURE_MAG_FILTER, MaxwellToGL::TextureFilterMode(mag_filter));
+ glSamplerParameteri(
+ s, GL_TEXTURE_MAG_FILTER,
+ MaxwellToGL::TextureFilterMode(mag_filter, Tegra::Texture::TextureMipmapFilter::None));
}
- if (min_filter != config.min_filter) {
+ if (min_filter != config.min_filter || mip_filter != config.mip_filter) {
min_filter = config.min_filter;
- glSamplerParameteri(s, GL_TEXTURE_MIN_FILTER, MaxwellToGL::TextureFilterMode(min_filter));
+ mip_filter = config.mip_filter;
+ glSamplerParameteri(s, GL_TEXTURE_MIN_FILTER,
+ MaxwellToGL::TextureFilterMode(min_filter, mip_filter));
}
if (wrap_u != config.wrap_u) {
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h
index 5020a5392..7b0615125 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.h
+++ b/src/video_core/renderer_opengl/gl_rasterizer.h
@@ -93,6 +93,7 @@ private:
private:
Tegra::Texture::TextureFilter mag_filter;
Tegra::Texture::TextureFilter min_filter;
+ Tegra::Texture::TextureMipmapFilter mip_filter;
Tegra::Texture::WrapMode wrap_u;
Tegra::Texture::WrapMode wrap_v;
Tegra::Texture::WrapMode wrap_p;
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
index b057e2efa..1d43a419d 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
@@ -16,6 +16,7 @@
#include "core/settings.h"
#include "video_core/engines/maxwell_3d.h"
#include "video_core/renderer_opengl/gl_rasterizer_cache.h"
+#include "video_core/renderer_opengl/utils.h"
#include "video_core/textures/astc.h"
#include "video_core/textures/decoders.h"
#include "video_core/utils.h"
@@ -90,27 +91,36 @@ void SurfaceParams::InitCacheParameters(Tegra::GPUVAddr gpu_addr_) {
}
}
-std::size_t SurfaceParams::InnerMemorySize(bool layer_only) const {
+std::size_t SurfaceParams::InnerMipmapMemorySize(u32 mip_level, bool force_gl, bool layer_only,
+ bool uncompressed) const {
const u32 compression_factor{GetCompressionFactor(pixel_format)};
const u32 bytes_per_pixel{GetBytesPerPixel(pixel_format)};
u32 m_depth = (layer_only ? 1U : depth);
- u32 m_width = std::max(1U, width / compression_factor);
- u32 m_height = std::max(1U, height / compression_factor);
- std::size_t size = Tegra::Texture::CalculateSize(is_tiled, bytes_per_pixel, m_width, m_height,
- m_depth, block_height, block_depth);
- u32 m_block_height = block_height;
- u32 m_block_depth = block_depth;
- std::size_t block_size_bytes = 512 * block_height * block_depth; // 512 is GOB size
- for (u32 i = 1; i < max_mip_level; i++) {
- m_width = std::max(1U, m_width / 2);
- m_height = std::max(1U, m_height / 2);
- m_depth = std::max(1U, m_depth / 2);
- m_block_height = std::max(1U, m_block_height / 2);
- m_block_depth = std::max(1U, m_block_depth / 2);
- size += Tegra::Texture::CalculateSize(is_tiled, bytes_per_pixel, m_width, m_height, m_depth,
- m_block_height, m_block_depth);
+ u32 m_width = MipWidth(mip_level);
+ u32 m_height = MipHeight(mip_level);
+ m_width = uncompressed ? m_width
+ : std::max(1U, (m_width + compression_factor - 1) / compression_factor);
+ m_height = uncompressed
+ ? m_height
+ : std::max(1U, (m_height + compression_factor - 1) / compression_factor);
+ m_depth = std::max(1U, m_depth >> mip_level);
+ u32 m_block_height = MipBlockHeight(mip_level);
+ u32 m_block_depth = MipBlockDepth(mip_level);
+ return Tegra::Texture::CalculateSize(force_gl ? false : is_tiled, bytes_per_pixel, m_width,
+ m_height, m_depth, m_block_height, m_block_depth);
+}
+
+std::size_t SurfaceParams::InnerMemorySize(bool force_gl, bool layer_only,
+ bool uncompressed) const {
+ std::size_t block_size_bytes = Tegra::Texture::GetGOBSize() * block_height * block_depth;
+ std::size_t size = 0;
+ for (u32 i = 0; i < max_mip_level; i++) {
+ size += InnerMipmapMemorySize(i, force_gl, layer_only, uncompressed);
+ }
+ if (!force_gl && is_tiled) {
+ size = Common::AlignUp(size, block_size_bytes);
}
- return is_tiled ? Common::AlignUp(size, block_size_bytes) : size;
+ return size;
}
/*static*/ SurfaceParams SurfaceParams::CreateForTexture(
@@ -188,7 +198,7 @@ std::size_t SurfaceParams::InnerMemorySize(bool layer_only) const {
params.unaligned_height = config.height;
params.target = SurfaceTarget::Texture2D;
params.depth = 1;
- params.max_mip_level = 0;
+ params.max_mip_level = 1;
params.is_layered = false;
// Render target specific parameters, not used for caching
@@ -222,7 +232,7 @@ std::size_t SurfaceParams::InnerMemorySize(bool layer_only) const {
params.unaligned_height = zeta_height;
params.target = SurfaceTarget::Texture2D;
params.depth = 1;
- params.max_mip_level = 0;
+ params.max_mip_level = 1;
params.is_layered = false;
params.rt = {};
@@ -249,7 +259,7 @@ std::size_t SurfaceParams::InnerMemorySize(bool layer_only) const {
params.unaligned_height = config.height;
params.target = SurfaceTarget::Texture2D;
params.depth = 1;
- params.max_mip_level = 0;
+ params.max_mip_level = 1;
params.rt = {};
params.InitCacheParameters(config.Address());
@@ -273,7 +283,7 @@ static constexpr std::array<FormatTuple, SurfaceParams::MaxPixelFormat> tex_form
{GL_R11F_G11F_B10F, GL_RGB, GL_UNSIGNED_INT_10F_11F_11F_REV, ComponentType::Float,
false}, // R11FG11FB10F
{GL_RGBA32UI, GL_RGBA_INTEGER, GL_UNSIGNED_INT, ComponentType::UInt, false}, // RGBA32UI
- {GL_COMPRESSED_RGB_S3TC_DXT1_EXT, GL_RGB, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
+ {GL_COMPRESSED_RGBA_S3TC_DXT1_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
true}, // DXT1
{GL_COMPRESSED_RGBA_S3TC_DXT3_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
true}, // DXT23
@@ -318,7 +328,7 @@ static constexpr std::array<FormatTuple, SurfaceParams::MaxPixelFormat> tex_form
{GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_5X4
{GL_SRGB8_ALPHA8, GL_BGRA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // BGRA8
// Compressed sRGB formats
- {GL_COMPRESSED_SRGB_S3TC_DXT1_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
+ {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
true}, // DXT1_SRGB
{GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
true}, // DXT23_SRGB
@@ -373,13 +383,13 @@ static const FormatTuple& GetFormatTuple(PixelFormat pixel_format, ComponentType
return format;
}
-MathUtil::Rectangle<u32> SurfaceParams::GetRect() const {
- u32 actual_height{unaligned_height};
+MathUtil::Rectangle<u32> SurfaceParams::GetRect(u32 mip_level) const {
+ u32 actual_height{std::max(1U, unaligned_height >> mip_level)};
if (IsPixelFormatASTC(pixel_format)) {
// ASTC formats must stop at the ATSC block size boundary
actual_height = Common::AlignDown(actual_height, GetASTCBlockSize(pixel_format).second);
}
- return {0, actual_height, width, 0};
+ return {0, actual_height, MipWidth(mip_level), 0};
}
/// Returns true if the specified PixelFormat is a BCn format, e.g. DXT or DXN
@@ -563,28 +573,31 @@ static constexpr GLConversionArray gl_to_morton_fns = {
};
void SwizzleFunc(const GLConversionArray& functions, const SurfaceParams& params,
- std::vector<u8>& gl_buffer) {
- u32 depth = params.depth;
+ std::vector<u8>& gl_buffer, u32 mip_level) {
+ u32 depth = params.MipDepth(mip_level);
if (params.target == SurfaceParams::SurfaceTarget::Texture2D) {
// TODO(Blinkhawk): Eliminate this condition once all texture types are implemented.
depth = 1U;
}
if (params.is_layered) {
- u64 offset = 0;
+ u64 offset = params.GetMipmapLevelOffset(mip_level);
u64 offset_gl = 0;
u64 layer_size = params.LayerMemorySize();
- u64 gl_size = params.LayerSizeGL();
- for (u32 i = 0; i < depth; i++) {
+ u64 gl_size = params.LayerSizeGL(mip_level);
+ for (u32 i = 0; i < params.depth; i++) {
functions[static_cast<std::size_t>(params.pixel_format)](
- params.width, params.block_height, params.height, params.block_depth, 1,
+ params.MipWidth(mip_level), params.MipBlockHeight(mip_level),
+ params.MipHeight(mip_level), params.MipBlockDepth(mip_level), 1,
gl_buffer.data() + offset_gl, gl_size, params.addr + offset);
offset += layer_size;
offset_gl += gl_size;
}
} else {
+ u64 offset = params.GetMipmapLevelOffset(mip_level);
functions[static_cast<std::size_t>(params.pixel_format)](
- params.width, params.block_height, params.height, params.block_depth, depth,
- gl_buffer.data(), gl_buffer.size(), params.addr);
+ params.MipWidth(mip_level), params.MipBlockHeight(mip_level),
+ params.MipHeight(mip_level), params.MipBlockDepth(mip_level), depth, gl_buffer.data(),
+ gl_buffer.size(), params.addr + offset);
}
}
@@ -839,34 +852,41 @@ CachedSurface::CachedSurface(const SurfaceParams& params)
// Only pre-create the texture for non-compressed textures.
switch (params.target) {
case SurfaceParams::SurfaceTarget::Texture1D:
- glTexStorage1D(SurfaceTargetToGL(params.target), 1, format_tuple.internal_format,
- rect.GetWidth());
+ glTexStorage1D(SurfaceTargetToGL(params.target), params.max_mip_level,
+ format_tuple.internal_format, rect.GetWidth());
break;
case SurfaceParams::SurfaceTarget::Texture2D:
case SurfaceParams::SurfaceTarget::TextureCubemap:
- glTexStorage2D(SurfaceTargetToGL(params.target), 1, format_tuple.internal_format,
- rect.GetWidth(), rect.GetHeight());
+ glTexStorage2D(SurfaceTargetToGL(params.target), params.max_mip_level,
+ format_tuple.internal_format, rect.GetWidth(), rect.GetHeight());
break;
case SurfaceParams::SurfaceTarget::Texture3D:
case SurfaceParams::SurfaceTarget::Texture2DArray:
- glTexStorage3D(SurfaceTargetToGL(params.target), 1, format_tuple.internal_format,
- rect.GetWidth(), rect.GetHeight(), params.depth);
+ glTexStorage3D(SurfaceTargetToGL(params.target), params.max_mip_level,
+ format_tuple.internal_format, rect.GetWidth(), rect.GetHeight(),
+ params.depth);
break;
default:
LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}",
static_cast<u32>(params.target));
UNREACHABLE();
- glTexStorage2D(GL_TEXTURE_2D, 1, format_tuple.internal_format, rect.GetWidth(),
- rect.GetHeight());
+ glTexStorage2D(GL_TEXTURE_2D, params.max_mip_level, format_tuple.internal_format,
+ rect.GetWidth(), rect.GetHeight());
}
}
glTexParameteri(SurfaceTargetToGL(params.target), GL_TEXTURE_MIN_FILTER, GL_LINEAR);
+ glTexParameteri(SurfaceTargetToGL(params.target), GL_TEXTURE_MAG_FILTER, GL_LINEAR);
glTexParameteri(SurfaceTargetToGL(params.target), GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
glTexParameteri(SurfaceTargetToGL(params.target), GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
+ glTexParameteri(SurfaceTargetToGL(params.target), GL_TEXTURE_MAX_LEVEL,
+ params.max_mip_level - 1);
+ if (params.max_mip_level == 1) {
+ glTexParameterf(SurfaceTargetToGL(params.target), GL_TEXTURE_LOD_BIAS, 1000.0);
+ }
- VideoCore::LabelGLObject(GL_TEXTURE, texture.handle, params.addr,
- SurfaceParams::SurfaceTargetName(params.target));
+ LabelGLObject(GL_TEXTURE, texture.handle, params.addr,
+ SurfaceParams::SurfaceTargetName(params.target));
// Clamp size to mapped GPU memory region
// TODO(bunnei): Super Mario Odyssey maps a 0x40000 byte region and then uses it for a 0x80000
@@ -992,20 +1012,22 @@ static void ConvertFormatAsNeeded_FlushGLBuffer(std::vector<u8>& data, PixelForm
MICROPROFILE_DEFINE(OpenGL_SurfaceLoad, "OpenGL", "Surface Load", MP_RGB(128, 64, 192));
void CachedSurface::LoadGLBuffer() {
MICROPROFILE_SCOPE(OpenGL_SurfaceLoad);
-
- gl_buffer.resize(params.size_in_bytes_gl);
+ gl_buffer.resize(params.max_mip_level);
+ for (u32 i = 0; i < params.max_mip_level; i++)
+ gl_buffer[i].resize(params.GetMipmapSizeGL(i));
if (params.is_tiled) {
ASSERT_MSG(params.block_width == 1, "Block width is defined as {} on texture type {}",
params.block_width, static_cast<u32>(params.target));
-
- SwizzleFunc(morton_to_gl_fns, params, gl_buffer);
+ for (u32 i = 0; i < params.max_mip_level; i++)
+ SwizzleFunc(morton_to_gl_fns, params, gl_buffer[i], i);
} else {
const auto texture_src_data{Memory::GetPointer(params.addr)};
const auto texture_src_data_end{texture_src_data + params.size_in_bytes_gl};
- gl_buffer.assign(texture_src_data, texture_src_data_end);
+ gl_buffer[0].assign(texture_src_data, texture_src_data_end);
}
-
- ConvertFormatAsNeeded_LoadGLBuffer(gl_buffer, params.pixel_format, params.width, params.height);
+ for (u32 i = 0; i < params.max_mip_level; i++)
+ ConvertFormatAsNeeded_LoadGLBuffer(gl_buffer[i], params.pixel_format, params.MipWidth(i),
+ params.MipHeight(i));
}
MICROPROFILE_DEFINE(OpenGL_SurfaceFlush, "OpenGL", "Surface Flush", MP_RGB(128, 192, 64));
@@ -1015,7 +1037,8 @@ void CachedSurface::FlushGLBuffer() {
ASSERT_MSG(!IsPixelFormatASTC(params.pixel_format), "Unimplemented");
// OpenGL temporary buffer needs to be big enough to store raw texture size
- gl_buffer.resize(GetSizeInBytes());
+ gl_buffer.resize(1);
+ gl_buffer[0].resize(GetSizeInBytes());
const FormatTuple& tuple = GetFormatTuple(params.pixel_format, params.component_type);
// Ensure no bad interactions with GL_UNPACK_ALIGNMENT
@@ -1024,9 +1047,9 @@ void CachedSurface::FlushGLBuffer() {
ASSERT(!tuple.compressed);
glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
glGetTextureImage(texture.handle, 0, tuple.format, tuple.type,
- static_cast<GLsizei>(gl_buffer.size()), gl_buffer.data());
+ static_cast<GLsizei>(gl_buffer[0].size()), gl_buffer[0].data());
glPixelStorei(GL_PACK_ROW_LENGTH, 0);
- ConvertFormatAsNeeded_FlushGLBuffer(gl_buffer, params.pixel_format, params.width,
+ ConvertFormatAsNeeded_FlushGLBuffer(gl_buffer[0], params.pixel_format, params.width,
params.height);
ASSERT(params.type != SurfaceType::Fill);
const u8* const texture_src_data = Memory::GetPointer(params.addr);
@@ -1035,26 +1058,21 @@ void CachedSurface::FlushGLBuffer() {
ASSERT_MSG(params.block_width == 1, "Block width is defined as {} on texture type {}",
params.block_width, static_cast<u32>(params.target));
- SwizzleFunc(gl_to_morton_fns, params, gl_buffer);
+ SwizzleFunc(gl_to_morton_fns, params, gl_buffer[0], 0);
} else {
- std::memcpy(Memory::GetPointer(GetAddr()), gl_buffer.data(), GetSizeInBytes());
+ std::memcpy(Memory::GetPointer(GetAddr()), gl_buffer[0].data(), GetSizeInBytes());
}
}
-MICROPROFILE_DEFINE(OpenGL_TextureUL, "OpenGL", "Texture Upload", MP_RGB(128, 64, 192));
-void CachedSurface::UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle) {
- if (params.type == SurfaceType::Fill)
- return;
-
- MICROPROFILE_SCOPE(OpenGL_TextureUL);
-
- const auto& rect{params.GetRect()};
+void CachedSurface::UploadGLMipmapTexture(u32 mip_map, GLuint read_fb_handle,
+ GLuint draw_fb_handle) {
+ const auto& rect{params.GetRect(mip_map)};
// Load data from memory to the surface
const GLint x0 = static_cast<GLint>(rect.left);
const GLint y0 = static_cast<GLint>(rect.bottom);
std::size_t buffer_offset =
- static_cast<std::size_t>(static_cast<std::size_t>(y0) * params.width +
+ static_cast<std::size_t>(static_cast<std::size_t>(y0) * params.MipWidth(mip_map) +
static_cast<std::size_t>(x0)) *
SurfaceParams::GetBytesPerPixel(params.pixel_format);
@@ -1072,88 +1090,117 @@ void CachedSurface::UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle
cur_state.Apply();
// Ensure no bad interactions with GL_UNPACK_ALIGNMENT
- ASSERT(params.width * SurfaceParams::GetBytesPerPixel(params.pixel_format) % 4 == 0);
- glPixelStorei(GL_UNPACK_ROW_LENGTH, static_cast<GLint>(params.width));
+ ASSERT(params.MipWidth(mip_map) * SurfaceParams::GetBytesPerPixel(params.pixel_format) % 4 ==
+ 0);
+ glPixelStorei(GL_UNPACK_ROW_LENGTH, static_cast<GLint>(params.MipWidth(mip_map)));
+ GLsizei image_size = static_cast<GLsizei>(params.GetMipmapSizeGL(mip_map, false));
glActiveTexture(GL_TEXTURE0);
if (tuple.compressed) {
switch (params.target) {
case SurfaceParams::SurfaceTarget::Texture2D:
- glCompressedTexImage2D(
- SurfaceTargetToGL(params.target), 0, tuple.internal_format,
- static_cast<GLsizei>(params.width), static_cast<GLsizei>(params.height), 0,
- static_cast<GLsizei>(params.size_in_bytes_gl), &gl_buffer[buffer_offset]);
+ glCompressedTexImage2D(SurfaceTargetToGL(params.target), mip_map, tuple.internal_format,
+ static_cast<GLsizei>(params.MipWidth(mip_map)),
+ static_cast<GLsizei>(params.MipHeight(mip_map)), 0, image_size,
+ &gl_buffer[mip_map][buffer_offset]);
break;
case SurfaceParams::SurfaceTarget::Texture3D:
+ glCompressedTexImage3D(SurfaceTargetToGL(params.target), mip_map, tuple.internal_format,
+ static_cast<GLsizei>(params.MipWidth(mip_map)),
+ static_cast<GLsizei>(params.MipHeight(mip_map)),
+ static_cast<GLsizei>(params.MipDepth(mip_map)), 0, image_size,
+ &gl_buffer[mip_map][buffer_offset]);
+ break;
case SurfaceParams::SurfaceTarget::Texture2DArray:
- glCompressedTexImage3D(
- SurfaceTargetToGL(params.target), 0, tuple.internal_format,
- static_cast<GLsizei>(params.width), static_cast<GLsizei>(params.height),
- static_cast<GLsizei>(params.depth), 0,
- static_cast<GLsizei>(params.size_in_bytes_gl), &gl_buffer[buffer_offset]);
+ glCompressedTexImage3D(SurfaceTargetToGL(params.target), mip_map, tuple.internal_format,
+ static_cast<GLsizei>(params.MipWidth(mip_map)),
+ static_cast<GLsizei>(params.MipHeight(mip_map)),
+ static_cast<GLsizei>(params.depth), 0, image_size,
+ &gl_buffer[mip_map][buffer_offset]);
break;
- case SurfaceParams::SurfaceTarget::TextureCubemap:
+ case SurfaceParams::SurfaceTarget::TextureCubemap: {
+ GLsizei layer_size = static_cast<GLsizei>(params.LayerSizeGL(mip_map));
for (std::size_t face = 0; face < params.depth; ++face) {
glCompressedTexImage2D(static_cast<GLenum>(GL_TEXTURE_CUBE_MAP_POSITIVE_X + face),
- 0, tuple.internal_format, static_cast<GLsizei>(params.width),
- static_cast<GLsizei>(params.height), 0,
- static_cast<GLsizei>(params.SizeInBytesCubeFaceGL()),
- &gl_buffer[buffer_offset]);
- buffer_offset += params.SizeInBytesCubeFace();
+ mip_map, tuple.internal_format,
+ static_cast<GLsizei>(params.MipWidth(mip_map)),
+ static_cast<GLsizei>(params.MipHeight(mip_map)), 0,
+ layer_size, &gl_buffer[mip_map][buffer_offset]);
+ buffer_offset += layer_size;
}
break;
+ }
default:
LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}",
static_cast<u32>(params.target));
UNREACHABLE();
- glCompressedTexImage2D(
- GL_TEXTURE_2D, 0, tuple.internal_format, static_cast<GLsizei>(params.width),
- static_cast<GLsizei>(params.height), 0,
- static_cast<GLsizei>(params.size_in_bytes_gl), &gl_buffer[buffer_offset]);
+ glCompressedTexImage2D(GL_TEXTURE_2D, mip_map, tuple.internal_format,
+ static_cast<GLsizei>(params.MipWidth(mip_map)),
+ static_cast<GLsizei>(params.MipHeight(mip_map)), 0,
+ static_cast<GLsizei>(params.size_in_bytes_gl),
+ &gl_buffer[mip_map][buffer_offset]);
}
} else {
switch (params.target) {
case SurfaceParams::SurfaceTarget::Texture1D:
- glTexSubImage1D(SurfaceTargetToGL(params.target), 0, x0,
+ glTexSubImage1D(SurfaceTargetToGL(params.target), mip_map, x0,
static_cast<GLsizei>(rect.GetWidth()), tuple.format, tuple.type,
- &gl_buffer[buffer_offset]);
+ &gl_buffer[mip_map][buffer_offset]);
break;
case SurfaceParams::SurfaceTarget::Texture2D:
- glTexSubImage2D(SurfaceTargetToGL(params.target), 0, x0, y0,
+ glTexSubImage2D(SurfaceTargetToGL(params.target), mip_map, x0, y0,
static_cast<GLsizei>(rect.GetWidth()),
static_cast<GLsizei>(rect.GetHeight()), tuple.format, tuple.type,
- &gl_buffer[buffer_offset]);
+ &gl_buffer[mip_map][buffer_offset]);
break;
case SurfaceParams::SurfaceTarget::Texture3D:
+ glTexSubImage3D(SurfaceTargetToGL(params.target), mip_map, x0, y0, 0,
+ static_cast<GLsizei>(rect.GetWidth()),
+ static_cast<GLsizei>(rect.GetHeight()), params.MipDepth(mip_map),
+ tuple.format, tuple.type, &gl_buffer[mip_map][buffer_offset]);
+ break;
case SurfaceParams::SurfaceTarget::Texture2DArray:
- glTexSubImage3D(SurfaceTargetToGL(params.target), 0, x0, y0, 0,
+ glTexSubImage3D(SurfaceTargetToGL(params.target), mip_map, x0, y0, 0,
static_cast<GLsizei>(rect.GetWidth()),
static_cast<GLsizei>(rect.GetHeight()), params.depth, tuple.format,
- tuple.type, &gl_buffer[buffer_offset]);
+ tuple.type, &gl_buffer[mip_map][buffer_offset]);
break;
- case SurfaceParams::SurfaceTarget::TextureCubemap:
+ case SurfaceParams::SurfaceTarget::TextureCubemap: {
+ std::size_t start = buffer_offset;
for (std::size_t face = 0; face < params.depth; ++face) {
- glTexSubImage2D(static_cast<GLenum>(GL_TEXTURE_CUBE_MAP_POSITIVE_X + face), 0, x0,
- y0, static_cast<GLsizei>(rect.GetWidth()),
+ glTexSubImage2D(static_cast<GLenum>(GL_TEXTURE_CUBE_MAP_POSITIVE_X + face), mip_map,
+ x0, y0, static_cast<GLsizei>(rect.GetWidth()),
static_cast<GLsizei>(rect.GetHeight()), tuple.format, tuple.type,
- &gl_buffer[buffer_offset]);
- buffer_offset += params.SizeInBytesCubeFace();
+ &gl_buffer[mip_map][buffer_offset]);
+ buffer_offset += params.LayerSizeGL(mip_map);
}
break;
+ }
default:
LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}",
static_cast<u32>(params.target));
UNREACHABLE();
- glTexSubImage2D(GL_TEXTURE_2D, 0, x0, y0, static_cast<GLsizei>(rect.GetWidth()),
+ glTexSubImage2D(GL_TEXTURE_2D, mip_map, x0, y0, static_cast<GLsizei>(rect.GetWidth()),
static_cast<GLsizei>(rect.GetHeight()), tuple.format, tuple.type,
- &gl_buffer[buffer_offset]);
+ &gl_buffer[mip_map][buffer_offset]);
}
}
glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
}
+MICROPROFILE_DEFINE(OpenGL_TextureUL, "OpenGL", "Texture Upload", MP_RGB(128, 64, 192));
+void CachedSurface::UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle) {
+ if (params.type == SurfaceType::Fill)
+ return;
+
+ MICROPROFILE_SCOPE(OpenGL_TextureUL);
+
+ for (u32 i = 0; i < params.max_mip_level; i++)
+ UploadGLMipmapTexture(i, read_fb_handle, draw_fb_handle);
+}
+
RasterizerCacheOpenGL::RasterizerCacheOpenGL() {
read_framebuffer.Create();
draw_framebuffer.Create();
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.h b/src/video_core/renderer_opengl/gl_rasterizer_cache.h
index be8c00e99..e72f4f2d2 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.h
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.h
@@ -404,7 +404,7 @@ struct SurfaceParams {
128, // BC7U
32, // ASTC_2D_4X4_SRGB
16, // ASTC_2D_8X8_SRGB
- 32, // ASTC_2D_8X5_SRGB
+ 16, // ASTC_2D_8X5_SRGB
32, // ASTC_2D_5X4_SRGB
32, // Z32F
16, // Z16
@@ -834,7 +834,7 @@ struct SurfaceParams {
}
/// Returns the rectangle corresponding to this surface
- MathUtil::Rectangle<u32> GetRect() const;
+ MathUtil::Rectangle<u32> GetRect(u32 mip_level = 0) const;
/// Returns the total size of this surface in bytes, adjusted for compression
std::size_t SizeInBytesRaw(bool ignore_tiled = false) const {
@@ -865,7 +865,7 @@ struct SurfaceParams {
/// Returns the exact size of memory occupied by the texture in VRAM, including mipmaps.
std::size_t MemorySize() const {
- std::size_t size = InnerMemorySize(is_layered);
+ std::size_t size = InnerMemorySize(false, is_layered);
if (is_layered)
return size * depth;
return size;
@@ -874,12 +874,78 @@ struct SurfaceParams {
/// Returns the exact size of the memory occupied by a layer in a texture in VRAM, including
/// mipmaps.
std::size_t LayerMemorySize() const {
- return InnerMemorySize(true);
+ return InnerMemorySize(false, true);
}
/// Returns the size of a layer of this surface in OpenGL.
- std::size_t LayerSizeGL() const {
- return SizeInBytesRaw(true) / depth;
+ std::size_t LayerSizeGL(u32 mip_level) const {
+ return InnerMipmapMemorySize(mip_level, true, is_layered, false);
+ }
+
+ std::size_t GetMipmapSizeGL(u32 mip_level, bool ignore_compressed = true) const {
+ std::size_t size = InnerMipmapMemorySize(mip_level, true, is_layered, ignore_compressed);
+ if (is_layered)
+ return size * depth;
+ return size;
+ }
+
+ std::size_t GetMipmapLevelOffset(u32 mip_level) const {
+ std::size_t offset = 0;
+ for (u32 i = 0; i < mip_level; i++)
+ offset += InnerMipmapMemorySize(i, false, is_layered);
+ return offset;
+ }
+
+ std::size_t GetMipmapLevelOffsetGL(u32 mip_level) const {
+ std::size_t offset = 0;
+ for (u32 i = 0; i < mip_level; i++)
+ offset += InnerMipmapMemorySize(i, true, is_layered);
+ return offset;
+ }
+
+ u32 MipWidth(u32 mip_level) const {
+ return std::max(1U, width >> mip_level);
+ }
+
+ u32 MipHeight(u32 mip_level) const {
+ return std::max(1U, height >> mip_level);
+ }
+
+ u32 MipDepth(u32 mip_level) const {
+ return std::max(1U, depth >> mip_level);
+ }
+
+ // Auto block resizing algorithm from:
+ // https://cgit.freedesktop.org/mesa/mesa/tree/src/gallium/drivers/nouveau/nv50/nv50_miptree.c
+ u32 MipBlockHeight(u32 mip_level) const {
+ if (mip_level == 0)
+ return block_height;
+ u32 alt_height = MipHeight(mip_level);
+ u32 h = GetDefaultBlockHeight(pixel_format);
+ u32 blocks_in_y = (alt_height + h - 1) / h;
+ u32 bh = 16;
+ while (bh > 1 && blocks_in_y <= bh * 4) {
+ bh >>= 1;
+ }
+ return bh;
+ }
+
+ u32 MipBlockDepth(u32 mip_level) const {
+ if (mip_level == 0)
+ return block_depth;
+ if (is_layered)
+ return 1;
+ u32 depth = MipDepth(mip_level);
+ u32 bd = 32;
+ while (bd > 1 && depth * 2 <= bd) {
+ bd >>= 1;
+ }
+ if (bd == 32) {
+ u32 bh = MipBlockHeight(mip_level);
+ if (bh >= 4)
+ return 16;
+ }
+ return bd;
}
/// Creates SurfaceParams from a texture configuration
@@ -940,7 +1006,10 @@ struct SurfaceParams {
} rt;
private:
- std::size_t InnerMemorySize(bool layer_only = false) const;
+ std::size_t InnerMipmapMemorySize(u32 mip_level, bool force_gl = false, bool layer_only = false,
+ bool uncompressed = false) const;
+ std::size_t InnerMemorySize(bool force_gl = false, bool layer_only = false,
+ bool uncompressed = false) const;
};
}; // namespace OpenGL
@@ -1002,8 +1071,10 @@ public:
void UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle);
private:
+ void UploadGLMipmapTexture(u32 mip_map, GLuint read_fb_handle, GLuint draw_fb_handle);
+
OGLTexture texture;
- std::vector<u8> gl_buffer;
+ std::vector<std::vector<u8>> gl_buffer;
SurfaceParams params;
GLenum gl_target;
std::size_t cached_size_in_bytes;
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp
index 1a03a677f..9522fd344 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp
@@ -8,6 +8,7 @@
#include "video_core/engines/maxwell_3d.h"
#include "video_core/renderer_opengl/gl_shader_cache.h"
#include "video_core/renderer_opengl/gl_shader_manager.h"
+#include "video_core/renderer_opengl/utils.h"
#include "video_core/utils.h"
namespace OpenGL {
@@ -89,7 +90,7 @@ CachedShader::CachedShader(VAddr addr, Maxwell::ShaderProgram program_type)
shader.Create(program_result.first.c_str(), gl_type);
program.Create(true, shader.handle);
SetShaderUniformBlockBindings(program.handle);
- VideoCore::LabelGLObject(GL_PROGRAM, program.handle, addr);
+ LabelGLObject(GL_PROGRAM, program.handle, addr);
} else {
// Store shader's code to lazily build it on draw
geometry_programs.code = program_result.first;
@@ -130,7 +131,7 @@ GLuint CachedShader::LazyGeometryProgram(OGLProgram& target_program,
shader.Create(source.c_str(), GL_GEOMETRY_SHADER);
target_program.Create(true, shader.handle);
SetShaderUniformBlockBindings(target_program.handle);
- VideoCore::LabelGLObject(GL_PROGRAM, target_program.handle, addr, debug_name);
+ LabelGLObject(GL_PROGRAM, target_program.handle, addr, debug_name);
return target_program.handle;
};
diff --git a/src/video_core/renderer_opengl/maxwell_to_gl.h b/src/video_core/renderer_opengl/maxwell_to_gl.h
index 0f6dcab2b..87d511c38 100644
--- a/src/video_core/renderer_opengl/maxwell_to_gl.h
+++ b/src/video_core/renderer_opengl/maxwell_to_gl.h
@@ -135,12 +135,29 @@ inline GLenum PrimitiveTopology(Maxwell::PrimitiveTopology topology) {
return {};
}
-inline GLenum TextureFilterMode(Tegra::Texture::TextureFilter filter_mode) {
+inline GLenum TextureFilterMode(Tegra::Texture::TextureFilter filter_mode,
+ Tegra::Texture::TextureMipmapFilter mip_filter_mode) {
switch (filter_mode) {
- case Tegra::Texture::TextureFilter::Linear:
- return GL_LINEAR;
- case Tegra::Texture::TextureFilter::Nearest:
- return GL_NEAREST;
+ case Tegra::Texture::TextureFilter::Linear: {
+ switch (mip_filter_mode) {
+ case Tegra::Texture::TextureMipmapFilter::None:
+ return GL_LINEAR;
+ case Tegra::Texture::TextureMipmapFilter::Nearest:
+ return GL_NEAREST_MIPMAP_LINEAR;
+ case Tegra::Texture::TextureMipmapFilter::Linear:
+ return GL_LINEAR_MIPMAP_LINEAR;
+ }
+ }
+ case Tegra::Texture::TextureFilter::Nearest: {
+ switch (mip_filter_mode) {
+ case Tegra::Texture::TextureMipmapFilter::None:
+ return GL_NEAREST;
+ case Tegra::Texture::TextureMipmapFilter::Nearest:
+ return GL_NEAREST_MIPMAP_NEAREST;
+ case Tegra::Texture::TextureMipmapFilter::Linear:
+ return GL_LINEAR_MIPMAP_NEAREST;
+ }
+ }
}
LOG_CRITICAL(Render_OpenGL, "Unimplemented texture filter mode={}",
static_cast<u32>(filter_mode));
diff --git a/src/video_core/renderer_opengl/utils.cpp b/src/video_core/renderer_opengl/utils.cpp
new file mode 100644
index 000000000..d84634cb3
--- /dev/null
+++ b/src/video_core/renderer_opengl/utils.cpp
@@ -0,0 +1,38 @@
+// Copyright 2014 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <string>
+#include <fmt/format.h>
+#include <glad/glad.h>
+#include "common/common_types.h"
+#include "video_core/renderer_opengl/utils.h"
+
+namespace OpenGL {
+
+void LabelGLObject(GLenum identifier, GLuint handle, VAddr addr, std::string extra_info) {
+ if (!GLAD_GL_KHR_debug) {
+ return; // We don't need to throw an error as this is just for debugging
+ }
+ const std::string nice_addr = fmt::format("0x{:016x}", addr);
+ std::string object_label;
+
+ if (extra_info.empty()) {
+ switch (identifier) {
+ case GL_TEXTURE:
+ object_label = "Texture@" + nice_addr;
+ break;
+ case GL_PROGRAM:
+ object_label = "Shader@" + nice_addr;
+ break;
+ default:
+ object_label = fmt::format("Object(0x{:x})@{}", identifier, nice_addr);
+ break;
+ }
+ } else {
+ object_label = extra_info + '@' + nice_addr;
+ }
+ glObjectLabel(identifier, handle, -1, static_cast<const GLchar*>(object_label.c_str()));
+}
+
+} // namespace OpenGL \ No newline at end of file
diff --git a/src/video_core/renderer_opengl/utils.h b/src/video_core/renderer_opengl/utils.h
new file mode 100644
index 000000000..1fcb6fc11
--- /dev/null
+++ b/src/video_core/renderer_opengl/utils.h
@@ -0,0 +1,15 @@
+// Copyright 2014 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <string>
+#include <glad/glad.h>
+#include "common/common_types.h"
+
+namespace OpenGL {
+
+void LabelGLObject(GLenum identifier, GLuint handle, VAddr addr, std::string extra_info = "");
+
+} // namespace OpenGL \ No newline at end of file
diff --git a/src/video_core/textures/decoders.h b/src/video_core/textures/decoders.h
index 4726f54a5..b390219e4 100644
--- a/src/video_core/textures/decoders.h
+++ b/src/video_core/textures/decoders.h
@@ -10,6 +10,12 @@
namespace Tegra::Texture {
+// GOBSize constant. Calculated by 64 bytes in x multiplied by 8 y coords, represents
+// an small rect of (64/bytes_per_pixel)X8.
+inline std::size_t GetGOBSize() {
+ return 512;
+}
+
/**
* Unswizzles a swizzled texture without changing its format.
*/
diff --git a/src/video_core/utils.h b/src/video_core/utils.h
index 237cc1307..e0a14d48f 100644
--- a/src/video_core/utils.h
+++ b/src/video_core/utils.h
@@ -161,30 +161,4 @@ static inline void MortonCopyPixels128(u32 width, u32 height, u32 bytes_per_pixe
}
}
-static void LabelGLObject(GLenum identifier, GLuint handle, VAddr addr,
- std::string extra_info = "") {
- if (!GLAD_GL_KHR_debug) {
- return; // We don't need to throw an error as this is just for debugging
- }
- const std::string nice_addr = fmt::format("0x{:016x}", addr);
- std::string object_label;
-
- if (extra_info.empty()) {
- switch (identifier) {
- case GL_TEXTURE:
- object_label = "Texture@" + nice_addr;
- break;
- case GL_PROGRAM:
- object_label = "Shader@" + nice_addr;
- break;
- default:
- object_label = fmt::format("Object(0x{:x})@{}", identifier, nice_addr);
- break;
- }
- } else {
- object_label = extra_info + '@' + nice_addr;
- }
- glObjectLabel(identifier, handle, -1, static_cast<const GLchar*>(object_label.c_str()));
-}
-
} // namespace VideoCore
generated by cgit v1.2.3 (git 2.25.1) at 2024-10-30 18:01:10 +0100