ahmed2007/suyu
1
0
Fork 0
forked from suyu/suyu
Code Pull requests Activity

Pica/VertexShader: Fix a bug caused due to incorrect assumptions of consecutive output register tables.

Browse source 
We now write create a temporary buffer for output registers and copy all of them to the actual output vertex structure after the shader has run. This is technically not necessary, but it's easier to vectorize in the future.
This commit is contained in:
Tony Wasserka 2015-03-12 14:18:46 +01:00
parent ed5b275d21
commit e4f5ec6272
1 changed files with 24 additions and 20 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

44
src/video_core/vertex_shader.cpp
View file

@ -72,7 +72,7 @@ struct VertexShaderState {
u32* program_counter;
const float24* input_register_table[16];
float24* output_register_table[7*4];
Math::Vec4<float24> output_registers[16];
Math::Vec4<float24> temporary_registers[16];
bool conditional_code[2];
@ -198,8 +198,7 @@ static void ProcessShaderCode(VertexShaderState& state) {
src2[3] = src2[3] * float24::FromFloat32(-1);
}
float24* dest = (instr.common.dest.Value() < 0x08) ? state.output_register_table[4*instr.common.dest.Value().GetIndex()]
: (instr.common.dest.Value() < 0x10) ? dummy_vec4_float24
float24* dest = (instr.common.dest.Value() < 0x10) ? &state.output_registers[instr.common.dest.Value().GetIndex()][0]
: (instr.common.dest.Value() < 0x20) ? &state.temporary_registers[instr.common.dest.Value().GetIndex()][0]
: dummy_vec4_float24;
@ -409,8 +408,7 @@ static void ProcessShaderCode(VertexShaderState& state) {
src3[3] = src3[3] * float24::FromFloat32(-1);
}
float24* dest = (instr.mad.dest.Value() < 0x08) ? state.output_register_table[4*instr.mad.dest.Value().GetIndex()]
: (instr.mad.dest.Value() < 0x10) ? dummy_vec4_float24
float24* dest = (instr.mad.dest.Value() < 0x10) ? &state.output_registers[instr.mad.dest.Value().GetIndex()][0]
: (instr.mad.dest.Value() < 0x20) ? &state.temporary_registers[instr.mad.dest.Value().GetIndex()][0]
: dummy_vec4_float24;
@ -587,12 +585,18 @@ OutputVertex RunShader(const InputVertex& input, int num_attributes) {
if(num_attributes > 14) state.input_register_table[attribute_register_map.attribute14_register] = &input.attr[14].x;
if(num_attributes > 15) state.input_register_table[attribute_register_map.attribute15_register] = &input.attr[15].x;
// Setup output register table
OutputVertex ret;
// Zero output so that attributes which aren't output won't have denormals in them, which will
// slow us down later.
memset(&ret, 0, sizeof(ret));
state.conditional_code[0] = false;
state.conditional_code[1] = false;
ProcessShaderCode(state);
DebugUtils::DumpShader(shader_memory.data(), state.debug.max_offset, swizzle_data.data(),
state.debug.max_opdesc_id, registers.vs_main_offset,
registers.vs_output_attributes);
// Setup output data
OutputVertex ret;
// TODO(neobrain): Under some circumstances, up to 16 attributes may be output. We need to
// figure out what those circumstances are and enable the remaining outputs then.
for (int i = 0; i < 7; ++i) {
const auto& output_register_map = registers.vs_output_attributes[i];
@ -601,18 +605,18 @@ OutputVertex RunShader(const InputVertex& input, int num_attributes) {
output_register_map.map_z, output_register_map.map_w
};
for (int comp = 0; comp < 4; ++comp)
state.output_register_table[4*i+comp] = ((float24*)&ret) + semantics[comp];
for (int comp = 0; comp < 4; ++comp) {
float24* out = ((float24*)&ret) + semantics[comp];
if (semantics[comp] != Regs::VSOutputAttributes::INVALID) {
*out = state.output_registers[i][comp];
} else {
// Zero output so that attributes which aren't output won't have denormals in them,
// which would slow us down later.
memset(out, 0, sizeof(*out));
}
}
}
state.conditional_code[0] = false;
state.conditional_code[1] = false;
ProcessShaderCode(state);
DebugUtils::DumpShader(shader_memory.data(), state.debug.max_offset, swizzle_data.data(),
state.debug.max_opdesc_id, registers.vs_main_offset,
registers.vs_output_attributes);
LOG_TRACE(Render_Software, "Output vertex: pos (%.2f, %.2f, %.2f, %.2f), col(%.2f, %.2f, %.2f, %.2f), tc0(%.2f, %.2f)",
ret.pos.x.ToFloat32(), ret.pos.y.ToFloat32(), ret.pos.z.ToFloat32(), ret.pos.w.ToFloat32(),
ret.color.x.ToFloat32(), ret.color.y.ToFloat32(), ret.color.z.ToFloat32(), ret.color.w.ToFloat32(),