diff options
Diffstat (limited to 'src/video_core/vertex_shader.cpp')
| -rw-r--r-- | src/video_core/vertex_shader.cpp | 144 |
1 files changed, 121 insertions, 23 deletions
diff --git a/src/video_core/vertex_shader.cpp b/src/video_core/vertex_shader.cpp index ff825e2e..48977380 100644 --- a/src/video_core/vertex_shader.cpp +++ b/src/video_core/vertex_shader.cpp @@ -348,13 +348,114 @@ static void ProcessShaderCode(VertexShaderState& state) { break; } + + case Instruction::OpCodeType::MultiplyAdd: + { + if (instr.opcode.EffectiveOpCode() == Instruction::OpCode::MAD) { + const SwizzlePattern& swizzle = *(SwizzlePattern*)&swizzle_data[instr.mad.operand_desc_id]; + + const float24* src1_ = LookupSourceRegister(instr.mad.src1); + const float24* src2_ = LookupSourceRegister(instr.mad.src2); + const float24* src3_ = LookupSourceRegister(instr.mad.src3); + + const bool negate_src1 = ((bool)swizzle.negate_src1 != false); + const bool negate_src2 = ((bool)swizzle.negate_src2 != false); + const bool negate_src3 = ((bool)swizzle.negate_src3 != false); + + float24 src1[4] = { + src1_[(int)swizzle.GetSelectorSrc1(0)], + src1_[(int)swizzle.GetSelectorSrc1(1)], + src1_[(int)swizzle.GetSelectorSrc1(2)], + src1_[(int)swizzle.GetSelectorSrc1(3)], + }; + if (negate_src1) { + src1[0] = src1[0] * float24::FromFloat32(-1); + src1[1] = src1[1] * float24::FromFloat32(-1); + src1[2] = src1[2] * float24::FromFloat32(-1); + src1[3] = src1[3] * float24::FromFloat32(-1); + } + float24 src2[4] = { + src2_[(int)swizzle.GetSelectorSrc2(0)], + src2_[(int)swizzle.GetSelectorSrc2(1)], + src2_[(int)swizzle.GetSelectorSrc2(2)], + src2_[(int)swizzle.GetSelectorSrc2(3)], + }; + if (negate_src2) { + src2[0] = src2[0] * float24::FromFloat32(-1); + src2[1] = src2[1] * float24::FromFloat32(-1); + src2[2] = src2[2] * float24::FromFloat32(-1); + src2[3] = src2[3] * float24::FromFloat32(-1); + } + float24 src3[4] = { + src3_[(int)swizzle.GetSelectorSrc3(0)], + src3_[(int)swizzle.GetSelectorSrc3(1)], + src3_[(int)swizzle.GetSelectorSrc3(2)], + src3_[(int)swizzle.GetSelectorSrc3(3)], + }; + if (negate_src3) { + src3[0] = src3[0] * float24::FromFloat32(-1); + src3[1] = src3[1] * float24::FromFloat32(-1); + src3[2] = src3[2] * float24::FromFloat32(-1); + src3[3] = src3[3] * float24::FromFloat32(-1); + } + + float24* dest = (instr.mad.dest < 0x08) ? state.output_register_table[4*instr.mad.dest.GetIndex()] + : (instr.mad.dest < 0x10) ? dummy_vec4_float24 + : (instr.mad.dest < 0x20) ? &state.temporary_registers[instr.mad.dest.GetIndex()][0] + : dummy_vec4_float24; + + for (int i = 0; i < 4; ++i) { + if (!swizzle.DestComponentEnabled(i)) + continue; + + dest[i] = src1[i] * src2[i] + src3[i]; + } + } else { + LOG_ERROR(HW_GPU, "Unhandled multiply-add instruction: 0x%02x (%s): 0x%08x", + (int)instr.opcode.Value(), instr.opcode.GetInfo().name, instr.hex); + } + break; + } + default: + { + static auto evaluate_condition = [](const VertexShaderState& state, bool refx, bool refy, Instruction::FlowControlType flow_control) { + bool results[2] = { refx == state.conditional_code[0], + refy == state.conditional_code[1] }; + + switch (flow_control.op) { + case flow_control.Or: + return results[0] || results[1]; + + case flow_control.And: + return results[0] && results[1]; + + case flow_control.JustX: + return results[0]; + + case flow_control.JustY: + return results[1]; + } + }; + // Handle each instruction on its own switch (instr.opcode) { case Instruction::OpCode::END: exit_loop = true; break; + case Instruction::OpCode::JMPC: + if (evaluate_condition(state, instr.flow_control.refx, instr.flow_control.refy, instr.flow_control)) { + state.program_counter = &shader_memory[instr.flow_control.dest_offset] - 1; + } + break; + + case Instruction::OpCode::JMPU: + if (shader_uniforms.b[instr.flow_control.bool_uniform_id]) { + state.program_counter = &shader_memory[instr.flow_control.dest_offset] - 1; + } + break; + case Instruction::OpCode::CALL: call(state, instr.flow_control.dest_offset, @@ -362,6 +463,24 @@ static void ProcessShaderCode(VertexShaderState& state) { binary_offset + 1); break; + case Instruction::OpCode::CALLU: + if (shader_uniforms.b[instr.flow_control.bool_uniform_id]) { + call(state, + instr.flow_control.dest_offset, + instr.flow_control.num_instructions, + binary_offset + 1); + } + break; + + case Instruction::OpCode::CALLC: + if (evaluate_condition(state, instr.flow_control.refx, instr.flow_control.refy, instr.flow_control)) { + call(state, + instr.flow_control.dest_offset, + instr.flow_control.num_instructions, + binary_offset + 1); + } + break; + case Instruction::OpCode::NOP: break; @@ -384,29 +503,7 @@ static void ProcessShaderCode(VertexShaderState& state) { { // TODO: Do we need to consider swizzlers here? - auto flow_control = instr.flow_control; - bool results[3] = { (bool)flow_control.refx == state.conditional_code[0], - (bool)flow_control.refy == state.conditional_code[1] }; - - switch (flow_control.op) { - case flow_control.Or: - results[2] = results[0] || results[1]; - break; - - case flow_control.And: - results[2] = results[0] && results[1]; - break; - - case flow_control.JustX: - results[2] = results[0]; - break; - - case flow_control.JustY: - results[2] = results[1]; - break; - } - - if (results[2]) { + if (evaluate_condition(state, instr.flow_control.refx, instr.flow_control.refy, instr.flow_control)) { call(state, binary_offset + 1, instr.flow_control.dest_offset - binary_offset - 1, @@ -429,6 +526,7 @@ static void ProcessShaderCode(VertexShaderState& state) { break; } + } ++state.program_counter; |