mirror of
https://github.com/shadps4-emu/shadPS4.git
synced 2025-08-02 07:22:24 +00:00
video_core: add vertex_offset to non-indexed draw calls
renamed fetch offset fields
This commit is contained in:
parent
d9ea23badc
commit
a5e468af47
@ -35,7 +35,7 @@ namespace Shader::Gcn {
|
|||||||
**/
|
**/
|
||||||
|
|
||||||
FetchShaderData ParseFetchShader(const u32* code, u32* out_size) {
|
FetchShaderData ParseFetchShader(const u32* code, u32* out_size) {
|
||||||
std::vector<VertexAttribute> attributes;
|
FetchShaderData data{};
|
||||||
GcnCodeSlice code_slice(code, code + std::numeric_limits<u32>::max());
|
GcnCodeSlice code_slice(code, code + std::numeric_limits<u32>::max());
|
||||||
GcnDecodeContext decoder;
|
GcnDecodeContext decoder;
|
||||||
|
|
||||||
@ -46,9 +46,6 @@ FetchShaderData ParseFetchShader(const u32* code, u32* out_size) {
|
|||||||
};
|
};
|
||||||
boost::container::static_vector<VsharpLoad, 16> loads;
|
boost::container::static_vector<VsharpLoad, 16> loads;
|
||||||
|
|
||||||
s8 fetch_index_sgpr = -1;
|
|
||||||
s8 fetch_offset_sgpr = -1;
|
|
||||||
|
|
||||||
u32 semantic_index = 0;
|
u32 semantic_index = 0;
|
||||||
while (!code_slice.atEnd()) {
|
while (!code_slice.atEnd()) {
|
||||||
const auto inst = decoder.decodeInstruction(code_slice);
|
const auto inst = decoder.decodeInstruction(code_slice);
|
||||||
@ -67,11 +64,11 @@ FetchShaderData ParseFetchShader(const u32* code, u32* out_size) {
|
|||||||
const auto vgpr = inst.dst[0].code;
|
const auto vgpr = inst.dst[0].code;
|
||||||
const auto sgpr = s8(inst.src[0].code);
|
const auto sgpr = s8(inst.src[0].code);
|
||||||
switch (vgpr) {
|
switch (vgpr) {
|
||||||
case 0: // V0 is always the index
|
case 0: // V0 is always the vertex offset
|
||||||
fetch_index_sgpr = sgpr;
|
data.vertex_offset_sgpr = sgpr;
|
||||||
break;
|
break;
|
||||||
case 3: // V3 is always the offset
|
case 3: // V3 is always the instance offset
|
||||||
fetch_offset_sgpr = sgpr;
|
data.instance_offset_sgpr = sgpr;
|
||||||
break;
|
break;
|
||||||
default:
|
default:
|
||||||
UNREACHABLE();
|
UNREACHABLE();
|
||||||
@ -87,7 +84,7 @@ FetchShaderData ParseFetchShader(const u32* code, u32* out_size) {
|
|||||||
const auto it = std::ranges::find_if(
|
const auto it = std::ranges::find_if(
|
||||||
loads, [&](VsharpLoad& load) { return load.dst_reg == base_sgpr; });
|
loads, [&](VsharpLoad& load) { return load.dst_reg == base_sgpr; });
|
||||||
|
|
||||||
auto& attrib = attributes.emplace_back();
|
auto& attrib = data.attributes.emplace_back();
|
||||||
attrib.semantic = semantic_index++;
|
attrib.semantic = semantic_index++;
|
||||||
attrib.dest_vgpr = inst.src[1].code;
|
attrib.dest_vgpr = inst.src[1].code;
|
||||||
attrib.num_elements = inst.control.mubuf.count;
|
attrib.num_elements = inst.control.mubuf.count;
|
||||||
@ -102,11 +99,7 @@ FetchShaderData ParseFetchShader(const u32* code, u32* out_size) {
|
|||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
return FetchShaderData{
|
return data;
|
||||||
.attributes = std::move(attributes),
|
|
||||||
.fetch_index_sgpr = fetch_index_sgpr,
|
|
||||||
.fetch_offset_sgpr = fetch_offset_sgpr,
|
|
||||||
};
|
|
||||||
}
|
}
|
||||||
|
|
||||||
} // namespace Shader::Gcn
|
} // namespace Shader::Gcn
|
||||||
|
@ -19,8 +19,8 @@ struct VertexAttribute {
|
|||||||
|
|
||||||
struct FetchShaderData {
|
struct FetchShaderData {
|
||||||
std::vector<VertexAttribute> attributes;
|
std::vector<VertexAttribute> attributes;
|
||||||
s8 fetch_index_sgpr; ///< Read index from VADDR
|
s8 vertex_offset_sgpr = -1; ///< SGPR of vertex offset from VADDR
|
||||||
s8 fetch_offset_sgpr; ///< Read offset from VADDR
|
s8 instance_offset_sgpr = -1; ///< SGPR of instance offset from VADDR
|
||||||
};
|
};
|
||||||
|
|
||||||
FetchShaderData ParseFetchShader(const u32* code, u32* out_size);
|
FetchShaderData ParseFetchShader(const u32* code, u32* out_size);
|
||||||
|
@ -359,8 +359,8 @@ void Translator::EmitFetch(const GcnInst& inst) {
|
|||||||
file.WriteRaw<u8>(code, fetch_size);
|
file.WriteRaw<u8>(code, fetch_size);
|
||||||
}
|
}
|
||||||
|
|
||||||
info.fetch_index_sgpr = fetch_data.fetch_index_sgpr;
|
info.vertex_offset_sgpr = fetch_data.vertex_offset_sgpr;
|
||||||
info.fetch_offset_sgpr = fetch_data.fetch_offset_sgpr;
|
info.instance_offset_sgpr = fetch_data.instance_offset_sgpr;
|
||||||
|
|
||||||
for (const auto& attrib : fetch_data.attributes) {
|
for (const auto& attrib : fetch_data.attributes) {
|
||||||
const IR::Attribute attr{IR::Attribute::Param0 + attrib.semantic};
|
const IR::Attribute attr{IR::Attribute::Param0 + attrib.semantic};
|
||||||
|
@ -175,8 +175,8 @@ struct Info {
|
|||||||
AttributeFlags stores{};
|
AttributeFlags stores{};
|
||||||
boost::container::static_vector<VsOutputMap, 3> vs_outputs;
|
boost::container::static_vector<VsOutputMap, 3> vs_outputs;
|
||||||
|
|
||||||
s8 fetch_index_sgpr = -1;
|
s8 vertex_offset_sgpr = -1;
|
||||||
s8 fetch_offset_sgpr = -1;
|
s8 instance_offset_sgpr = -1;
|
||||||
|
|
||||||
BufferResourceList buffers;
|
BufferResourceList buffers;
|
||||||
ImageResourceList images;
|
ImageResourceList images;
|
||||||
@ -213,6 +213,18 @@ struct Info {
|
|||||||
std::memcpy(&data, base + dword_offset, sizeof(T));
|
std::memcpy(&data, base + dword_offset, sizeof(T));
|
||||||
return data;
|
return data;
|
||||||
}
|
}
|
||||||
|
|
||||||
|
[[nodiscard]] std::pair<u32, u32> GetDrawOffsets() const noexcept {
|
||||||
|
u32 vertex_offset = 0;
|
||||||
|
u32 instance_offset = 0;
|
||||||
|
if (vertex_offset_sgpr != -1) {
|
||||||
|
vertex_offset = user_data[vertex_offset_sgpr];
|
||||||
|
}
|
||||||
|
if (instance_offset_sgpr != -1) {
|
||||||
|
instance_offset = user_data[instance_offset_sgpr];
|
||||||
|
}
|
||||||
|
return {vertex_offset, instance_offset};
|
||||||
|
}
|
||||||
};
|
};
|
||||||
|
|
||||||
constexpr AmdGpu::Buffer BufferResource::GetVsharp(const Info& info) const noexcept {
|
constexpr AmdGpu::Buffer BufferResource::GetVsharp(const Info& info) const noexcept {
|
||||||
|
@ -51,24 +51,17 @@ void Rasterizer::Draw(bool is_indexed, u32 index_offset) {
|
|||||||
BeginRendering();
|
BeginRendering();
|
||||||
UpdateDynamicState(*pipeline);
|
UpdateDynamicState(*pipeline);
|
||||||
|
|
||||||
u32 instance_offset = 0;
|
const auto [vertex_offset, instance_offset] = vs_info.GetDrawOffsets();
|
||||||
if (vs_info.fetch_offset_sgpr != -1) {
|
|
||||||
instance_offset = vs_info.user_data[vs_info.fetch_offset_sgpr];
|
|
||||||
}
|
|
||||||
|
|
||||||
if (is_indexed) {
|
if (is_indexed) {
|
||||||
u32 vertex_offset = 0;
|
|
||||||
if (vs_info.fetch_index_sgpr != -1) {
|
|
||||||
vertex_offset = vs_info.user_data[vs_info.fetch_index_sgpr];
|
|
||||||
}
|
|
||||||
|
|
||||||
cmdbuf.drawIndexed(num_indices, regs.num_instances.NumInstances(), 0, s32(vertex_offset),
|
cmdbuf.drawIndexed(num_indices, regs.num_instances.NumInstances(), 0, s32(vertex_offset),
|
||||||
instance_offset);
|
instance_offset);
|
||||||
} else {
|
} else {
|
||||||
const u32 num_vertices = regs.primitive_type == AmdGpu::Liverpool::PrimitiveType::RectList
|
const u32 num_vertices = regs.primitive_type == AmdGpu::Liverpool::PrimitiveType::RectList
|
||||||
? 4
|
? 4
|
||||||
: regs.num_indices;
|
: regs.num_indices;
|
||||||
cmdbuf.draw(num_vertices, regs.num_instances.NumInstances(), 0, instance_offset);
|
cmdbuf.draw(num_vertices, regs.num_instances.NumInstances(), vertex_offset,
|
||||||
|
instance_offset);
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
|
Loading…
Reference in New Issue
Block a user