Add accurate shader multiplication option

src/core/PICA/dynapica/shader_rec.cpp

@@ -16,7 +16,7 @@ void ShaderJIT::prepare(PICAShader& shaderUnit) {
 	auto it = cache.find(hash);
 
 	if (it == cache.end()) { // Block has not been compiled yet
-		auto emitter = std::make_unique<ShaderEmitter>();
+		auto emitter = std::make_unique<ShaderEmitter>(accurateMul);
 		emitter->compile(shaderUnit);
 		// Get pointer to callbacks
 		entrypointCallback = emitter->getInstructionCallback(shaderUnit.entrypoint);

src/core/PICA/dynapica/shader_rec_emitter_arm64.cpp

@@ -7,9 +7,6 @@ using namespace Helpers;
 using namespace oaknut;
 using namespace oaknut::util;
 
-// TODO: Expose safe/unsafe optimizations to the user
-constexpr bool useSafeMUL = true;
-
 // Similar to the x64 recompiler, we use an odd internal ABI, which abuses the fact that we'll very rarely be calling C++ functions
 // So to avoid pushing and popping, we'll be making use of volatile registers as much as possible
 static constexpr QReg src1Vec = Q1;

src/core/PICA/dynapica/shader_rec_emitter_x64.cpp

@@ -12,9 +12,6 @@ using namespace Xbyak;
 using namespace Xbyak::util;
 using namespace Helpers;
 
-// TODO: Expose safe/unsafe optimizations to the user
-constexpr bool useSafeMUL = false;
-
 // The shader recompiler uses quite an odd internal ABI
 // We make use of the fact that in regular conditions, we should pretty much never be calling C++ code from recompiled shader code
 // This allows us to establish an ABI that's optimized for this sort of workflow, statically allocating volatile host registers

src/core/PICA/gpu.cpp

@@ -64,6 +64,8 @@ void GPU::reset() {
 	regs.fill(0);
 	shaderUnit.reset();
 	shaderJIT.reset();
+	shaderJIT.setAccurateMul(config.accurateShaderMul);
+
 	std::memset(vram, 0, vramSize);
 	lightingLUT.fill(0);
 	lightingLUTDirty = true;