1
0
Fork 0
mirror of https://github.com/Atmosphere-NX/Atmosphere.git synced 2024-11-18 09:56:40 +00:00
Atmosphere/troposphere/daybreak/nanovg/source/dk_renderer.cpp

545 lines
22 KiB
C++
Raw Normal View History

#include "dk_renderer.hpp"
#include <stdarg.h>
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <math.h>
#include <switch.h>
#define GLM_FORCE_DEFAULT_ALIGNED_GENTYPES /* Enforces GLSL std140/std430 alignment rules for glm types. */
#define GLM_FORCE_INTRINSICS /* Enables usage of SIMD CPU instructions (requiring the above as well). */
#include <glm/vec2.hpp>
namespace nvg {
namespace {
constexpr std::array VertexBufferState = { DkVtxBufferState{sizeof(NVGvertex), 0}, };
constexpr std::array VertexAttribState = {
DkVtxAttribState{0, 0, offsetof(NVGvertex, x), DkVtxAttribSize_2x32, DkVtxAttribType_Float, 0},
DkVtxAttribState{0, 0, offsetof(NVGvertex, u), DkVtxAttribSize_2x32, DkVtxAttribType_Float, 0},
};
struct View {
glm::vec2 size;
};
void UpdateImage(dk::Image &image, CMemPool &scratchPool, dk::Device device, dk::Queue transferQueue, int type, int x, int y, int w, int h, const u8 *data) {
/* Do not proceed if no data is provided upfront. */
if (data == nullptr) {
return;
}
/* Allocate memory from the pool for the image. */
const size_t imageSize = type == NVG_TEXTURE_RGBA ? w * h * 4 : w * h;
CMemPool::Handle tempimgmem = scratchPool.allocate(imageSize, DK_IMAGE_LINEAR_STRIDE_ALIGNMENT);
memcpy(tempimgmem.getCpuAddr(), data, imageSize);
dk::UniqueCmdBuf tempcmdbuf = dk::CmdBufMaker{device}.create();
CMemPool::Handle tempcmdmem = scratchPool.allocate(DK_MEMBLOCK_ALIGNMENT);
tempcmdbuf.addMemory(tempcmdmem.getMemBlock(), tempcmdmem.getOffset(), tempcmdmem.getSize());
dk::ImageView imageView{image};
tempcmdbuf.copyBufferToImage({ tempimgmem.getGpuAddr() }, imageView, { static_cast<uint32_t>(x), static_cast<uint32_t>(y), 0, static_cast<uint32_t>(w), static_cast<uint32_t>(h), 1 });
transferQueue.submitCommands(tempcmdbuf.finishList());
transferQueue.waitIdle();
/* Destroy temp mem. */
tempcmdmem.destroy();
tempimgmem.destroy();
}
}
Texture::Texture(int id) : m_id(id) { /* ... */ }
Texture::~Texture() {
m_image_mem.destroy();
}
void Texture::Initialize(CMemPool &image_pool, CMemPool &scratch_pool, dk::Device device, dk::Queue queue, int type, int w, int h, int image_flags, const u8 *data) {
m_texture_descriptor = {
.width = w,
.height = h,
.type = type,
.flags = image_flags,
};
/* Create an image layout. */
dk::ImageLayout layout;
auto layout_maker = dk::ImageLayoutMaker{device}.setFlags(0).setDimensions(w, h);
if (type == NVG_TEXTURE_RGBA) {
layout_maker.setFormat(DkImageFormat_RGBA8_Unorm);
} else {
layout_maker.setFormat(DkImageFormat_R8_Unorm);
}
layout_maker.initialize(layout);
/* Initialize image. */
m_image_mem = image_pool.allocate(layout.getSize(), layout.getAlignment());
m_image.initialize(layout, m_image_mem.getMemBlock(), m_image_mem.getOffset());
m_image_descriptor.initialize(m_image);
/* Only update the image if the data isn't null. */
if (data != nullptr) {
UpdateImage(m_image, scratch_pool, device, queue, type, 0, 0, w, h, data);
}
}
int Texture::GetId() {
return m_id;
}
const DKNVGtextureDescriptor &Texture::GetDescriptor() {
return m_texture_descriptor;
}
dk::Image &Texture::GetImage() {
return m_image;
}
dk::ImageDescriptor &Texture::GetImageDescriptor() {
return m_image_descriptor;
}
DkRenderer::DkRenderer(unsigned int view_width, unsigned int view_height, dk::Device device, dk::Queue queue, CMemPool &image_mem_pool, CMemPool &code_mem_pool, CMemPool &data_mem_pool) :
m_view_width(view_width), m_view_height(view_height), m_device(device), m_queue(queue), m_image_mem_pool(image_mem_pool), m_code_mem_pool(code_mem_pool), m_data_mem_pool(data_mem_pool), m_image_descriptor_mappings({0})
{
/* Create a dynamic command buffer and allocate memory for it. */
m_dyn_cmd_buf = dk::CmdBufMaker{m_device}.create();
m_dyn_cmd_mem.allocate(m_data_mem_pool, DynamicCmdSize);
m_image_descriptor_set.allocate(m_data_mem_pool);
m_sampler_descriptor_set.allocate(m_data_mem_pool);
m_view_uniform_buffer = m_data_mem_pool.allocate(sizeof(View), DK_UNIFORM_BUF_ALIGNMENT);
m_frag_uniform_buffer = m_data_mem_pool.allocate(sizeof(FragmentUniformSize), DK_UNIFORM_BUF_ALIGNMENT);
/* Create and bind preset samplers. */
dk::UniqueCmdBuf init_cmd_buf = dk::CmdBufMaker{m_device}.create();
CMemPool::Handle init_cmd_mem = m_data_mem_pool.allocate(DK_MEMBLOCK_ALIGNMENT);
init_cmd_buf.addMemory(init_cmd_mem.getMemBlock(), init_cmd_mem.getOffset(), init_cmd_mem.getSize());
for (u8 i = 0; i < SamplerType_Total; i++) {
const DkFilter filter = (i & SamplerType_Nearest) ? DkFilter_Nearest : DkFilter_Linear;
const DkMipFilter mip_filter = (i & SamplerType_Nearest) ? DkMipFilter_Nearest : DkMipFilter_Linear;
const DkWrapMode u_wrap_mode = (i & SamplerType_RepeatX) ? DkWrapMode_Repeat : DkWrapMode_ClampToEdge;
const DkWrapMode v_wrap_mode = (i & SamplerType_RepeatY) ? DkWrapMode_Repeat : DkWrapMode_ClampToEdge;
auto sampler = dk::Sampler{};
auto sampler_descriptor = dk::SamplerDescriptor{};
sampler.setFilter(filter, filter, (i & SamplerType_MipFilter) ? mip_filter : DkMipFilter_None);
sampler.setWrapMode(u_wrap_mode, v_wrap_mode);
sampler_descriptor.initialize(sampler);
m_sampler_descriptor_set.update(init_cmd_buf, i, sampler_descriptor);
}
/* Flush the descriptor cache. */
init_cmd_buf.barrier(DkBarrier_None, DkInvalidateFlags_Descriptors);
m_sampler_descriptor_set.bindForSamplers(init_cmd_buf);
m_image_descriptor_set.bindForImages(init_cmd_buf);
m_queue.submitCommands(init_cmd_buf.finishList());
m_queue.waitIdle();
init_cmd_mem.destroy();
init_cmd_buf.destroy();
}
DkRenderer::~DkRenderer() {
if (m_vertex_buffer) {
m_vertex_buffer->destroy();
}
m_view_uniform_buffer.destroy();
m_frag_uniform_buffer.destroy();
m_textures.clear();
}
int DkRenderer::AcquireImageDescriptor(std::shared_ptr<Texture> texture, int image) {
int free_image_descriptor = m_last_image_descriptor + 1;
int mapping = 0;
for (int desc = 0; desc <= m_last_image_descriptor; desc++) {
mapping = m_image_descriptor_mappings[desc];
/* We've found the image descriptor requested. */
if (mapping == image) {
return desc;
}
/* Update the free image descriptor. */
if (mapping == 0 && free_image_descriptor == m_last_image_descriptor + 1) {
free_image_descriptor = desc;
}
}
/* No descriptors are free. */
if (free_image_descriptor >= static_cast<int>(MaxImages)) {
return -1;
}
/* Update descriptor sets. */
m_image_descriptor_set.update(m_dyn_cmd_buf, free_image_descriptor, texture->GetImageDescriptor());
/* Flush the descriptor cache. */
m_dyn_cmd_buf.barrier(DkBarrier_None, DkInvalidateFlags_Descriptors);
/* Update the map. */
m_image_descriptor_mappings[free_image_descriptor] = image;
m_last_image_descriptor = free_image_descriptor;
return free_image_descriptor;
}
void DkRenderer::FreeImageDescriptor(int image) {
for (int desc = 0; desc <= m_last_image_descriptor; desc++) {
if (m_image_descriptor_mappings[desc] == image) {
m_image_descriptor_mappings[desc] = 0;
}
}
}
void DkRenderer::UpdateVertexBuffer(const void *data, size_t size) {
/* Destroy the existing vertex buffer if it is too small. */
if (m_vertex_buffer && m_vertex_buffer->getSize() < size) {
m_vertex_buffer->destroy();
m_vertex_buffer.reset();
}
/* Create a new buffer if needed. */
if (!m_vertex_buffer) {
m_vertex_buffer = m_data_mem_pool.allocate(size);
}
/* Copy data to the vertex buffer if it exists. */
if (m_vertex_buffer) {
memcpy(m_vertex_buffer->getCpuAddr(), data, size);
}
}
void DkRenderer::SetUniforms(const DKNVGcontext &ctx, int offset, int image) {
m_dyn_cmd_buf.pushConstants(m_frag_uniform_buffer.getGpuAddr(), m_frag_uniform_buffer.getSize(), 0, ctx.fragSize, ctx.uniforms + offset);
m_dyn_cmd_buf.bindUniformBuffer(DkStage_Fragment, 0, m_frag_uniform_buffer.getGpuAddr(), m_frag_uniform_buffer.getSize());
/* Attempt to find a texture. */
const auto texture = this->FindTexture(image);
if (texture == nullptr) {
return;
}
/* Acquire an image descriptor. */
const int image_desc_id = this->AcquireImageDescriptor(texture, image);
if (image_desc_id == -1) {
return;
}
const int image_flags = texture->GetDescriptor().flags;
uint32_t sampler_id = 0;
if (image_flags & NVG_IMAGE_GENERATE_MIPMAPS) sampler_id |= SamplerType_MipFilter;
if (image_flags & NVG_IMAGE_NEAREST) sampler_id |= SamplerType_Nearest;
if (image_flags & NVG_IMAGE_REPEATX) sampler_id |= SamplerType_RepeatX;
if (image_flags & NVG_IMAGE_REPEATY) sampler_id |= SamplerType_RepeatY;
m_dyn_cmd_buf.bindTextures(DkStage_Fragment, 0, dkMakeTextureHandle(image_desc_id, sampler_id));
}
void DkRenderer::DrawFill(const DKNVGcontext &ctx, const DKNVGcall &call) {
DKNVGpath *paths = &ctx.paths[call.pathOffset];
int npaths = call.pathCount;
/* Set the stencils to be used. */
m_dyn_cmd_buf.setStencil(DkFace_FrontAndBack, 0xFF, 0x0, 0xFF);
/* Set the depth stencil state. */
auto depth_stencil_state = dk::DepthStencilState{}
.setStencilTestEnable(true)
.setStencilFrontCompareOp(DkCompareOp_Always)
.setStencilFrontFailOp(DkStencilOp_Keep)
.setStencilFrontDepthFailOp(DkStencilOp_Keep)
.setStencilFrontPassOp(DkStencilOp_IncrWrap)
.setStencilBackCompareOp(DkCompareOp_Always)
.setStencilBackFailOp(DkStencilOp_Keep)
.setStencilBackDepthFailOp(DkStencilOp_Keep)
.setStencilBackPassOp(DkStencilOp_DecrWrap);
m_dyn_cmd_buf.bindDepthStencilState(depth_stencil_state);
/* Configure for shape drawing. */
m_dyn_cmd_buf.bindColorWriteState(dk::ColorWriteState{}.setMask(0, 0));
this->SetUniforms(ctx, call.uniformOffset, 0);
m_dyn_cmd_buf.bindRasterizerState(dk::RasterizerState{}.setCullMode(DkFace_None));
/* Draw vertices. */
for (int i = 0; i < npaths; i++) {
m_dyn_cmd_buf.draw(DkPrimitive_TriangleFan, paths[i].fillCount, 1, paths[i].fillOffset, 0);
}
m_dyn_cmd_buf.bindColorWriteState(dk::ColorWriteState{});
this->SetUniforms(ctx, call.uniformOffset + ctx.fragSize, call.image);
m_dyn_cmd_buf.bindRasterizerState(dk::RasterizerState{});
if (ctx.flags & NVG_ANTIALIAS) {
/* Configure stencil anti-aliasing. */
depth_stencil_state
.setStencilFrontCompareOp(DkCompareOp_Equal)
.setStencilFrontFailOp(DkStencilOp_Keep)
.setStencilFrontDepthFailOp(DkStencilOp_Keep)
.setStencilFrontPassOp(DkStencilOp_Keep)
.setStencilBackCompareOp(DkCompareOp_Equal)
.setStencilBackFailOp(DkStencilOp_Keep)
.setStencilBackDepthFailOp(DkStencilOp_Keep)
.setStencilBackPassOp(DkStencilOp_Keep);
m_dyn_cmd_buf.bindDepthStencilState(depth_stencil_state);
/* Draw fringes. */
for (int i = 0; i < npaths; i++) {
m_dyn_cmd_buf.draw(DkPrimitive_TriangleStrip, paths[i].strokeCount, 1, paths[i].strokeOffset, 0);
}
}
/* Configure and draw fill. */
depth_stencil_state
.setStencilFrontCompareOp(DkCompareOp_NotEqual)
.setStencilFrontFailOp(DkStencilOp_Zero)
.setStencilFrontDepthFailOp(DkStencilOp_Zero)
.setStencilFrontPassOp(DkStencilOp_Zero)
.setStencilBackCompareOp(DkCompareOp_NotEqual)
.setStencilBackFailOp(DkStencilOp_Zero)
.setStencilBackDepthFailOp(DkStencilOp_Zero)
.setStencilBackPassOp(DkStencilOp_Zero);
m_dyn_cmd_buf.bindDepthStencilState(depth_stencil_state);
m_dyn_cmd_buf.draw(DkPrimitive_TriangleStrip, call.triangleCount, 1, call.triangleOffset, 0);
/* Reset the depth stencil state to default. */
m_dyn_cmd_buf.bindDepthStencilState(dk::DepthStencilState{});
}
void DkRenderer::DrawConvexFill(const DKNVGcontext &ctx, const DKNVGcall &call) {
DKNVGpath *paths = &ctx.paths[call.pathOffset];
int npaths = call.pathCount;
this->SetUniforms(ctx, call.uniformOffset, call.image);
for (int i = 0; i < npaths; i++) {
m_dyn_cmd_buf.draw(DkPrimitive_TriangleFan, paths[i].fillCount, 1, paths[i].fillOffset, 0);
/* Draw fringes. */
if (paths[i].strokeCount > 0) {
m_dyn_cmd_buf.draw(DkPrimitive_TriangleStrip, paths[i].strokeCount, 1, paths[i].strokeOffset, 0);
}
}
}
void DkRenderer::DrawStroke(const DKNVGcontext &ctx, const DKNVGcall &call) {
DKNVGpath* paths = &ctx.paths[call.pathOffset];
int npaths = call.pathCount;
if (ctx.flags & NVG_STENCIL_STROKES) {
/* Set the stencil to be used. */
m_dyn_cmd_buf.setStencil(DkFace_Front, 0xFF, 0x0, 0xFF);
/* Configure for filling the stroke base without overlap. */
auto depth_stencil_state = dk::DepthStencilState{}
.setStencilTestEnable(true)
.setStencilFrontCompareOp(DkCompareOp_Equal)
.setStencilFrontFailOp(DkStencilOp_Keep)
.setStencilFrontDepthFailOp(DkStencilOp_Keep)
.setStencilFrontPassOp(DkStencilOp_Incr);
m_dyn_cmd_buf.bindDepthStencilState(depth_stencil_state);
this->SetUniforms(ctx, call.uniformOffset + ctx.fragSize, call.image);
/* Draw vertices. */
for (int i = 0; i < npaths; i++) {
m_dyn_cmd_buf.draw(DkPrimitive_TriangleStrip, paths[i].strokeCount, 1, paths[i].strokeOffset, 0);
}
/* Configure for drawing anti-aliased pixels. */
depth_stencil_state.setStencilFrontPassOp(DkStencilOp_Keep);
m_dyn_cmd_buf.bindDepthStencilState(depth_stencil_state);
this->SetUniforms(ctx, call.uniformOffset, call.image);
/* Draw vertices. */
for (int i = 0; i < npaths; i++) {
m_dyn_cmd_buf.draw(DkPrimitive_TriangleStrip, paths[i].strokeCount, 1, paths[i].strokeOffset, 0);
}
/* Configure for clearing the stencil buffer. */
depth_stencil_state
.setStencilTestEnable(true)
.setStencilFrontCompareOp(DkCompareOp_Always)
.setStencilFrontFailOp(DkStencilOp_Zero)
.setStencilFrontDepthFailOp(DkStencilOp_Zero)
.setStencilFrontPassOp(DkStencilOp_Zero);
m_dyn_cmd_buf.bindDepthStencilState(depth_stencil_state);
/* Draw vertices. */
for (int i = 0; i < npaths; i++) {
m_dyn_cmd_buf.draw(DkPrimitive_TriangleStrip, paths[i].strokeCount, 1, paths[i].strokeOffset, 0);
}
/* Reset the depth stencil state to default. */
m_dyn_cmd_buf.bindDepthStencilState(dk::DepthStencilState{});
} else {
this->SetUniforms(ctx, call.uniformOffset, call.image);
/* Draw vertices. */
for (int i = 0; i < npaths; i++) {
m_dyn_cmd_buf.draw(DkPrimitive_TriangleStrip, paths[i].strokeCount, 1, paths[i].strokeOffset, 0);
}
}
}
void DkRenderer::DrawTriangles(const DKNVGcontext &ctx, const DKNVGcall &call) {
this->SetUniforms(ctx, call.uniformOffset, call.image);
m_dyn_cmd_buf.draw(DkPrimitive_Triangles, call.triangleCount, 1, call.triangleOffset, 0);
}
int DkRenderer::Create(DKNVGcontext &ctx) {
m_vertex_shader.load(m_code_mem_pool, "romfs:/shaders/fill_vsh.dksh");
/* Load the appropriate fragment shader depending on whether AA is enabled. */
if (ctx.flags & NVG_ANTIALIAS) {
m_fragment_shader.load(m_code_mem_pool, "romfs:/shaders/fill_aa_fsh.dksh");
} else {
m_fragment_shader.load(m_code_mem_pool, "romfs:/shaders/fill_fsh.dksh");
}
/* Set the size of fragment uniforms. */
ctx.fragSize = FragmentUniformSize;
return 1;
}
std::shared_ptr<Texture> DkRenderer::FindTexture(int id) {
for (auto it = m_textures.begin(); it != m_textures.end(); it++) {
if ((*it)->GetId() == id) {
return *it;
}
}
return nullptr;
}
int DkRenderer::CreateTexture(const DKNVGcontext &ctx, int type, int w, int h, int image_flags, const unsigned char* data) {
const auto texture_id = m_next_texture_id++;
auto texture = std::make_shared<Texture>(texture_id);
texture->Initialize(m_image_mem_pool, m_data_mem_pool, m_device, m_queue, type, w, h, image_flags, data);
m_textures.push_back(texture);
return texture->GetId();
}
int DkRenderer::DeleteTexture(const DKNVGcontext &ctx, int image) {
bool found = false;
for (auto it = m_textures.begin(); it != m_textures.end();) {
/* Remove textures with the given id. */
if ((*it)->GetId() == image) {
it = m_textures.erase(it);
found = true;
} else {
++it;
}
}
/* Free any used image descriptors. */
this->FreeImageDescriptor(image);
return found;
}
int DkRenderer::UpdateTexture(const DKNVGcontext &ctx, int image, int x, int y, int w, int h, const unsigned char *data) {
const std::shared_ptr<Texture> texture = this->FindTexture(image);
/* Could not find a texture. */
if (texture == nullptr) {
return 0;
}
const DKNVGtextureDescriptor &tex_desc = texture->GetDescriptor();
if (tex_desc.type == NVG_TEXTURE_RGBA) {
data += y * tex_desc.width*4;
} else {
data += y * tex_desc.width;
}
x = 0;
w = tex_desc.width;
UpdateImage(texture->GetImage(), m_data_mem_pool, m_device, m_queue, tex_desc.type, x, y, w, h, data);
return 1;
}
int DkRenderer::GetTextureSize(const DKNVGcontext &ctx, int image, int *w, int *h) {
const auto descriptor = this->GetTextureDescriptor(ctx, image);
if (descriptor == nullptr) {
return 0;
}
*w = descriptor->width;
*h = descriptor->height;
return 1;
}
const DKNVGtextureDescriptor *DkRenderer::GetTextureDescriptor(const DKNVGcontext &ctx, int id) {
for (auto it = m_textures.begin(); it != m_textures.end(); it++) {
if ((*it)->GetId() == id) {
return &(*it)->GetDescriptor();
}
}
return nullptr;
}
void DkRenderer::Flush(DKNVGcontext &ctx) {
if (ctx.ncalls > 0) {
/* Prepare dynamic command buffer. */
m_dyn_cmd_mem.begin(m_dyn_cmd_buf);
/* Update buffers with data. */
this->UpdateVertexBuffer(ctx.verts, ctx.nverts * sizeof(NVGvertex));
/* Enable blending. */
m_dyn_cmd_buf.bindColorState(dk::ColorState{}.setBlendEnable(0, true));
/* Setup. */
m_dyn_cmd_buf.bindShaders(DkStageFlag_GraphicsMask, { m_vertex_shader, m_fragment_shader });
m_dyn_cmd_buf.bindVtxAttribState(VertexAttribState);
m_dyn_cmd_buf.bindVtxBufferState(VertexBufferState);
m_dyn_cmd_buf.bindVtxBuffer(0, m_vertex_buffer->getGpuAddr(), m_vertex_buffer->getSize());
/* Push the view size to the uniform buffer and bind it. */
const auto view = View{glm::vec2{m_view_width, m_view_height}};
m_dyn_cmd_buf.pushConstants(m_view_uniform_buffer.getGpuAddr(), m_view_uniform_buffer.getSize(), 0, sizeof(view), &view);
m_dyn_cmd_buf.bindUniformBuffer(DkStage_Vertex, 0, m_view_uniform_buffer.getGpuAddr(), m_view_uniform_buffer.getSize());
/* Iterate over calls. */
for (int i = 0; i < ctx.ncalls; i++) {
const DKNVGcall &call = ctx.calls[i];
/* Perform blending. */
m_dyn_cmd_buf.bindBlendStates(0, { dk::BlendState{}.setFactors(static_cast<DkBlendFactor>(call.blendFunc.srcRGB), static_cast<DkBlendFactor>(call.blendFunc.dstRGB), static_cast<DkBlendFactor>(call.blendFunc.srcAlpha), static_cast<DkBlendFactor>(call.blendFunc.dstRGB)) });
if (call.type == DKNVG_FILL) {
this->DrawFill(ctx, call);
} else if (call.type == DKNVG_CONVEXFILL) {
this->DrawConvexFill(ctx, call);
} else if (call.type == DKNVG_STROKE) {
this->DrawStroke(ctx, call);
} else if (call.type == DKNVG_TRIANGLES) {
this->DrawTriangles(ctx, call);
}
}
m_queue.submitCommands(m_dyn_cmd_mem.end(m_dyn_cmd_buf));
}
/* Reset calls. */
ctx.nverts = 0;
ctx.npaths = 0;
ctx.ncalls = 0;
ctx.nuniforms = 0;
}
}