mirrors/shadPS4
1
0
Fork 0
mirror of https://github.com/shadps4-emu/shadPS4.git synced 2025-01-07 15:46:01 +00:00
Code Issues Actions 4 Packages Projects Releases Wiki Activity

renderer_vulkan: Add support for indexed QuadList draw. (#1661)

Browse source
This commit is contained in:
squidbus 2024-12-04 01:15:58 -08:00 committed by GitHub
parent be9c797003
commit f8041b8446
3 changed files with 54 additions and 15 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

28
src/video_core/buffer_cache/buffer_cache.cpp
View file

@ -236,7 +236,7 @@ bool BufferCache::BindVertexBuffers(const Shader::Info& vs_info) {
u32 BufferCache::BindIndexBuffer(bool& is_indexed, u32 index_offset) { u32 BufferCache::BindIndexBuffer(bool& is_indexed, u32 index_offset) {
// Emulate QuadList primitive type with CPU made index buffer. // Emulate QuadList primitive type with CPU made index buffer.
const auto& regs = liverpool->regs; const auto& regs = liverpool->regs;
if (regs.primitive_type == AmdGpu::PrimitiveType::QuadList) { if (regs.primitive_type == AmdGpu::PrimitiveType::QuadList && !is_indexed) {
is_indexed = true; is_indexed = true;
// Emit indices. // Emit indices.
@ -262,6 +262,32 @@ u32 BufferCache::BindIndexBuffer(bool& is_indexed, u32 index_offset) {
VAddr index_address = regs.index_base_address.Address<VAddr>(); VAddr index_address = regs.index_base_address.Address<VAddr>();
index_address += index_offset * index_size; index_address += index_offset * index_size;
if (regs.primitive_type == AmdGpu::PrimitiveType::QuadList) {
// Convert indices.
const u32 new_index_size = regs.num_indices * index_size * 6 / 4;
const auto [data, offset] = stream_buffer.Map(new_index_size);
const auto index_ptr = reinterpret_cast<u8*>(index_address);
switch (index_type) {
case vk::IndexType::eUint16:
Vulkan::LiverpoolToVK::ConvertQuadToTriangleListIndices<u16>(data, index_ptr,
regs.num_indices);
break;
case vk::IndexType::eUint32:
Vulkan::LiverpoolToVK::ConvertQuadToTriangleListIndices<u32>(data, index_ptr,
regs.num_indices);
break;
default:
UNREACHABLE_MSG("Unsupported QuadList index type {}", vk::to_string(index_type));
break;
}
stream_buffer.Commit();
// Bind index buffer.
const auto cmdbuf = scheduler.CommandBuffer();
cmdbuf.bindIndexBuffer(stream_buffer.Handle(), offset, index_type);
return new_index_size / index_size;
}
// Bind index buffer. // Bind index buffer.
const u32 index_buffer_size = regs.num_indices * index_size; const u32 index_buffer_size = regs.num_indices * index_size;
const auto [vk_buffer, offset] = ObtainBuffer(index_address, index_buffer_size, false); const auto [vk_buffer, offset] = ObtainBuffer(index_address, index_buffer_size, false);

13
src/video_core/renderer_vulkan/liverpool_to_vk.cpp
View file

@ -726,19 +726,6 @@ vk::Format DepthFormat(DepthBuffer::ZFormat z_format, DepthBuffer::StencilFormat
return format->vk_format; return format->vk_format;
} }
void EmitQuadToTriangleListIndices(u8* out_ptr, u32 num_vertices) {
static constexpr u16 NumVerticesPerQuad = 4;
u16* out_data = reinterpret_cast<u16*>(out_ptr);
for (u16 i = 0; i < num_vertices; i += NumVerticesPerQuad) {
*out_data++ = i;
*out_data++ = i + 1;
*out_data++ = i + 2;
*out_data++ = i;
*out_data++ = i + 2;
*out_data++ = i + 3;
}
}
vk::ClearValue ColorBufferClearValue(const AmdGpu::Liverpool::ColorBuffer& color_buffer) { vk::ClearValue ColorBufferClearValue(const AmdGpu::Liverpool::ColorBuffer& color_buffer) {
const auto comp_swap = color_buffer.info.comp_swap.Value(); const auto comp_swap = color_buffer.info.comp_swap.Value();
const auto format = color_buffer.info.format.Value(); const auto format = color_buffer.info.format.Value();

28
src/video_core/renderer_vulkan/liverpool_to_vk.h
View file

@ -68,7 +68,33 @@ vk::ClearValue ColorBufferClearValue(const AmdGpu::Liverpool::ColorBuffer& color
vk::SampleCountFlagBits NumSamples(u32 num_samples, vk::SampleCountFlags supported_flags); vk::SampleCountFlagBits NumSamples(u32 num_samples, vk::SampleCountFlags supported_flags);
void EmitQuadToTriangleListIndices(u8* out_indices, u32 num_vertices); static constexpr u16 NumVerticesPerQuad = 4;
inline void EmitQuadToTriangleListIndices(u8* out_ptr, u32 num_vertices) {
u16* out_data = reinterpret_cast<u16*>(out_ptr);
for (u16 i = 0; i < num_vertices; i += NumVerticesPerQuad) {
*out_data++ = i;
*out_data++ = i + 1;
*out_data++ = i + 2;
*out_data++ = i;
*out_data++ = i + 2;
*out_data++ = i + 3;
}
}
template <typename T>
void ConvertQuadToTriangleListIndices(u8* out_ptr, const u8* in_ptr, u32 num_vertices) {
T* out_data = reinterpret_cast<T*>(out_ptr);
const T* in_data = reinterpret_cast<const T*>(in_ptr);
for (u16 i = 0; i < num_vertices; i += NumVerticesPerQuad) {
*out_data++ = in_data[i];
*out_data++ = in_data[i + 1];
*out_data++ = in_data[i + 2];
*out_data++ = in_data[i];
*out_data++ = in_data[i + 2];
*out_data++ = in_data[i + 3];
}
}
static inline vk::Format PromoteFormatToDepth(vk::Format fmt) { static inline vk::Format PromoteFormatToDepth(vk::Format fmt) {
if (fmt == vk::Format::eR32Sfloat) { if (fmt == vk::Format::eR32Sfloat) {