3
0
Fork 0
forked from suyu/suyu

Merge pull request #963 from yuriks/gpu-fixes

Misc. GPU vertex loading fixes
This commit is contained in:
bunnei 2015-07-29 16:45:17 -04:00
commit 31c1bb901b
2 changed files with 44 additions and 42 deletions

View file

@ -221,20 +221,14 @@ static inline void WritePicaReg(u32 id, u32 value, u32 mask) {
// Initialize data for the current vertex // Initialize data for the current vertex
VertexShader::InputVertex input; VertexShader::InputVertex input;
// Load a debugging token to check whether this gets loaded by the running
// application or not.
static const float24 debug_token = float24::FromRawFloat24(0x00abcdef);
input.attr[0].w = debug_token;
for (int i = 0; i < attribute_config.GetNumTotalAttributes(); ++i) { for (int i = 0; i < attribute_config.GetNumTotalAttributes(); ++i) {
// Load the default attribute if we're configured to do so, this data will be overwritten by the loader data if it's set if (vertex_attribute_elements[i] != 0) {
if (attribute_config.IsDefaultAttribute(i)) { // Default attribute values set if array elements have < 4 components. This
input.attr[i] = g_state.vs.default_attributes[i]; // is *not* carried over from the default attribute settings even if they're
LOG_TRACE(HW_GPU, "Loaded default attribute %x for vertex %x (index %x): (%f, %f, %f, %f)", // enabled for this attribute.
i, vertex, index, static const float24 zero = float24::FromFloat32(0.0f);
input.attr[i][0].ToFloat32(), input.attr[i][1].ToFloat32(), static const float24 one = float24::FromFloat32(1.0f);
input.attr[i][2].ToFloat32(), input.attr[i][3].ToFloat32()); input.attr[i] = Math::Vec4<float24>(zero, zero, zero, one);
}
// Load per-vertex data from the loader arrays // Load per-vertex data from the loader arrays
for (unsigned int comp = 0; comp < vertex_attribute_elements[i]; ++comp) { for (unsigned int comp = 0; comp < vertex_attribute_elements[i]; ++comp) {
@ -260,17 +254,19 @@ static inline void WritePicaReg(u32 id, u32 value, u32 mask) {
vertex_attribute_strides[i] * vertex + comp * vertex_attribute_element_size[i], vertex_attribute_strides[i] * vertex + comp * vertex_attribute_element_size[i],
input.attr[i][comp].ToFloat32()); input.attr[i][comp].ToFloat32());
} }
} else if (attribute_config.IsDefaultAttribute(i)) {
// Load the default attribute if we're configured to do so
input.attr[i] = g_state.vs.default_attributes[i];
LOG_TRACE(HW_GPU, "Loaded default attribute %x for vertex %x (index %x): (%f, %f, %f, %f)",
i, vertex, index,
input.attr[i][0].ToFloat32(), input.attr[i][1].ToFloat32(),
input.attr[i][2].ToFloat32(), input.attr[i][3].ToFloat32());
} else {
// TODO(yuriks): In this case, no data gets loaded and the vertex remains
// with the last value it had. This isn't currently maintained
// as global state, however, and so won't work in Cita yet.
}
} }
// HACK: Some games do not initialize the vertex position's w component. This leads
// to critical issues since it messes up perspective division. As a
// workaround, we force the fourth component to 1.0 if we find this to be the
// case.
// To do this, we additionally have to assume that the first input attribute
// is the vertex position, since there's no information about this other than
// the empiric observation that this is usually the case.
if (input.attr[0].w == debug_token)
input.attr[0].w = float24::FromFloat32(1.0);
if (g_debug_context) if (g_debug_context)
g_debug_context->OnEvent(DebugContext::Event::VertexLoaded, (void*)&input); g_debug_context->OnEvent(DebugContext::Event::VertexLoaded, (void*)&input);

View file

@ -609,6 +609,12 @@ OutputVertex RunShader(const InputVertex& input, int num_attributes, const Regs:
} }
} }
// The hardware takes the absolute and saturates vertex colors like this, *before* doing interpolation
for (int i = 0; i < 4; ++i) {
ret.color[i] = float24::FromFloat32(
std::fmin(std::fabs(ret.color[i].ToFloat32()), 1.0f));
}
LOG_TRACE(Render_Software, "Output vertex: pos (%.2f, %.2f, %.2f, %.2f), col(%.2f, %.2f, %.2f, %.2f), tc0(%.2f, %.2f)", 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.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(), ret.color.x.ToFloat32(), ret.color.y.ToFloat32(), ret.color.z.ToFloat32(), ret.color.w.ToFloat32(),