summaryrefslogtreecommitdiffstats
path: root/src/video_core/renderer_opengl
diff options
context:
space:
mode:
Diffstat (limited to '')
-rw-r--r--src/video_core/renderer_opengl/gl_arb_decompiler.cpp63
-rw-r--r--src/video_core/renderer_opengl/gl_buffer_cache.cpp62
-rw-r--r--src/video_core/renderer_opengl/gl_buffer_cache.h36
-rw-r--r--src/video_core/renderer_opengl/gl_device.cpp50
-rw-r--r--src/video_core/renderer_opengl/gl_device.h10
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer.cpp70
-rw-r--r--src/video_core/renderer_opengl/gl_shader_cache.cpp12
-rw-r--r--src/video_core/renderer_opengl/gl_shader_cache.h1
-rw-r--r--src/video_core/renderer_opengl/gl_shader_decompiler.cpp50
-rw-r--r--src/video_core/renderer_opengl/gl_stream_buffer.cpp56
-rw-r--r--src/video_core/renderer_opengl/gl_stream_buffer.h16
-rw-r--r--src/video_core/renderer_opengl/maxwell_to_gl.h82
-rw-r--r--src/video_core/renderer_opengl/renderer_opengl.cpp17
-rw-r--r--src/video_core/renderer_opengl/renderer_opengl.h3
14 files changed, 310 insertions, 218 deletions
diff --git a/src/video_core/renderer_opengl/gl_arb_decompiler.cpp b/src/video_core/renderer_opengl/gl_arb_decompiler.cpp
index 1e96b0310..eb5158407 100644
--- a/src/video_core/renderer_opengl/gl_arb_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_arb_decompiler.cpp
@@ -281,14 +281,14 @@ private:
template <const std::string_view& op>
std::string Unary(Operation operation) {
- const std::string temporary = AllocTemporary();
+ std::string temporary = AllocTemporary();
AddLine("{}{} {}, {};", op, Modifiers(operation), temporary, Visit(operation[0]));
return temporary;
}
template <const std::string_view& op>
std::string Binary(Operation operation) {
- const std::string temporary = AllocTemporary();
+ std::string temporary = AllocTemporary();
AddLine("{}{} {}, {}, {};", op, Modifiers(operation), temporary, Visit(operation[0]),
Visit(operation[1]));
return temporary;
@@ -296,7 +296,7 @@ private:
template <const std::string_view& op>
std::string Trinary(Operation operation) {
- const std::string temporary = AllocTemporary();
+ std::string temporary = AllocTemporary();
AddLine("{}{} {}, {}, {}, {};", op, Modifiers(operation), temporary, Visit(operation[0]),
Visit(operation[1]), Visit(operation[2]));
return temporary;
@@ -304,7 +304,7 @@ private:
template <const std::string_view& op, bool unordered>
std::string FloatComparison(Operation operation) {
- const std::string temporary = AllocTemporary();
+ std::string temporary = AllocTemporary();
AddLine("TRUNC.U.CC RC.x, {};", Binary<op>(operation));
AddLine("MOV.S {}, 0;", temporary);
AddLine("MOV.S {} (NE.x), -1;", temporary);
@@ -331,7 +331,7 @@ private:
template <const std::string_view& op, bool is_nan>
std::string HalfComparison(Operation operation) {
- const std::string tmp1 = AllocVectorTemporary();
+ std::string tmp1 = AllocVectorTemporary();
const std::string tmp2 = AllocVectorTemporary();
const std::string op_a = Visit(operation[0]);
const std::string op_b = Visit(operation[1]);
@@ -367,15 +367,14 @@ private:
AddLine("MOV.F {}.{}, {};", value, Swizzle(i), Visit(meta.values[i]));
}
- const std::string result = coord;
- AddLine("ATOMIM.{}.{} {}.x, {}, {}, image[{}], {};", op, type, result, value, coord,
+ AddLine("ATOMIM.{}.{} {}.x, {}, {}, image[{}], {};", op, type, coord, value, coord,
image_id, ImageType(meta.image.type));
- return fmt::format("{}.x", result);
+ return fmt::format("{}.x", coord);
}
template <const std::string_view& op, const std::string_view& type>
std::string Atomic(Operation operation) {
- const std::string temporary = AllocTemporary();
+ std::string temporary = AllocTemporary();
std::string address;
std::string_view opname;
if (const auto gmem = std::get_if<GmemNode>(&*operation[0])) {
@@ -396,7 +395,7 @@ private:
template <char type>
std::string Negate(Operation operation) {
- const std::string temporary = AllocTemporary();
+ std::string temporary = AllocTemporary();
if constexpr (type == 'F') {
AddLine("MOV.F32 {}, -{};", temporary, Visit(operation[0]));
} else {
@@ -407,7 +406,7 @@ private:
template <char type>
std::string Absolute(Operation operation) {
- const std::string temporary = AllocTemporary();
+ std::string temporary = AllocTemporary();
AddLine("MOV.{} {}, |{}|;", type, temporary, Visit(operation[0]));
return temporary;
}
@@ -1156,20 +1155,20 @@ void ARBDecompiler::VisitAST(const ASTNode& node) {
}
std::string ARBDecompiler::VisitExpression(const Expr& node) {
- const std::string result = AllocTemporary();
if (const auto expr = std::get_if<ExprAnd>(&*node)) {
+ std::string result = AllocTemporary();
AddLine("AND.U {}, {}, {};", result, VisitExpression(expr->operand1),
VisitExpression(expr->operand2));
return result;
}
if (const auto expr = std::get_if<ExprOr>(&*node)) {
- const std::string result = AllocTemporary();
+ std::string result = AllocTemporary();
AddLine("OR.U {}, {}, {};", result, VisitExpression(expr->operand1),
VisitExpression(expr->operand2));
return result;
}
if (const auto expr = std::get_if<ExprNot>(&*node)) {
- const std::string result = AllocTemporary();
+ std::string result = AllocTemporary();
AddLine("CMP.S {}, {}, 0, -1;", result, VisitExpression(expr->operand1));
return result;
}
@@ -1186,7 +1185,7 @@ std::string ARBDecompiler::VisitExpression(const Expr& node) {
return expr->value ? "0xffffffff" : "0";
}
if (const auto expr = std::get_if<ExprGprEqual>(&*node)) {
- const std::string result = AllocTemporary();
+ std::string result = AllocTemporary();
AddLine("SEQ.U {}, R{}.x, {};", result, expr->gpr, expr->value);
return result;
}
@@ -1231,13 +1230,13 @@ std::string ARBDecompiler::Visit(const Node& node) {
}
if (const auto immediate = std::get_if<ImmediateNode>(&*node)) {
- const std::string temporary = AllocTemporary();
+ std::string temporary = AllocTemporary();
AddLine("MOV.U {}, {};", temporary, immediate->GetValue());
return temporary;
}
if (const auto predicate = std::get_if<PredicateNode>(&*node)) {
- const std::string temporary = AllocTemporary();
+ std::string temporary = AllocTemporary();
switch (const auto index = predicate->GetIndex(); index) {
case Tegra::Shader::Pred::UnusedIndex:
AddLine("MOV.S {}, -1;", temporary);
@@ -1333,13 +1332,13 @@ std::string ARBDecompiler::Visit(const Node& node) {
} else {
offset_string = Visit(offset);
}
- const std::string temporary = AllocTemporary();
+ std::string temporary = AllocTemporary();
AddLine("LDC.F32 {}, cbuf{}[{}];", temporary, cbuf->GetIndex(), offset_string);
return temporary;
}
if (const auto gmem = std::get_if<GmemNode>(&*node)) {
- const std::string temporary = AllocTemporary();
+ std::string temporary = AllocTemporary();
AddLine("SUB.U {}, {}, {};", temporary, Visit(gmem->GetRealAddress()),
Visit(gmem->GetBaseAddress()));
AddLine("LDB.U32 {}, {}[{}];", temporary, GlobalMemoryName(gmem->GetDescriptor()),
@@ -1348,14 +1347,14 @@ std::string ARBDecompiler::Visit(const Node& node) {
}
if (const auto lmem = std::get_if<LmemNode>(&*node)) {
- const std::string temporary = Visit(lmem->GetAddress());
+ std::string temporary = Visit(lmem->GetAddress());
AddLine("SHR.U {}, {}, 2;", temporary, temporary);
AddLine("MOV.U {}, lmem[{}].x;", temporary, temporary);
return temporary;
}
if (const auto smem = std::get_if<SmemNode>(&*node)) {
- const std::string temporary = Visit(smem->GetAddress());
+ std::string temporary = Visit(smem->GetAddress());
AddLine("LDS.U32 {}, shared_mem[{}];", temporary, temporary);
return temporary;
}
@@ -1535,7 +1534,7 @@ std::string ARBDecompiler::Assign(Operation operation) {
}
std::string ARBDecompiler::Select(Operation operation) {
- const std::string temporary = AllocTemporary();
+ std::string temporary = AllocTemporary();
AddLine("CMP.S {}, {}, {}, {};", temporary, Visit(operation[0]), Visit(operation[1]),
Visit(operation[2]));
return temporary;
@@ -1545,12 +1544,12 @@ std::string ARBDecompiler::FClamp(Operation operation) {
// 1.0f in hex, replace with std::bit_cast on C++20
static constexpr u32 POSITIVE_ONE = 0x3f800000;
- const std::string temporary = AllocTemporary();
+ std::string temporary = AllocTemporary();
const Node& value = operation[0];
const Node& low = operation[1];
const Node& high = operation[2];
- const auto imm_low = std::get_if<ImmediateNode>(&*low);
- const auto imm_high = std::get_if<ImmediateNode>(&*high);
+ const auto* const imm_low = std::get_if<ImmediateNode>(&*low);
+ const auto* const imm_high = std::get_if<ImmediateNode>(&*high);
if (imm_low && imm_high && imm_low->GetValue() == 0 && imm_high->GetValue() == POSITIVE_ONE) {
AddLine("MOV.F32.SAT {}, {};", temporary, Visit(value));
} else {
@@ -1574,7 +1573,7 @@ std::string ARBDecompiler::FCastHalf1(Operation operation) {
}
std::string ARBDecompiler::FSqrt(Operation operation) {
- const std::string temporary = AllocTemporary();
+ std::string temporary = AllocTemporary();
AddLine("RSQ.F32 {}, {};", temporary, Visit(operation[0]));
AddLine("RCP.F32 {}, {};", temporary, temporary);
return temporary;
@@ -1588,7 +1587,7 @@ std::string ARBDecompiler::FSwizzleAdd(Operation operation) {
AddLine("ADD.F {}.x, {}, {};", temporary, Visit(operation[0]), Visit(operation[1]));
return fmt::format("{}.x", temporary);
}
- const std::string lut = AllocVectorTemporary();
+
AddLine("AND.U {}.z, {}.threadid, 3;", temporary, StageInputName(stage));
AddLine("SHL.U {}.z, {}.z, 1;", temporary, temporary);
AddLine("SHR.U {}.z, {}, {}.z;", temporary, Visit(operation[2]), temporary);
@@ -1766,21 +1765,21 @@ std::string ARBDecompiler::LogicalAssign(Operation operation) {
}
std::string ARBDecompiler::LogicalPick2(Operation operation) {
- const std::string temporary = AllocTemporary();
+ std::string temporary = AllocTemporary();
const u32 index = std::get<ImmediateNode>(*operation[1]).GetValue();
AddLine("MOV.U {}, {}.{};", temporary, Visit(operation[0]), Swizzle(index));
return temporary;
}
std::string ARBDecompiler::LogicalAnd2(Operation operation) {
- const std::string temporary = AllocTemporary();
+ std::string temporary = AllocTemporary();
const std::string op = Visit(operation[0]);
AddLine("AND.U {}, {}.x, {}.y;", temporary, op, op);
return temporary;
}
std::string ARBDecompiler::FloatOrdered(Operation operation) {
- const std::string temporary = AllocTemporary();
+ std::string temporary = AllocTemporary();
AddLine("MOVC.F32 RC.x, {};", Visit(operation[0]));
AddLine("MOVC.F32 RC.y, {};", Visit(operation[1]));
AddLine("MOV.S {}, -1;", temporary);
@@ -1790,7 +1789,7 @@ std::string ARBDecompiler::FloatOrdered(Operation operation) {
}
std::string ARBDecompiler::FloatUnordered(Operation operation) {
- const std::string temporary = AllocTemporary();
+ std::string temporary = AllocTemporary();
AddLine("MOVC.F32 RC.x, {};", Visit(operation[0]));
AddLine("MOVC.F32 RC.y, {};", Visit(operation[1]));
AddLine("MOV.S {}, 0;", temporary);
@@ -1800,7 +1799,7 @@ std::string ARBDecompiler::FloatUnordered(Operation operation) {
}
std::string ARBDecompiler::LogicalAddCarry(Operation operation) {
- const std::string temporary = AllocTemporary();
+ std::string temporary = AllocTemporary();
AddLine("ADDC.U RC, {}, {};", Visit(operation[0]), Visit(operation[1]));
AddLine("MOV.S {}, 0;", temporary);
AddLine("IF CF.x;");
diff --git a/src/video_core/renderer_opengl/gl_buffer_cache.cpp b/src/video_core/renderer_opengl/gl_buffer_cache.cpp
index ad0577a4f..d9f7b4cc6 100644
--- a/src/video_core/renderer_opengl/gl_buffer_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_buffer_cache.cpp
@@ -22,21 +22,46 @@ using Maxwell = Tegra::Engines::Maxwell3D::Regs;
MICROPROFILE_DEFINE(OpenGL_Buffer_Download, "OpenGL", "Buffer Download", MP_RGB(192, 192, 128));
-Buffer::Buffer(VAddr cpu_addr, const std::size_t size) : VideoCommon::BufferBlock{cpu_addr, size} {
+Buffer::Buffer(const Device& device, VAddr cpu_addr, std::size_t size)
+ : VideoCommon::BufferBlock{cpu_addr, size} {
gl_buffer.Create();
glNamedBufferData(gl_buffer.handle, static_cast<GLsizeiptr>(size), nullptr, GL_DYNAMIC_DRAW);
+ if (device.HasVertexBufferUnifiedMemory()) {
+ glMakeNamedBufferResidentNV(gl_buffer.handle, GL_READ_WRITE);
+ glGetNamedBufferParameterui64vNV(gl_buffer.handle, GL_BUFFER_GPU_ADDRESS_NV, &gpu_address);
+ }
}
Buffer::~Buffer() = default;
+void Buffer::Upload(std::size_t offset, std::size_t size, const u8* data) const {
+ glNamedBufferSubData(Handle(), static_cast<GLintptr>(offset), static_cast<GLsizeiptr>(size),
+ data);
+}
+
+void Buffer::Download(std::size_t offset, std::size_t size, u8* data) const {
+ MICROPROFILE_SCOPE(OpenGL_Buffer_Download);
+ glMemoryBarrier(GL_BUFFER_UPDATE_BARRIER_BIT);
+ glGetNamedBufferSubData(Handle(), static_cast<GLintptr>(offset), static_cast<GLsizeiptr>(size),
+ data);
+}
+
+void Buffer::CopyFrom(const Buffer& src, std::size_t src_offset, std::size_t dst_offset,
+ std::size_t size) const {
+ glCopyNamedBufferSubData(src.Handle(), Handle(), static_cast<GLintptr>(src_offset),
+ static_cast<GLintptr>(dst_offset), static_cast<GLsizeiptr>(size));
+}
+
OGLBufferCache::OGLBufferCache(RasterizerOpenGL& rasterizer, Core::System& system,
- const Device& device, std::size_t stream_size)
- : GenericBufferCache{rasterizer, system, std::make_unique<OGLStreamBuffer>(stream_size, true)} {
+ const Device& device_, std::size_t stream_size)
+ : GenericBufferCache{rasterizer, system,
+ std::make_unique<OGLStreamBuffer>(device_, stream_size, true)},
+ device{device_} {
if (!device.HasFastBufferSubData()) {
return;
}
- static constexpr auto size = static_cast<GLsizeiptr>(Maxwell::MaxConstBufferSize);
+ static constexpr GLsizeiptr size = static_cast<GLsizeiptr>(Maxwell::MaxConstBufferSize);
glCreateBuffers(static_cast<GLsizei>(std::size(cbufs)), std::data(cbufs));
for (const GLuint cbuf : cbufs) {
glNamedBufferData(cbuf, size, nullptr, GL_STREAM_DRAW);
@@ -48,39 +73,20 @@ OGLBufferCache::~OGLBufferCache() {
}
std::shared_ptr<Buffer> OGLBufferCache::CreateBlock(VAddr cpu_addr, std::size_t size) {
- return std::make_shared<Buffer>(cpu_addr, size);
+ return std::make_shared<Buffer>(device, cpu_addr, size);
}
-GLuint OGLBufferCache::GetEmptyBuffer(std::size_t) {
- return 0;
-}
-
-void OGLBufferCache::UploadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
- const u8* data) {
- glNamedBufferSubData(buffer.Handle(), static_cast<GLintptr>(offset),
- static_cast<GLsizeiptr>(size), data);
-}
-
-void OGLBufferCache::DownloadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
- u8* data) {
- MICROPROFILE_SCOPE(OpenGL_Buffer_Download);
- glMemoryBarrier(GL_BUFFER_UPDATE_BARRIER_BIT);
- glGetNamedBufferSubData(buffer.Handle(), static_cast<GLintptr>(offset),
- static_cast<GLsizeiptr>(size), data);
-}
-
-void OGLBufferCache::CopyBlock(const Buffer& src, const Buffer& dst, std::size_t src_offset,
- std::size_t dst_offset, std::size_t size) {
- glCopyNamedBufferSubData(src.Handle(), dst.Handle(), static_cast<GLintptr>(src_offset),
- static_cast<GLintptr>(dst_offset), static_cast<GLsizeiptr>(size));
+OGLBufferCache::BufferInfo OGLBufferCache::GetEmptyBuffer(std::size_t) {
+ return {0, 0, 0};
}
OGLBufferCache::BufferInfo OGLBufferCache::ConstBufferUpload(const void* raw_pointer,
std::size_t size) {
DEBUG_ASSERT(cbuf_cursor < std::size(cbufs));
const GLuint cbuf = cbufs[cbuf_cursor++];
+
glNamedBufferSubData(cbuf, 0, static_cast<GLsizeiptr>(size), raw_pointer);
- return {cbuf, 0};
+ return {cbuf, 0, 0};
}
} // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_buffer_cache.h b/src/video_core/renderer_opengl/gl_buffer_cache.h
index a49aaf9c4..59d95adbc 100644
--- a/src/video_core/renderer_opengl/gl_buffer_cache.h
+++ b/src/video_core/renderer_opengl/gl_buffer_cache.h
@@ -25,15 +25,27 @@ class RasterizerOpenGL;
class Buffer : public VideoCommon::BufferBlock {
public:
- explicit Buffer(VAddr cpu_addr, const std::size_t size);
+ explicit Buffer(const Device& device, VAddr cpu_addr, std::size_t size);
~Buffer();
- GLuint Handle() const {
+ void Upload(std::size_t offset, std::size_t size, const u8* data) const;
+
+ void Download(std::size_t offset, std::size_t size, u8* data) const;
+
+ void CopyFrom(const Buffer& src, std::size_t src_offset, std::size_t dst_offset,
+ std::size_t size) const;
+
+ GLuint Handle() const noexcept {
return gl_buffer.handle;
}
+ u64 Address() const noexcept {
+ return gpu_address;
+ }
+
private:
OGLBuffer gl_buffer;
+ u64 gpu_address = 0;
};
using GenericBufferCache = VideoCommon::BufferCache<Buffer, GLuint, OGLStreamBuffer>;
@@ -43,7 +55,7 @@ public:
const Device& device, std::size_t stream_size);
~OGLBufferCache();
- GLuint GetEmptyBuffer(std::size_t) override;
+ BufferInfo GetEmptyBuffer(std::size_t) override;
void Acquire() noexcept {
cbuf_cursor = 0;
@@ -52,22 +64,16 @@ public:
protected:
std::shared_ptr<Buffer> CreateBlock(VAddr cpu_addr, std::size_t size) override;
- void UploadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
- const u8* data) override;
-
- void DownloadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
- u8* data) override;
-
- void CopyBlock(const Buffer& src, const Buffer& dst, std::size_t src_offset,
- std::size_t dst_offset, std::size_t size) override;
-
BufferInfo ConstBufferUpload(const void* raw_pointer, std::size_t size) override;
private:
+ static constexpr std::size_t NUM_CBUFS = Tegra::Engines::Maxwell3D::Regs::MaxConstBuffers *
+ Tegra::Engines::Maxwell3D::Regs::MaxShaderProgram;
+
+ const Device& device;
+
std::size_t cbuf_cursor = 0;
- std::array<GLuint, Tegra::Engines::Maxwell3D::Regs::MaxConstBuffers *
- Tegra::Engines::Maxwell3D::Regs::MaxShaderProgram>
- cbufs;
+ std::array<GLuint, NUM_CBUFS> cbufs{};
};
} // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_device.cpp b/src/video_core/renderer_opengl/gl_device.cpp
index e245e27ec..b6b6659c1 100644
--- a/src/video_core/renderer_opengl/gl_device.cpp
+++ b/src/video_core/renderer_opengl/gl_device.cpp
@@ -123,16 +123,24 @@ std::array<Device::BaseBindings, Tegra::Engines::MaxShaderTypes> BuildBaseBindin
u32 num_images = GetInteger<u32>(GL_MAX_IMAGE_UNITS);
u32 base_images = 0;
- // Reserve more image bindings on fragment and vertex stages.
+ // GL_MAX_IMAGE_UNITS is guaranteed by the spec to have a minimum value of 8.
+ // Due to the limitation of GL_MAX_IMAGE_UNITS, reserve at least 4 image bindings on the
+ // fragment stage, and at least 1 for the rest of the stages.
+ // So far games are observed to use 1 image binding on vertex and 4 on fragment stages.
+
+ // Reserve at least 4 image bindings on the fragment stage.
bindings[4].image =
- Extract(base_images, num_images, num_images / NumStages + 2, LimitImages[4]);
- bindings[0].image =
- Extract(base_images, num_images, num_images / NumStages + 1, LimitImages[0]);
+ Extract(base_images, num_images, std::max(4U, num_images / NumStages), LimitImages[4]);
+
+ // This is guaranteed to be at least 1.
+ const u32 total_extracted_images = num_images / (NumStages - 1);
// Reserve the other image bindings.
- const u32 total_extracted_images = num_images / (NumStages - 2);
- for (std::size_t i = 2; i < NumStages; ++i) {
+ for (std::size_t i = 0; i < NumStages; ++i) {
const std::size_t stage = stage_swizzle[i];
+ if (stage == 4) {
+ continue;
+ }
bindings[stage].image =
Extract(base_images, num_images, total_extracted_images, LimitImages[stage]);
}
@@ -170,7 +178,7 @@ bool IsASTCSupported() {
for (const GLenum format : formats) {
for (const GLenum support : required_support) {
GLint value;
- glGetInternalformativ(GL_TEXTURE_2D, format, support, 1, &value);
+ glGetInternalformativ(target, format, support, 1, &value);
if (value != GL_FULL_SUPPORT) {
return false;
}
@@ -180,16 +188,32 @@ bool IsASTCSupported() {
return true;
}
+/// @brief Returns true when a GL_RENDERER is a Turing GPU
+/// @param renderer GL_RENDERER string
+bool IsTuring(std::string_view renderer) {
+ static constexpr std::array<std::string_view, 12> TURING_GPUS = {
+ "GTX 1650", "GTX 1660", "RTX 2060", "RTX 2070",
+ "RTX 2080", "TITAN RTX", "Quadro RTX 3000", "Quadro RTX 4000",
+ "Quadro RTX 5000", "Quadro RTX 6000", "Quadro RTX 8000", "Tesla T4",
+ };
+ return std::any_of(TURING_GPUS.begin(), TURING_GPUS.end(),
+ [renderer](std::string_view candidate) {
+ return renderer.find(candidate) != std::string_view::npos;
+ });
+}
+
} // Anonymous namespace
Device::Device()
: max_uniform_buffers{BuildMaxUniformBuffers()}, base_bindings{BuildBaseBindings()} {
const std::string_view vendor = reinterpret_cast<const char*>(glGetString(GL_VENDOR));
+ const std::string_view renderer = reinterpret_cast<const char*>(glGetString(GL_RENDERER));
const std::string_view version = reinterpret_cast<const char*>(glGetString(GL_VERSION));
const std::vector extensions = GetExtensions();
const bool is_nvidia = vendor == "NVIDIA Corporation";
const bool is_amd = vendor == "ATI Technologies Inc.";
+ const bool is_turing = is_nvidia && IsTuring(renderer);
bool disable_fast_buffer_sub_data = false;
if (is_nvidia && version == "4.6.0 NVIDIA 443.24") {
@@ -208,12 +232,21 @@ Device::Device()
has_shader_ballot = GLAD_GL_ARB_shader_ballot;
has_vertex_viewport_layer = GLAD_GL_ARB_shader_viewport_layer_array;
has_image_load_formatted = HasExtension(extensions, "GL_EXT_shader_image_load_formatted");
+ has_texture_shadow_lod = HasExtension(extensions, "GL_EXT_texture_shadow_lod");
has_astc = IsASTCSupported();
has_variable_aoffi = TestVariableAoffi();
has_component_indexing_bug = is_amd;
has_precise_bug = TestPreciseBug();
- has_fast_buffer_sub_data = is_nvidia && !disable_fast_buffer_sub_data;
has_nv_viewport_array2 = GLAD_GL_NV_viewport_array2;
+
+ // At the moment of writing this, only Nvidia's driver optimizes BufferSubData on exclusive
+ // uniform buffers as "push constants"
+ has_fast_buffer_sub_data = is_nvidia && !disable_fast_buffer_sub_data;
+
+ // Nvidia's driver on Turing GPUs randomly crashes when the buffer is made resident, or on
+ // DeleteBuffers. Disable unified memory on these devices.
+ has_vertex_buffer_unified_memory = GLAD_GL_NV_vertex_buffer_unified_memory && !is_turing;
+
use_assembly_shaders = Settings::values.use_assembly_shaders && GLAD_GL_NV_gpu_program5 &&
GLAD_GL_NV_compute_program5 && GLAD_GL_NV_transform_feedback &&
GLAD_GL_NV_transform_feedback2;
@@ -237,6 +270,7 @@ Device::Device(std::nullptr_t) {
has_shader_ballot = true;
has_vertex_viewport_layer = true;
has_image_load_formatted = true;
+ has_texture_shadow_lod = true;
has_variable_aoffi = true;
}
diff --git a/src/video_core/renderer_opengl/gl_device.h b/src/video_core/renderer_opengl/gl_device.h
index 145347943..e1d811966 100644
--- a/src/video_core/renderer_opengl/gl_device.h
+++ b/src/video_core/renderer_opengl/gl_device.h
@@ -68,6 +68,14 @@ public:
return has_image_load_formatted;
}
+ bool HasTextureShadowLod() const {
+ return has_texture_shadow_lod;
+ }
+
+ bool HasVertexBufferUnifiedMemory() const {
+ return has_vertex_buffer_unified_memory;
+ }
+
bool HasASTC() const {
return has_astc;
}
@@ -110,6 +118,8 @@ private:
bool has_shader_ballot{};
bool has_vertex_viewport_layer{};
bool has_image_load_formatted{};
+ bool has_texture_shadow_lod{};
+ bool has_vertex_buffer_unified_memory{};
bool has_astc{};
bool has_variable_aoffi{};
bool has_component_indexing_bug{};
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index 2d6c11320..362457ffe 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -61,7 +61,8 @@ constexpr std::size_t NUM_CONST_BUFFERS_BYTES_PER_STAGE =
constexpr std::size_t TOTAL_CONST_BUFFER_BYTES =
NUM_CONST_BUFFERS_BYTES_PER_STAGE * Maxwell::MaxShaderStage;
-constexpr std::size_t NumSupportedVertexAttributes = 16;
+constexpr std::size_t NUM_SUPPORTED_VERTEX_ATTRIBUTES = 16;
+constexpr std::size_t NUM_SUPPORTED_VERTEX_BINDINGS = 16;
template <typename Engine, typename Entry>
Tegra::Texture::FullTextureInfo GetTextureInfo(const Engine& engine, const Entry& entry,
@@ -193,7 +194,7 @@ void RasterizerOpenGL::SetupVertexFormat() {
// avoid OpenGL errors.
// TODO(Subv): Analyze the shader to identify which attributes are actually used and don't
// assume every shader uses them all.
- for (std::size_t index = 0; index < NumSupportedVertexAttributes; ++index) {
+ for (std::size_t index = 0; index < NUM_SUPPORTED_VERTEX_ATTRIBUTES; ++index) {
if (!flags[Dirty::VertexFormat0 + index]) {
continue;
}
@@ -231,9 +232,11 @@ void RasterizerOpenGL::SetupVertexBuffer() {
MICROPROFILE_SCOPE(OpenGL_VB);
+ const bool use_unified_memory = device.HasVertexBufferUnifiedMemory();
+
// Upload all guest vertex arrays sequentially to our buffer
const auto& regs = gpu.regs;
- for (std::size_t index = 0; index < Maxwell::NumVertexArrays; ++index) {
+ for (std::size_t index = 0; index < NUM_SUPPORTED_VERTEX_BINDINGS; ++index) {
if (!flags[Dirty::VertexBuffer0 + index]) {
continue;
}
@@ -246,16 +249,25 @@ void RasterizerOpenGL::SetupVertexBuffer() {
const GPUVAddr start = vertex_array.StartAddress();
const GPUVAddr end = regs.vertex_array_limit[index].LimitAddress();
-
ASSERT(end >= start);
+
+ const GLuint gl_index = static_cast<GLuint>(index);
const u64 size = end - start;
if (size == 0) {
- glBindVertexBuffer(static_cast<GLuint>(index), 0, 0, vertex_array.stride);
+ glBindVertexBuffer(gl_index, 0, 0, vertex_array.stride);
+ if (use_unified_memory) {
+ glBufferAddressRangeNV(GL_VERTEX_ATTRIB_ARRAY_ADDRESS_NV, gl_index, 0, 0);
+ }
continue;
}
- const auto [vertex_buffer, vertex_buffer_offset] = buffer_cache.UploadMemory(start, size);
- glBindVertexBuffer(static_cast<GLuint>(index), vertex_buffer, vertex_buffer_offset,
- vertex_array.stride);
+ const auto info = buffer_cache.UploadMemory(start, size);
+ if (use_unified_memory) {
+ glBindVertexBuffer(gl_index, 0, 0, vertex_array.stride);
+ glBufferAddressRangeNV(GL_VERTEX_ATTRIB_ARRAY_ADDRESS_NV, gl_index,
+ info.address + info.offset, size);
+ } else {
+ glBindVertexBuffer(gl_index, info.handle, info.offset, vertex_array.stride);
+ }
}
}
@@ -268,7 +280,7 @@ void RasterizerOpenGL::SetupVertexInstances() {
flags[Dirty::VertexInstances] = false;
const auto& regs = gpu.regs;
- for (std::size_t index = 0; index < NumSupportedVertexAttributes; ++index) {
+ for (std::size_t index = 0; index < NUM_SUPPORTED_VERTEX_ATTRIBUTES; ++index) {
if (!flags[Dirty::VertexInstance0 + index]) {
continue;
}
@@ -285,9 +297,9 @@ GLintptr RasterizerOpenGL::SetupIndexBuffer() {
MICROPROFILE_SCOPE(OpenGL_Index);
const auto& regs = system.GPU().Maxwell3D().regs;
const std::size_t size = CalculateIndexBufferSize();
- const auto [buffer, offset] = buffer_cache.UploadMemory(regs.index_array.IndexStart(), size);
- glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, buffer);
- return offset;
+ const auto info = buffer_cache.UploadMemory(regs.index_array.IndexStart(), size);
+ glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, info.handle);
+ return info.offset;
}
void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
@@ -643,9 +655,9 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
if (!device.UseAssemblyShaders()) {
MaxwellUniformData ubo;
ubo.SetFromRegs(gpu);
- const auto [buffer, offset] =
+ const auto info =
buffer_cache.UploadHostMemory(&ubo, sizeof(ubo), device.GetUniformBufferAlignment());
- glBindBufferRange(GL_UNIFORM_BUFFER, EmulationUniformBlockBinding, buffer, offset,
+ glBindBufferRange(GL_UNIFORM_BUFFER, EmulationUniformBlockBinding, info.handle, info.offset,
static_cast<GLsizeiptr>(sizeof(ubo)));
}
@@ -956,8 +968,7 @@ void RasterizerOpenGL::SetupConstBuffer(GLenum stage, u32 binding,
if (device.UseAssemblyShaders()) {
glBindBufferRangeNV(stage, entry.GetIndex(), 0, 0, 0);
} else {
- glBindBufferRange(GL_UNIFORM_BUFFER, binding,
- buffer_cache.GetEmptyBuffer(sizeof(float)), 0, sizeof(float));
+ glBindBufferRange(GL_UNIFORM_BUFFER, binding, 0, 0, sizeof(float));
}
return;
}
@@ -970,24 +981,25 @@ void RasterizerOpenGL::SetupConstBuffer(GLenum stage, u32 binding,
const std::size_t alignment = use_unified ? 4 : device.GetUniformBufferAlignment();
const GPUVAddr gpu_addr = buffer.address;
- auto [cbuf, offset] = buffer_cache.UploadMemory(gpu_addr, size, alignment, false, fast_upload);
+ auto info = buffer_cache.UploadMemory(gpu_addr, size, alignment, false, fast_upload);
if (device.UseAssemblyShaders()) {
UNIMPLEMENTED_IF(use_unified);
- if (offset != 0) {
+ if (info.offset != 0) {
const GLuint staging_cbuf = staging_cbufs[current_cbuf++];
- glCopyNamedBufferSubData(cbuf, staging_cbuf, offset, 0, size);
- cbuf = staging_cbuf;
- offset = 0;
+ glCopyNamedBufferSubData(info.handle, staging_cbuf, info.offset, 0, size);
+ info.handle = staging_cbuf;
+ info.offset = 0;
}
- glBindBufferRangeNV(stage, binding, cbuf, offset, size);
+ glBindBufferRangeNV(stage, binding, info.handle, info.offset, size);
return;
}
if (use_unified) {
- glCopyNamedBufferSubData(cbuf, unified_uniform_buffer.handle, offset, unified_offset, size);
+ glCopyNamedBufferSubData(info.handle, unified_uniform_buffer.handle, info.offset,
+ unified_offset, size);
} else {
- glBindBufferRange(GL_UNIFORM_BUFFER, binding, cbuf, offset, size);
+ glBindBufferRange(GL_UNIFORM_BUFFER, binding, info.handle, info.offset, size);
}
}
@@ -1023,9 +1035,8 @@ void RasterizerOpenGL::SetupComputeGlobalMemory(Shader* kernel) {
void RasterizerOpenGL::SetupGlobalMemory(u32 binding, const GlobalMemoryEntry& entry,
GPUVAddr gpu_addr, std::size_t size) {
const auto alignment{device.GetShaderStorageBufferAlignment()};
- const auto [ssbo, buffer_offset] =
- buffer_cache.UploadMemory(gpu_addr, size, alignment, entry.is_written);
- glBindBufferRange(GL_SHADER_STORAGE_BUFFER, binding, ssbo, buffer_offset,
+ const auto info = buffer_cache.UploadMemory(gpu_addr, size, alignment, entry.is_written);
+ glBindBufferRange(GL_SHADER_STORAGE_BUFFER, binding, info.handle, info.offset,
static_cast<GLsizeiptr>(size));
}
@@ -1712,8 +1723,9 @@ void RasterizerOpenGL::EndTransformFeedback() {
const GLuint handle = transform_feedback_buffers[index].handle;
const GPUVAddr gpu_addr = binding.Address();
const std::size_t size = binding.buffer_size;
- const auto [dest_buffer, offset] = buffer_cache.UploadMemory(gpu_addr, size, 4, true);
- glCopyNamedBufferSubData(handle, dest_buffer, 0, offset, static_cast<GLsizeiptr>(size));
+ const auto info = buffer_cache.UploadMemory(gpu_addr, size, 4, true);
+ glCopyNamedBufferSubData(handle, info.handle, 0, info.offset,
+ static_cast<GLsizeiptr>(size));
}
}
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp
index 46e780a06..c6a3bf3a1 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp
@@ -460,8 +460,9 @@ Shader* ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
const u8* host_ptr_b = memory_manager.GetPointer(address_b);
code_b = GetShaderCode(memory_manager, address_b, host_ptr_b, false);
}
+ const std::size_t code_size = code.size() * sizeof(u64);
- const auto unique_identifier = GetUniqueIdentifier(
+ const u64 unique_identifier = GetUniqueIdentifier(
GetShaderType(program), program == Maxwell::ShaderProgram::VertexA, code, code_b);
const ShaderParameters params{system, disk_cache, device,
@@ -477,7 +478,7 @@ Shader* ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
Shader* const result = shader.get();
if (cpu_addr) {
- Register(std::move(shader), *cpu_addr, code.size() * sizeof(u64));
+ Register(std::move(shader), *cpu_addr, code_size);
} else {
null_shader = std::move(shader);
}
@@ -495,8 +496,9 @@ Shader* ShaderCacheOpenGL::GetComputeKernel(GPUVAddr code_addr) {
const auto host_ptr{memory_manager.GetPointer(code_addr)};
// No kernel found, create a new one
- auto code{GetShaderCode(memory_manager, code_addr, host_ptr, true)};
- const auto unique_identifier{GetUniqueIdentifier(ShaderType::Compute, false, code)};
+ ProgramCode code{GetShaderCode(memory_manager, code_addr, host_ptr, true)};
+ const std::size_t code_size{code.size() * sizeof(u64)};
+ const u64 unique_identifier{GetUniqueIdentifier(ShaderType::Compute, false, code)};
const ShaderParameters params{system, disk_cache, device,
*cpu_addr, host_ptr, unique_identifier};
@@ -511,7 +513,7 @@ Shader* ShaderCacheOpenGL::GetComputeKernel(GPUVAddr code_addr) {
Shader* const result = kernel.get();
if (cpu_addr) {
- Register(std::move(kernel), *cpu_addr, code.size() * sizeof(u64));
+ Register(std::move(kernel), *cpu_addr, code_size);
} else {
null_kernel = std::move(kernel);
}
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.h b/src/video_core/renderer_opengl/gl_shader_cache.h
index 6848f1388..994aaeaf2 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.h
+++ b/src/video_core/renderer_opengl/gl_shader_cache.h
@@ -37,7 +37,6 @@ namespace OpenGL {
class Device;
class RasterizerOpenGL;
-struct UnspecializedShader;
using Maxwell = Tegra::Engines::Maxwell3D::Regs;
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index d6e30b321..2c49aeaac 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -37,6 +37,7 @@ using Tegra::Shader::IpaMode;
using Tegra::Shader::IpaSampleMode;
using Tegra::Shader::PixelImap;
using Tegra::Shader::Register;
+using Tegra::Shader::TextureType;
using VideoCommon::Shader::BuildTransformFeedback;
using VideoCommon::Shader::Registry;
@@ -526,6 +527,9 @@ private:
if (device.HasImageLoadFormatted()) {
code.AddLine("#extension GL_EXT_shader_image_load_formatted : require");
}
+ if (device.HasTextureShadowLod()) {
+ code.AddLine("#extension GL_EXT_texture_shadow_lod : require");
+ }
if (device.HasWarpIntrinsics()) {
code.AddLine("#extension GL_NV_gpu_shader5 : require");
code.AddLine("#extension GL_NV_shader_thread_group : require");
@@ -909,13 +913,13 @@ private:
return "samplerBuffer";
}
switch (sampler.type) {
- case Tegra::Shader::TextureType::Texture1D:
+ case TextureType::Texture1D:
return "sampler1D";
- case Tegra::Shader::TextureType::Texture2D:
+ case TextureType::Texture2D:
return "sampler2D";
- case Tegra::Shader::TextureType::Texture3D:
+ case TextureType::Texture3D:
return "sampler3D";
- case Tegra::Shader::TextureType::TextureCube:
+ case TextureType::TextureCube:
return "samplerCube";
default:
UNREACHABLE();
@@ -1380,8 +1384,19 @@ private:
const std::size_t count = operation.GetOperandsCount();
const bool has_array = meta->sampler.is_array;
const bool has_shadow = meta->sampler.is_shadow;
+ const bool workaround_lod_array_shadow_as_grad =
+ !device.HasTextureShadowLod() && function_suffix == "Lod" && meta->sampler.is_shadow &&
+ ((meta->sampler.type == TextureType::Texture2D && meta->sampler.is_array) ||
+ meta->sampler.type == TextureType::TextureCube);
+
+ std::string expr = "texture";
+
+ if (workaround_lod_array_shadow_as_grad) {
+ expr += "Grad";
+ } else {
+ expr += function_suffix;
+ }
- std::string expr = "texture" + function_suffix;
if (!meta->aoffi.empty()) {
expr += "Offset";
} else if (!meta->ptp.empty()) {
@@ -1415,6 +1430,16 @@ private:
expr += ')';
}
+ if (workaround_lod_array_shadow_as_grad) {
+ switch (meta->sampler.type) {
+ case TextureType::Texture2D:
+ return expr + ", vec2(0.0), vec2(0.0))";
+ case TextureType::TextureCube:
+ return expr + ", vec3(0.0), vec3(0.0))";
+ }
+ UNREACHABLE();
+ }
+
for (const auto& variant : extras) {
if (const auto argument = std::get_if<TextureArgument>(&variant)) {
expr += GenerateTextureArgument(*argument);
@@ -2041,8 +2066,19 @@ private:
const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
ASSERT(meta);
- std::string expr = GenerateTexture(
- operation, "Lod", {TextureArgument{Type::Float, meta->lod}, TextureOffset{}});
+ std::string expr{};
+
+ if (!device.HasTextureShadowLod() && meta->sampler.is_shadow &&
+ ((meta->sampler.type == TextureType::Texture2D && meta->sampler.is_array) ||
+ meta->sampler.type == TextureType::TextureCube)) {
+ LOG_ERROR(Render_OpenGL,
+ "Device lacks GL_EXT_texture_shadow_lod, using textureGrad as a workaround");
+ expr = GenerateTexture(operation, "Lod", {});
+ } else {
+ expr = GenerateTexture(operation, "Lod",
+ {TextureArgument{Type::Float, meta->lod}, TextureOffset{}});
+ }
+
if (meta->sampler.is_shadow) {
expr = "vec4(" + expr + ')';
}
diff --git a/src/video_core/renderer_opengl/gl_stream_buffer.cpp b/src/video_core/renderer_opengl/gl_stream_buffer.cpp
index 932a2f69e..3655ff629 100644
--- a/src/video_core/renderer_opengl/gl_stream_buffer.cpp
+++ b/src/video_core/renderer_opengl/gl_stream_buffer.cpp
@@ -2,11 +2,13 @@
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
-#include <deque>
+#include <tuple>
#include <vector>
+
#include "common/alignment.h"
#include "common/assert.h"
#include "common/microprofile.h"
+#include "video_core/renderer_opengl/gl_device.h"
#include "video_core/renderer_opengl/gl_stream_buffer.h"
MICROPROFILE_DEFINE(OpenGL_StreamBuffer, "OpenGL", "Stream Buffer Orphaning",
@@ -14,8 +16,7 @@ MICROPROFILE_DEFINE(OpenGL_StreamBuffer, "OpenGL", "Stream Buffer Orphaning",
namespace OpenGL {
-OGLStreamBuffer::OGLStreamBuffer(GLsizeiptr size, bool vertex_data_usage, bool prefer_coherent,
- bool use_persistent)
+OGLStreamBuffer::OGLStreamBuffer(const Device& device, GLsizeiptr size, bool vertex_data_usage)
: buffer_size(size) {
gl_buffer.Create();
@@ -29,23 +30,19 @@ OGLStreamBuffer::OGLStreamBuffer(GLsizeiptr size, bool vertex_data_usage, bool p
allocate_size *= 2;
}
- if (use_persistent) {
- persistent = true;
- coherent = prefer_coherent;
- const GLbitfield flags =
- GL_MAP_WRITE_BIT | GL_MAP_PERSISTENT_BIT | (coherent ? GL_MAP_COHERENT_BIT : 0);
- glNamedBufferStorage(gl_buffer.handle, allocate_size, nullptr, flags);
- mapped_ptr = static_cast<u8*>(glMapNamedBufferRange(
- gl_buffer.handle, 0, buffer_size, flags | (coherent ? 0 : GL_MAP_FLUSH_EXPLICIT_BIT)));
- } else {
- glNamedBufferData(gl_buffer.handle, allocate_size, nullptr, GL_STREAM_DRAW);
+ static constexpr GLbitfield flags = GL_MAP_WRITE_BIT | GL_MAP_PERSISTENT_BIT;
+ glNamedBufferStorage(gl_buffer.handle, allocate_size, nullptr, flags);
+ mapped_ptr = static_cast<u8*>(
+ glMapNamedBufferRange(gl_buffer.handle, 0, buffer_size, flags | GL_MAP_FLUSH_EXPLICIT_BIT));
+
+ if (device.HasVertexBufferUnifiedMemory()) {
+ glMakeNamedBufferResidentNV(gl_buffer.handle, GL_READ_ONLY);
+ glGetNamedBufferParameterui64vNV(gl_buffer.handle, GL_BUFFER_GPU_ADDRESS_NV, &gpu_address);
}
}
OGLStreamBuffer::~OGLStreamBuffer() {
- if (persistent) {
- glUnmapNamedBuffer(gl_buffer.handle);
- }
+ glUnmapNamedBuffer(gl_buffer.handle);
gl_buffer.Release();
}
@@ -60,36 +57,21 @@ std::tuple<u8*, GLintptr, bool> OGLStreamBuffer::Map(GLsizeiptr size, GLintptr a
bool invalidate = false;
if (buffer_pos + size > buffer_size) {
+ MICROPROFILE_SCOPE(OpenGL_StreamBuffer);
+ glInvalidateBufferData(gl_buffer.handle);
+
buffer_pos = 0;
invalidate = true;
-
- if (persistent) {
- glUnmapNamedBuffer(gl_buffer.handle);
- }
}
- if (invalidate || !persistent) {
- MICROPROFILE_SCOPE(OpenGL_StreamBuffer);
- GLbitfield flags = GL_MAP_WRITE_BIT | (persistent ? GL_MAP_PERSISTENT_BIT : 0) |
- (coherent ? GL_MAP_COHERENT_BIT : GL_MAP_FLUSH_EXPLICIT_BIT) |
- (invalidate ? GL_MAP_INVALIDATE_BUFFER_BIT : GL_MAP_UNSYNCHRONIZED_BIT);
- mapped_ptr = static_cast<u8*>(
- glMapNamedBufferRange(gl_buffer.handle, buffer_pos, buffer_size - buffer_pos, flags));
- mapped_offset = buffer_pos;
- }
-
- return std::make_tuple(mapped_ptr + buffer_pos - mapped_offset, buffer_pos, invalidate);
+ return std::make_tuple(mapped_ptr + buffer_pos, buffer_pos, invalidate);
}
void OGLStreamBuffer::Unmap(GLsizeiptr size) {
ASSERT(size <= mapped_size);
- if (!coherent && size > 0) {
- glFlushMappedNamedBufferRange(gl_buffer.handle, buffer_pos - mapped_offset, size);
- }
-
- if (!persistent) {
- glUnmapNamedBuffer(gl_buffer.handle);
+ if (size > 0) {
+ glFlushMappedNamedBufferRange(gl_buffer.handle, buffer_pos, size);
}
buffer_pos += size;
diff --git a/src/video_core/renderer_opengl/gl_stream_buffer.h b/src/video_core/renderer_opengl/gl_stream_buffer.h
index 866da3594..307a67113 100644
--- a/src/video_core/renderer_opengl/gl_stream_buffer.h
+++ b/src/video_core/renderer_opengl/gl_stream_buffer.h
@@ -11,10 +11,11 @@
namespace OpenGL {
+class Device;
+
class OGLStreamBuffer : private NonCopyable {
public:
- explicit OGLStreamBuffer(GLsizeiptr size, bool vertex_data_usage, bool prefer_coherent = false,
- bool use_persistent = true);
+ explicit OGLStreamBuffer(const Device& device, GLsizeiptr size, bool vertex_data_usage);
~OGLStreamBuffer();
/*
@@ -33,19 +34,20 @@ public:
return gl_buffer.handle;
}
- GLsizeiptr Size() const {
+ u64 Address() const {
+ return gpu_address;
+ }
+
+ GLsizeiptr Size() const noexcept {
return buffer_size;
}
private:
OGLBuffer gl_buffer;
- bool coherent = false;
- bool persistent = false;
-
+ GLuint64EXT gpu_address = 0;
GLintptr buffer_pos = 0;
GLsizeiptr buffer_size = 0;
- GLintptr mapped_offset = 0;
GLsizeiptr mapped_size = 0;
u8* mapped_ptr = nullptr;
};
diff --git a/src/video_core/renderer_opengl/maxwell_to_gl.h b/src/video_core/renderer_opengl/maxwell_to_gl.h
index 994ae98eb..35e329240 100644
--- a/src/video_core/renderer_opengl/maxwell_to_gl.h
+++ b/src/video_core/renderer_opengl/maxwell_to_gl.h
@@ -46,10 +46,8 @@ inline GLenum VertexType(Maxwell::VertexAttribute attrib) {
return GL_UNSIGNED_INT;
case Maxwell::VertexAttribute::Size::Size_10_10_10_2:
return GL_UNSIGNED_INT_2_10_10_10_REV;
- default:
- LOG_ERROR(Render_OpenGL, "Unimplemented vertex size={}", attrib.SizeString());
- return {};
}
+ break;
case Maxwell::VertexAttribute::Type::SignedInt:
case Maxwell::VertexAttribute::Type::SignedNorm:
switch (attrib.size) {
@@ -70,10 +68,8 @@ inline GLenum VertexType(Maxwell::VertexAttribute attrib) {
return GL_INT;
case Maxwell::VertexAttribute::Size::Size_10_10_10_2:
return GL_INT_2_10_10_10_REV;
- default:
- LOG_ERROR(Render_OpenGL, "Unimplemented vertex size={}", attrib.SizeString());
- return {};
}
+ break;
case Maxwell::VertexAttribute::Type::Float:
switch (attrib.size) {
case Maxwell::VertexAttribute::Size::Size_16:
@@ -86,10 +82,8 @@ inline GLenum VertexType(Maxwell::VertexAttribute attrib) {
case Maxwell::VertexAttribute::Size::Size_32_32_32:
case Maxwell::VertexAttribute::Size::Size_32_32_32_32:
return GL_FLOAT;
- default:
- LOG_ERROR(Render_OpenGL, "Unimplemented vertex size={}", attrib.SizeString());
- return {};
}
+ break;
case Maxwell::VertexAttribute::Type::UnsignedScaled:
switch (attrib.size) {
case Maxwell::VertexAttribute::Size::Size_8:
@@ -102,10 +96,8 @@ inline GLenum VertexType(Maxwell::VertexAttribute attrib) {
case Maxwell::VertexAttribute::Size::Size_16_16_16:
case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
return GL_UNSIGNED_SHORT;
- default:
- LOG_ERROR(Render_OpenGL, "Unimplemented vertex size={}", attrib.SizeString());
- return {};
}
+ break;
case Maxwell::VertexAttribute::Type::SignedScaled:
switch (attrib.size) {
case Maxwell::VertexAttribute::Size::Size_8:
@@ -118,14 +110,12 @@ inline GLenum VertexType(Maxwell::VertexAttribute attrib) {
case Maxwell::VertexAttribute::Size::Size_16_16_16:
case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
return GL_SHORT;
- default:
- LOG_ERROR(Render_OpenGL, "Unimplemented vertex size={}", attrib.SizeString());
- return {};
}
- default:
- LOG_ERROR(Render_OpenGL, "Unimplemented vertex type={}", attrib.TypeString());
- return {};
+ break;
}
+ UNIMPLEMENTED_MSG("Unimplemented vertex type={} and size={}", attrib.TypeString(),
+ attrib.SizeString());
+ return {};
}
inline GLenum IndexFormat(Maxwell::IndexFormat index_format) {
@@ -137,8 +127,7 @@ inline GLenum IndexFormat(Maxwell::IndexFormat index_format) {
case Maxwell::IndexFormat::UnsignedInt:
return GL_UNSIGNED_INT;
}
- LOG_CRITICAL(Render_OpenGL, "Unimplemented index_format={}", static_cast<u32>(index_format));
- UNREACHABLE();
+ UNREACHABLE_MSG("Invalid index_format={}", static_cast<u32>(index_format));
return {};
}
@@ -180,33 +169,32 @@ inline GLenum PrimitiveTopology(Maxwell::PrimitiveTopology topology) {
}
inline GLenum TextureFilterMode(Tegra::Texture::TextureFilter filter_mode,
- Tegra::Texture::TextureMipmapFilter mip_filter_mode) {
+ Tegra::Texture::TextureMipmapFilter mipmap_filter_mode) {
switch (filter_mode) {
- case Tegra::Texture::TextureFilter::Linear: {
- switch (mip_filter_mode) {
+ case Tegra::Texture::TextureFilter::Nearest:
+ switch (mipmap_filter_mode) {
case Tegra::Texture::TextureMipmapFilter::None:
- return GL_LINEAR;
+ return GL_NEAREST;
case Tegra::Texture::TextureMipmapFilter::Nearest:
- return GL_LINEAR_MIPMAP_NEAREST;
+ return GL_NEAREST_MIPMAP_NEAREST;
case Tegra::Texture::TextureMipmapFilter::Linear:
- return GL_LINEAR_MIPMAP_LINEAR;
+ return GL_NEAREST_MIPMAP_LINEAR;
}
break;
- }
- case Tegra::Texture::TextureFilter::Nearest: {
- switch (mip_filter_mode) {
+ case Tegra::Texture::TextureFilter::Linear:
+ switch (mipmap_filter_mode) {
case Tegra::Texture::TextureMipmapFilter::None:
- return GL_NEAREST;
+ return GL_LINEAR;
case Tegra::Texture::TextureMipmapFilter::Nearest:
- return GL_NEAREST_MIPMAP_NEAREST;
+ return GL_LINEAR_MIPMAP_NEAREST;
case Tegra::Texture::TextureMipmapFilter::Linear:
- return GL_NEAREST_MIPMAP_LINEAR;
+ return GL_LINEAR_MIPMAP_LINEAR;
}
break;
}
- }
- LOG_ERROR(Render_OpenGL, "Unimplemented texture filter mode={}", static_cast<u32>(filter_mode));
- return GL_LINEAR;
+ UNREACHABLE_MSG("Invalid texture filter mode={} and mipmap filter mode={}",
+ static_cast<u32>(filter_mode), static_cast<u32>(mipmap_filter_mode));
+ return GL_NEAREST;
}
inline GLenum WrapMode(Tegra::Texture::WrapMode wrap_mode) {
@@ -229,10 +217,9 @@ inline GLenum WrapMode(Tegra::Texture::WrapMode wrap_mode) {
} else {
return GL_MIRROR_CLAMP_TO_EDGE;
}
- default:
- LOG_ERROR(Render_OpenGL, "Unimplemented texture wrap mode={}", static_cast<u32>(wrap_mode));
- return GL_REPEAT;
}
+ UNIMPLEMENTED_MSG("Unimplemented texture wrap mode={}", static_cast<u32>(wrap_mode));
+ return GL_REPEAT;
}
inline GLenum DepthCompareFunc(Tegra::Texture::DepthCompareFunc func) {
@@ -254,8 +241,7 @@ inline GLenum DepthCompareFunc(Tegra::Texture::DepthCompareFunc func) {
case Tegra::Texture::DepthCompareFunc::Always:
return GL_ALWAYS;
}
- LOG_ERROR(Render_OpenGL, "Unimplemented texture depth compare function ={}",
- static_cast<u32>(func));
+ UNIMPLEMENTED_MSG("Unimplemented texture depth compare function={}", static_cast<u32>(func));
return GL_GREATER;
}
@@ -277,7 +263,7 @@ inline GLenum BlendEquation(Maxwell::Blend::Equation equation) {
case Maxwell::Blend::Equation::MaxGL:
return GL_MAX;
}
- LOG_ERROR(Render_OpenGL, "Unimplemented blend equation={}", static_cast<u32>(equation));
+ UNIMPLEMENTED_MSG("Unimplemented blend equation={}", static_cast<u32>(equation));
return GL_FUNC_ADD;
}
@@ -341,7 +327,7 @@ inline GLenum BlendFunc(Maxwell::Blend::Factor factor) {
case Maxwell::Blend::Factor::OneMinusConstantAlphaGL:
return GL_ONE_MINUS_CONSTANT_ALPHA;
}
- LOG_ERROR(Render_OpenGL, "Unimplemented blend factor={}", static_cast<u32>(factor));
+ UNIMPLEMENTED_MSG("Unimplemented blend factor={}", static_cast<u32>(factor));
return GL_ZERO;
}
@@ -361,7 +347,7 @@ inline GLenum SwizzleSource(Tegra::Texture::SwizzleSource source) {
case Tegra::Texture::SwizzleSource::OneFloat:
return GL_ONE;
}
- LOG_ERROR(Render_OpenGL, "Unimplemented swizzle source={}", static_cast<u32>(source));
+ UNIMPLEMENTED_MSG("Unimplemented swizzle source={}", static_cast<u32>(source));
return GL_ZERO;
}
@@ -392,7 +378,7 @@ inline GLenum ComparisonOp(Maxwell::ComparisonOp comparison) {
case Maxwell::ComparisonOp::AlwaysOld:
return GL_ALWAYS;
}
- LOG_ERROR(Render_OpenGL, "Unimplemented comparison op={}", static_cast<u32>(comparison));
+ UNIMPLEMENTED_MSG("Unimplemented comparison op={}", static_cast<u32>(comparison));
return GL_ALWAYS;
}
@@ -423,7 +409,7 @@ inline GLenum StencilOp(Maxwell::StencilOp stencil) {
case Maxwell::StencilOp::DecrWrapOGL:
return GL_DECR_WRAP;
}
- LOG_ERROR(Render_OpenGL, "Unimplemented stencil op={}", static_cast<u32>(stencil));
+ UNIMPLEMENTED_MSG("Unimplemented stencil op={}", static_cast<u32>(stencil));
return GL_KEEP;
}
@@ -434,7 +420,7 @@ inline GLenum FrontFace(Maxwell::FrontFace front_face) {
case Maxwell::FrontFace::CounterClockWise:
return GL_CCW;
}
- LOG_ERROR(Render_OpenGL, "Unimplemented front face cull={}", static_cast<u32>(front_face));
+ UNIMPLEMENTED_MSG("Unimplemented front face cull={}", static_cast<u32>(front_face));
return GL_CCW;
}
@@ -447,7 +433,7 @@ inline GLenum CullFace(Maxwell::CullFace cull_face) {
case Maxwell::CullFace::FrontAndBack:
return GL_FRONT_AND_BACK;
}
- LOG_ERROR(Render_OpenGL, "Unimplemented cull face={}", static_cast<u32>(cull_face));
+ UNIMPLEMENTED_MSG("Unimplemented cull face={}", static_cast<u32>(cull_face));
return GL_BACK;
}
@@ -486,7 +472,7 @@ inline GLenum LogicOp(Maxwell::LogicOperation operation) {
case Maxwell::LogicOperation::Set:
return GL_SET;
}
- LOG_ERROR(Render_OpenGL, "Unimplemented logic operation={}", static_cast<u32>(operation));
+ UNIMPLEMENTED_MSG("Unimplemented logic operation={}", static_cast<u32>(operation));
return GL_COPY;
}
diff --git a/src/video_core/renderer_opengl/renderer_opengl.cpp b/src/video_core/renderer_opengl/renderer_opengl.cpp
index 6214fcbc3..c40adb6e7 100644
--- a/src/video_core/renderer_opengl/renderer_opengl.cpp
+++ b/src/video_core/renderer_opengl/renderer_opengl.cpp
@@ -488,6 +488,15 @@ void RendererOpenGL::InitOpenGLObjects() {
// Clear screen to black
LoadColorToActiveGLTexture(0, 0, 0, 0, screen_info.texture);
+
+ // Enable unified vertex attributes and query vertex buffer address when the driver supports it
+ if (device.HasVertexBufferUnifiedMemory()) {
+ glEnableClientState(GL_VERTEX_ATTRIB_ARRAY_UNIFIED_NV);
+
+ glMakeNamedBufferResidentNV(vertex_buffer.handle, GL_READ_ONLY);
+ glGetNamedBufferParameterui64vNV(vertex_buffer.handle, GL_BUFFER_GPU_ADDRESS_NV,
+ &vertex_buffer_address);
+ }
}
void RendererOpenGL::AddTelemetryFields() {
@@ -656,7 +665,13 @@ void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) {
offsetof(ScreenRectVertex, tex_coord));
glVertexAttribBinding(PositionLocation, 0);
glVertexAttribBinding(TexCoordLocation, 0);
- glBindVertexBuffer(0, vertex_buffer.handle, 0, sizeof(ScreenRectVertex));
+ if (device.HasVertexBufferUnifiedMemory()) {
+ glBindVertexBuffer(0, 0, 0, sizeof(ScreenRectVertex));
+ glBufferAddressRangeNV(GL_VERTEX_ATTRIB_ARRAY_ADDRESS_NV, 0, vertex_buffer_address,
+ sizeof(vertices));
+ } else {
+ glBindVertexBuffer(0, vertex_buffer.handle, 0, sizeof(ScreenRectVertex));
+ }
glBindTextureUnit(0, screen_info.display_texture);
glBindSampler(0, 0);
diff --git a/src/video_core/renderer_opengl/renderer_opengl.h b/src/video_core/renderer_opengl/renderer_opengl.h
index 61bf507f4..8b18d32e6 100644
--- a/src/video_core/renderer_opengl/renderer_opengl.h
+++ b/src/video_core/renderer_opengl/renderer_opengl.h
@@ -107,6 +107,9 @@ private:
OGLPipeline pipeline;
OGLFramebuffer screenshot_framebuffer;
+ // GPU address of the vertex buffer
+ GLuint64EXT vertex_buffer_address = 0;
+
/// Display information for Switch screen
ScreenInfo screen_info;
generated by cgit v1.2.3 (git 2.25.1) at 2024-09-11 12:50:41 +0200