Diamond75/Panda3DS
1
0
Fork 0
mirror of https://github.com/wheremyfoodat/Panda3DS.git synced 2025-06-03 12:27:21 +12:00
Code Issues Projects Releases Packages Wiki Activity Actions

[Shader JIT] Get first JIT trangle

Browse source
This commit is contained in:
wheremyfoodat 2023-06-09 02:28:59 +03:00
parent 9bb1f31fc9
commit fd411245fa
3 changed files with 82 additions and 18 deletions
Download patch file Download diff file Expand all files Collapse all files

6
include/PICA/dynapica/shader_rec_emitter_x64.hpp
View file

@ -51,6 +51,9 @@ class ShaderEmitter : public Xbyak::CodeGenerator {
const vec4f& getDestRef(const PICAShader& shader, u32 dest);
// Instruction recompilation functions
void recADD(const PICAShader& shader, u32 instruction);
void recDP4(const PICAShader& shader, u32 instruction);
void recEND(const PICAShader& shader, u32 instruction);
void recMOV(const PICAShader& shader, u32 instruction);
public:
@ -64,6 +67,9 @@ public:
const auto cpu = Xbyak::util::Cpu();
haveSSE4_1 = cpu.has(Xbyak::util::Cpu::tSSE41);
if (!cpu.has(Xbyak::util::Cpu::tSSE3)) {
Helpers::panic("This CPU does not support SSE3. Please use the shader interpreter instead");
}
}
void compile(const PICAShader& shaderUnit);

21
include/PICA/shader.hpp
View file

@ -73,6 +73,18 @@ class PICAShader {
std::array<u32, 4> floatUniformBuffer; // Buffer for temporarily caching float uniform data
public:
// These are placed close to the temp registers and co because it helps the JIT generate better code
u32 entrypoint = 0; // Initial shader PC
u32 boolUniform;
std::array<OpenGL::Vector<u8, 4>, 4> intUniforms;
alignas(16) std::array<vec4f, 96> floatUniforms;
alignas(16) std::array<vec4f, 16> fixedAttributes; // Fixed vertex attributes
alignas(16) std::array<vec4f, 16> inputs; // Attributes passed to the shader
alignas(16) std::array<vec4f, 16> outputs;
alignas(16) vec4f dummy = vec4f({ f24::zero(), f24::zero(), f24::zero(), f24::zero() }); // Dummy register used by the JIT
protected:
std::array<u32, 128> operandDescriptors;
alignas(16) std::array<vec4f, 16> tempRegisters; // General purpose registers the shader can use for temp values
@ -191,15 +203,6 @@ public:
std::array<u32, maxInstructionCount> loadedShader; // Currently loaded & active shader
std::array<u32, maxInstructionCount> bufferedShader; // Shader to be transferred when the SH_CODETRANSFER_END reg gets written to
u32 entrypoint = 0; // Initial shader PC
u32 boolUniform;
std::array<OpenGL::Vector<u8, 4>, 4> intUniforms;
alignas(16) std::array<vec4f, 96> floatUniforms;
alignas(16) std::array<vec4f, 16> fixedAttributes; // Fixed vertex attributes
alignas(16) std::array<vec4f, 16> inputs; // Attributes passed to the shader
alignas(16) std::array<vec4f, 16> outputs;
PICAShader(ShaderType type) : type(type) {}
// Theese functions are in the header to be inlined more easily, though with LTO I hope I'll be able to move them