Merge pull request #342 from bunnei/indexed-verts

Implement indexed mode rendering
This commit is contained in:
bunnei 2018-04-17 11:34:22 -04:00 committed by GitHub
commit 0905dc1ff4
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
5 changed files with 98 additions and 28 deletions

View File

@ -165,6 +165,7 @@ void Maxwell3D::ProcessQueryGet() {
void Maxwell3D::DrawArrays() {
LOG_DEBUG(HW_GPU, "called, topology=%d, count=%d", regs.draw.topology.Value(),
regs.vertex_buffer.count);
ASSERT_MSG(!(regs.index_array.count && regs.vertex_buffer.count), "Both indexed and direct?");
auto debug_context = Core::System::GetInstance().GetGPUDebugContext();
@ -176,7 +177,8 @@ void Maxwell3D::DrawArrays() {
debug_context->OnEvent(Tegra::DebugContext::Event::FinishedPrimitiveBatch, nullptr);
}
VideoCore::g_renderer->Rasterizer()->AccelerateDrawBatch(false /*is_indexed*/);
const bool is_indexed{regs.index_array.count && !regs.vertex_buffer.count};
VideoCore::g_renderer->Rasterizer()->AccelerateDrawBatch(is_indexed);
}
void Maxwell3D::ProcessCBBind(Regs::ShaderStage stage) {

View File

@ -248,6 +248,12 @@ public:
Patches = 0xe,
};
enum class IndexFormat : u32 {
UnsignedByte = 0x0,
UnsignedShort = 0x1,
UnsignedInt = 0x2,
};
union {
struct {
INSERT_PADDING_WORDS(0x200);
@ -375,7 +381,42 @@ public:
};
} draw;
INSERT_PADDING_WORDS(0x139);
INSERT_PADDING_WORDS(0x6B);
struct {
u32 start_addr_high;
u32 start_addr_low;
u32 end_addr_high;
u32 end_addr_low;
IndexFormat format;
u32 first;
u32 count;
unsigned FormatSizeInBytes() const {
switch (format) {
case IndexFormat::UnsignedByte:
return 1;
case IndexFormat::UnsignedShort:
return 2;
case IndexFormat::UnsignedInt:
return 4;
}
UNREACHABLE();
}
GPUVAddr StartAddress() const {
return static_cast<GPUVAddr>(
(static_cast<GPUVAddr>(start_addr_high) << 32) | start_addr_low);
}
GPUVAddr EndAddress() const {
return static_cast<GPUVAddr>((static_cast<GPUVAddr>(end_addr_high) << 32) |
end_addr_low);
}
} index_array;
INSERT_PADDING_WORDS(0xC7);
struct {
u32 query_address_high;
u32 query_address_low;
@ -572,6 +613,7 @@ ASSERT_REG_POSITION(tsc, 0x557);
ASSERT_REG_POSITION(tic, 0x55D);
ASSERT_REG_POSITION(code_address, 0x582);
ASSERT_REG_POSITION(draw, 0x585);
ASSERT_REG_POSITION(index_array, 0x5F2);
ASSERT_REG_POSITION(query, 0x6C0);
ASSERT_REG_POSITION(vertex_array[0], 0x700);
ASSERT_REG_POSITION(vertex_array_limit[0], 0x7C0);

View File

@ -97,7 +97,6 @@ RasterizerOpenGL::RasterizerOpenGL() {
state.draw.vertex_buffer = stream_buffer->GetHandle();
shader_program_manager = std::make_unique<GLShader::ProgramManager>();
state.draw.shader_program = 0;
state.draw.vertex_array = hw_vao.handle;
state.Apply();
@ -128,17 +127,6 @@ RasterizerOpenGL::~RasterizerOpenGL() {
}
}
void RasterizerOpenGL::AnalyzeVertexArray(bool is_indexed) {
const auto& regs = Core::System().GetInstance().GPU().Maxwell3D().regs;
if (is_indexed) {
UNREACHABLE();
}
// TODO(bunnei): Add support for 1+ vertex arrays
vs_input_size = regs.vertex_buffer.count * regs.vertex_array[0].stride;
}
void RasterizerOpenGL::SetupVertexArray(u8* array_ptr, GLintptr buffer_offset) {
MICROPROFILE_SCOPE(OpenGL_VAO);
const auto& regs = Core::System().GetInstance().GPU().Maxwell3D().regs;
@ -150,6 +138,7 @@ void RasterizerOpenGL::SetupVertexArray(u8* array_ptr, GLintptr buffer_offset) {
// TODO(bunnei): Add support for 1+ vertex arrays
const auto& vertex_array{regs.vertex_array[0]};
const auto& vertex_array_limit{regs.vertex_array_limit[0]};
ASSERT_MSG(vertex_array.enable, "vertex array 0 is disabled?");
ASSERT_MSG(!vertex_array.divisor, "vertex array 0 divisor is unimplemented!");
for (unsigned index = 1; index < Maxwell::NumVertexArrays; ++index) {
@ -162,6 +151,10 @@ void RasterizerOpenGL::SetupVertexArray(u8* array_ptr, GLintptr buffer_offset) {
// to avoid OpenGL errors.
for (unsigned index = 0; index < 16; ++index) {
auto& attrib = regs.vertex_attrib_format[index];
NGLOG_DEBUG(HW_GPU, "vertex attrib {}, count={}, size={}, type={}, offset={}, normalize={}",
index, attrib.ComponentCount(), attrib.SizeString(), attrib.TypeString(),
attrib.offset.Value(), attrib.IsNormalized());
glVertexAttribPointer(index, attrib.ComponentCount(), MaxwellToGL::VertexType(attrib),
attrib.IsNormalized() ? GL_TRUE : GL_FALSE, vertex_array.stride,
reinterpret_cast<GLvoid*>(buffer_offset + attrib.offset));
@ -170,7 +163,7 @@ void RasterizerOpenGL::SetupVertexArray(u8* array_ptr, GLintptr buffer_offset) {
}
// Copy vertex array data
const u32 data_size{vertex_array.stride * regs.vertex_buffer.count};
const u64 data_size{vertex_array_limit.LimitAddress() - vertex_array.StartAddress() + 1};
const VAddr data_addr{memory_manager->PhysicalToVirtualAddress(vertex_array.StartAddress())};
res_cache.FlushRegion(data_addr, data_size, nullptr);
Memory::ReadBlock(data_addr, array_ptr, data_size);
@ -333,13 +326,18 @@ void RasterizerOpenGL::DrawArrays() {
// Draw the vertex batch
const bool is_indexed = accelerate_draw == AccelDraw::Indexed;
AnalyzeVertexArray(is_indexed);
const u64 index_buffer_size{regs.index_array.count * regs.index_array.FormatSizeInBytes()};
const unsigned vertex_num{is_indexed ? regs.index_array.count : regs.vertex_buffer.count};
// TODO(bunnei): Add support for 1+ vertex arrays
vs_input_size = vertex_num * regs.vertex_array[0].stride;
state.draw.vertex_buffer = stream_buffer->GetHandle();
state.Apply();
size_t buffer_size = static_cast<size_t>(vs_input_size);
if (is_indexed) {
UNREACHABLE();
buffer_size = Common::AlignUp(buffer_size, 4) + index_buffer_size;
}
// Uniform space for the 5 shader stages
@ -354,9 +352,18 @@ void RasterizerOpenGL::DrawArrays() {
SetupVertexArray(buffer_ptr, buffer_offset);
ptr_pos += vs_input_size;
// If indexed mode, copy the index buffer
GLintptr index_buffer_offset = 0;
if (is_indexed) {
UNREACHABLE();
ptr_pos = Common::AlignUp(ptr_pos, 4);
const auto& memory_manager = Core::System().GetInstance().GPU().memory_manager;
const VAddr index_data_addr{
memory_manager->PhysicalToVirtualAddress(regs.index_array.StartAddress())};
Memory::ReadBlock(index_data_addr, &buffer_ptr[ptr_pos], index_buffer_size);
index_buffer_offset = buffer_offset + static_cast<GLintptr>(ptr_pos);
ptr_pos += index_buffer_size;
}
SetupShaders(buffer_ptr, buffer_offset, ptr_pos);
@ -366,11 +373,16 @@ void RasterizerOpenGL::DrawArrays() {
shader_program_manager->ApplyTo(state);
state.Apply();
const GLenum primitive_mode{MaxwellToGL::PrimitiveTopology(regs.draw.topology)};
if (is_indexed) {
UNREACHABLE();
const GLint index_min{static_cast<GLint>(regs.index_array.first)};
const GLint index_max{static_cast<GLint>(regs.index_array.first + regs.index_array.count)};
glDrawRangeElementsBaseVertex(primitive_mode, index_min, index_max, regs.index_array.count,
MaxwellToGL::IndexFormat(regs.index_array.format),
reinterpret_cast<const void*>(index_buffer_offset),
-index_min);
} else {
glDrawArrays(MaxwellToGL::PrimitiveTopology(regs.draw.topology), 0,
regs.vertex_buffer.count);
glDrawArrays(primitive_mode, 0, regs.vertex_buffer.count);
}
// Disable scissor test

View File

@ -155,7 +155,6 @@ private:
GLsizeiptr vs_input_size;
void AnalyzeVertexArray(bool is_indexed);
void SetupVertexArray(u8* array_ptr, GLintptr buffer_offset);
std::array<OGLBuffer, Tegra::Engines::Maxwell3D::Regs::MaxShaderStage> uniform_buffers;

View File

@ -31,7 +31,7 @@ inline GLenum VertexType(Maxwell::VertexAttribute attrib) {
return GL_UNSIGNED_BYTE;
}
LOG_CRITICAL(Render_OpenGL, "Unimplemented vertex size=%s", attrib.SizeString().c_str());
NGLOG_CRITICAL(Render_OpenGL, "Unimplemented vertex size={}", attrib.SizeString());
UNREACHABLE();
return {};
}
@ -40,7 +40,21 @@ inline GLenum VertexType(Maxwell::VertexAttribute attrib) {
return GL_FLOAT;
}
LOG_CRITICAL(Render_OpenGL, "Unimplemented vertex type=%s", attrib.TypeString().c_str());
NGLOG_CRITICAL(Render_OpenGL, "Unimplemented vertex type={}", attrib.TypeString());
UNREACHABLE();
return {};
}
inline GLenum IndexFormat(Maxwell::IndexFormat index_format) {
switch (index_format) {
case Maxwell::IndexFormat::UnsignedByte:
return GL_UNSIGNED_BYTE;
case Maxwell::IndexFormat::UnsignedShort:
return GL_UNSIGNED_SHORT;
case Maxwell::IndexFormat::UnsignedInt:
return GL_UNSIGNED_INT;
}
NGLOG_CRITICAL(Render_OpenGL, "Unimplemented index_format={}", static_cast<u32>(index_format));
UNREACHABLE();
return {};
}
@ -52,7 +66,7 @@ inline GLenum PrimitiveTopology(Maxwell::PrimitiveTopology topology) {
case Maxwell::PrimitiveTopology::TriangleStrip:
return GL_TRIANGLE_STRIP;
}
LOG_CRITICAL(Render_OpenGL, "Unimplemented primitive topology=%d", topology);
NGLOG_CRITICAL(Render_OpenGL, "Unimplemented topology={}", static_cast<u32>(topology));
UNREACHABLE();
return {};
}
@ -64,7 +78,7 @@ inline GLenum TextureFilterMode(Tegra::Texture::TextureFilter filter_mode) {
case Tegra::Texture::TextureFilter::Nearest:
return GL_NEAREST;
}
LOG_CRITICAL(Render_OpenGL, "Unimplemented texture filter mode=%u",
NGLOG_CRITICAL(Render_OpenGL, "Unimplemented texture filter mode={}",
static_cast<u32>(filter_mode));
UNREACHABLE();
return {};
@ -75,7 +89,8 @@ inline GLenum WrapMode(Tegra::Texture::WrapMode wrap_mode) {
case Tegra::Texture::WrapMode::ClampToEdge:
return GL_CLAMP_TO_EDGE;
}
LOG_CRITICAL(Render_OpenGL, "Unimplemented texture wrap mode=%u", static_cast<u32>(wrap_mode));
NGLOG_CRITICAL(Render_OpenGL, "Unimplemented texture wrap mode={}",
static_cast<u32>(wrap_mode));
UNREACHABLE();
return {};
}