portaldevice/suyu
1
0
Fork 0
forked from suyu/suyu
Code Pull requests Activity

Pica/VertexShader: Implement the LOOP instruction.

Browse source
This commit is contained in:
Tony Wasserka 2014-12-21 03:01:35 +01:00
parent 6c26ec72a5
commit 70a764d992
1 changed files with 36 additions and 14 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

50
src/video_core/vertex_shader.cpp
View file

@ -85,8 +85,11 @@ struct VertexShaderState {
}; };
struct CallStackElement { struct CallStackElement {
u32 final_address; u32 final_address; // Address upon which we jump to return_address
u32 return_address; u32 return_address; // Where to jump when leaving scope
u8 repeat_counter; // How often to repeat until this call stack element is removed
u8 loop_increment; // Which value to add to the loop counter after an iteration
// TODO: Should this be a signed value? Does it even matter?
}; };
// TODO: Is there a maximal size for this? // TODO: Is there a maximal size for this?
@ -105,9 +108,14 @@ static void ProcessShaderCode(VertexShaderState& state) {
while (true) { while (true) {
if (!state.call_stack.empty()) { if (!state.call_stack.empty()) {
if (state.program_counter - shader_memory.data() == state.call_stack.top().final_address) { auto& top = state.call_stack.top();
state.program_counter = &shader_memory[state.call_stack.top().return_address]; if (state.program_counter - shader_memory.data() == top.final_address) {
state.call_stack.pop(); state.address_registers[2] += top.loop_increment;
if (top.repeat_counter-- == 0) {
state.program_counter = &shader_memory[top.return_address];
state.call_stack.pop();
}
// TODO: Is "trying again" accurate to hardware? // TODO: Is "trying again" accurate to hardware?
continue; continue;
@ -118,9 +126,10 @@ static void ProcessShaderCode(VertexShaderState& state) {
const Instruction& instr = *(const Instruction*)state.program_counter; const Instruction& instr = *(const Instruction*)state.program_counter;
const SwizzlePattern& swizzle = *(SwizzlePattern*)&swizzle_data[instr.common.operand_desc_id]; const SwizzlePattern& swizzle = *(SwizzlePattern*)&swizzle_data[instr.common.operand_desc_id];
auto call = [&](VertexShaderState& state, u32 offset, u32 num_instructions, u32 return_offset) { static auto call = [](VertexShaderState& state, u32 offset, u32 num_instructions,
u32 return_offset, u8 repeat_count, u8 loop_increment) {
state.program_counter = &shader_memory[offset] - 1; // -1 to make sure when incrementing the PC we end up at the correct offset state.program_counter = &shader_memory[offset] - 1; // -1 to make sure when incrementing the PC we end up at the correct offset
state.call_stack.push({ offset + num_instructions, return_offset }); state.call_stack.push({ offset + num_instructions, return_offset, repeat_count, loop_increment });
}; };
u32 binary_offset = state.program_counter - shader_memory.data(); u32 binary_offset = state.program_counter - shader_memory.data();
@ -457,7 +466,7 @@ static void ProcessShaderCode(VertexShaderState& state) {
call(state, call(state,
instr.flow_control.dest_offset, instr.flow_control.dest_offset,
instr.flow_control.num_instructions, instr.flow_control.num_instructions,
binary_offset + 1); binary_offset + 1, 0, 0);
break; break;
case Instruction::OpCode::CALLU: case Instruction::OpCode::CALLU:
@ -465,7 +474,7 @@ static void ProcessShaderCode(VertexShaderState& state) {
call(state, call(state,
instr.flow_control.dest_offset, instr.flow_control.dest_offset,
instr.flow_control.num_instructions, instr.flow_control.num_instructions,
binary_offset + 1); binary_offset + 1, 0, 0);
} }
break; break;
@ -474,7 +483,7 @@ static void ProcessShaderCode(VertexShaderState& state) {
call(state, call(state,
instr.flow_control.dest_offset, instr.flow_control.dest_offset,
instr.flow_control.num_instructions, instr.flow_control.num_instructions,
binary_offset + 1); binary_offset + 1, 0, 0);
} }
break; break;
@ -486,12 +495,12 @@ static void ProcessShaderCode(VertexShaderState& state) {
call(state, call(state,
binary_offset + 1, binary_offset + 1,
instr.flow_control.dest_offset - binary_offset - 1, instr.flow_control.dest_offset - binary_offset - 1,
instr.flow_control.dest_offset + instr.flow_control.num_instructions); instr.flow_control.dest_offset + instr.flow_control.num_instructions, 0, 0);
} else { } else {
call(state, call(state,
instr.flow_control.dest_offset, instr.flow_control.dest_offset,
instr.flow_control.num_instructions, instr.flow_control.num_instructions,
instr.flow_control.dest_offset + instr.flow_control.num_instructions); instr.flow_control.dest_offset + instr.flow_control.num_instructions, 0, 0);
} }
break; break;
@ -504,17 +513,30 @@ static void ProcessShaderCode(VertexShaderState& state) {
call(state, call(state,
binary_offset + 1, binary_offset + 1,
instr.flow_control.dest_offset - binary_offset - 1, instr.flow_control.dest_offset - binary_offset - 1,
instr.flow_control.dest_offset + instr.flow_control.num_instructions); instr.flow_control.dest_offset + instr.flow_control.num_instructions, 0, 0);
} else { } else {
call(state, call(state,
instr.flow_control.dest_offset, instr.flow_control.dest_offset,
instr.flow_control.num_instructions, instr.flow_control.num_instructions,
instr.flow_control.dest_offset + instr.flow_control.num_instructions); instr.flow_control.dest_offset + instr.flow_control.num_instructions, 0, 0);
} }
break; break;
} }
case Instruction::OpCode::LOOP:
{
state.address_registers[2] = shader_uniforms.i[instr.flow_control.int_uniform_id].y;
call(state,
binary_offset + 1,
instr.flow_control.dest_offset - binary_offset + 1,
instr.flow_control.dest_offset + 1,
shader_uniforms.i[instr.flow_control.int_uniform_id].x,
shader_uniforms.i[instr.flow_control.int_uniform_id].z);
break;
}
default: default:
LOG_ERROR(HW_GPU, "Unhandled instruction: 0x%02x (%s): 0x%08x", LOG_ERROR(HW_GPU, "Unhandled instruction: 0x%02x (%s): 0x%08x",
(int)instr.opcode.Value(), instr.opcode.GetInfo().name, instr.hex); (int)instr.opcode.Value(), instr.opcode.GetInfo().name, instr.hex);