fabi/Xenia-Canary
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[GPU] Eliminate unused shader I/O, UCP culling, centroid on Vulkan

Browse Source 
For more optimal usage of exports and the parameter cache on the host regardless of how effective the optimizations in the host GPU driver are. Also reserve space for Vulkan/Metal/D3D11-specific HostVertexShaderTypes to use one more bit for the host vertex shader type in the shader modification bits, so that won't have to be done in the future as that would require invalidating shader storages (which are invalidated by this commit) again.
This commit is contained in:
Triang3l
2022-07-21 12:32:28 +03:00
parent 0a94b86cb8
commit 1a95bef8b3
22 changed files with 3366 additions and 3235 deletions
Download Patch File Download Diff File Expand all files Collapse all files

62
src/xenia/gpu/d3d12/d3d12_command_processor.cc
View File

@@ -2159,12 +2159,21 @@ bool D3D12CommandProcessor::IssueDraw(xenos::PrimitiveType primitive_type,
draw_util::GetNormalizedDepthControl(regs);
// Shader modifications.
uint32_t ps_param_gen_pos = UINT32_MAX;
uint32_t interpolator_mask =
pixel_shader ? (vertex_shader->writes_interpolators() &
pixel_shader->GetInterpolatorInputMask(
regs.Get<reg::SQ_PROGRAM_CNTL>(),
regs.Get<reg::SQ_CONTEXT_MISC>(), ps_param_gen_pos))
: 0;
DxbcShaderTranslator::Modification vertex_shader_modification =
pipeline_cache_->GetCurrentVertexShaderModification(
*vertex_shader, primitive_processing_result.host_vertex_shader_type);
*vertex_shader, primitive_processing_result.host_vertex_shader_type,
interpolator_mask);
DxbcShaderTranslator::Modification pixel_shader_modification =
pixel_shader ? pipeline_cache_->GetCurrentPixelShaderModification(
*pixel_shader, normalized_depth_control)
*pixel_shader, interpolator_mask, ps_param_gen_pos,
normalized_depth_control)
: DxbcShaderTranslator::Modification(0);
// Set up the render targets - this may perform dispatches and draws.
@@ -3227,25 +3236,14 @@ void D3D12CommandProcessor::UpdateSystemConstantValues(
if (pa_cl_vte_cntl.vtx_w0_fmt) {
flags |= DxbcShaderTranslator::kSysFlag_WNotReciprocal;
}
// User clip planes (UCP_ENA_#), when not CLIP_DISABLE.
if (!pa_cl_clip_cntl.clip_disable) {
flags |= (pa_cl_clip_cntl.value & 0b111111)
<< DxbcShaderTranslator::kSysFlag_UserClipPlane0_Shift;
}
// Whether the primitive is polygonal and SV_IsFrontFace matters.
if (primitive_polygonal) {
flags |= DxbcShaderTranslator::kSysFlag_PrimitivePolygonal;
}
// Primitive type.
if (vgt_draw_initiator.prim_type == xenos::PrimitiveType::kPointList) {
flags |= DxbcShaderTranslator::kSysFlag_PrimitivePoint;
} else if (draw_util::IsPrimitiveLine(regs)) {
if (draw_util::IsPrimitiveLine(regs)) {
flags |= DxbcShaderTranslator::kSysFlag_PrimitiveLine;
}
// Primitive killing condition.
if (pa_cl_clip_cntl.vtx_kill_or) {
flags |= DxbcShaderTranslator::kSysFlag_KillIfAnyVertexKilled;
}
// Depth format.
if (rb_depth_info.depth_format == xenos::DepthRenderTargetFormat::kD24FS8) {
flags |= DxbcShaderTranslator::kSysFlag_DepthFloat24;
@@ -3325,18 +3323,24 @@ void D3D12CommandProcessor::UpdateSystemConstantValues(
system_constants_.vertex_index_max = vgt_max_vtx_indx;
// User clip planes (UCP_ENA_#), when not CLIP_DISABLE.
// The shader knows only the total count - tightly packing the user clip
// planes that are actually used.
if (!pa_cl_clip_cntl.clip_disable) {
for (uint32_t i = 0; i < 6; ++i) {
if (!(pa_cl_clip_cntl.value & (1 << i))) {
continue;
}
const float* ucp = &regs[XE_GPU_REG_PA_CL_UCP_0_X + i * 4].f32;
if (std::memcmp(system_constants_.user_clip_planes[i], ucp,
float* user_clip_plane_write_ptr = system_constants_.user_clip_planes[0];
uint32_t user_clip_planes_remaining = pa_cl_clip_cntl.ucp_ena;
uint32_t user_clip_plane_index;
while (xe::bit_scan_forward(user_clip_planes_remaining,
&user_clip_plane_index)) {
user_clip_planes_remaining &= ~(UINT32_C(1) << user_clip_plane_index);
const float* user_clip_plane =
&regs[XE_GPU_REG_PA_CL_UCP_0_X + user_clip_plane_index * 4].f32;
if (std::memcmp(user_clip_plane_write_ptr, user_clip_plane,
4 * sizeof(float))) {
dirty = true;
std::memcpy(system_constants_.user_clip_planes[i], ucp,
std::memcpy(user_clip_plane_write_ptr, user_clip_plane,
4 * sizeof(float));
}
user_clip_plane_write_ptr += 4;
}
}
@@ -3387,22 +3391,6 @@ void D3D12CommandProcessor::UpdateSystemConstantValues(
system_constants_.point_screen_diameter_to_ndc_radius[1] =
point_screen_diameter_to_ndc_radius_y;
// Interpolator sampling pattern, centroid or center.
uint32_t interpolator_sampling_pattern =
xenos::GetInterpolatorSamplingPattern(
rb_surface_info.msaa_samples, sq_context_misc.sc_sample_cntl,
regs.Get<reg::SQ_INTERPOLATOR_CNTL>().sampling_pattern);
dirty |= system_constants_.interpolator_sampling_pattern !=
interpolator_sampling_pattern;
system_constants_.interpolator_sampling_pattern =
interpolator_sampling_pattern;
// Pixel parameter register.
uint32_t ps_param_gen =
sq_program_cntl.param_gen ? sq_context_misc.param_gen_pos : UINT_MAX;
dirty |= system_constants_.ps_param_gen != ps_param_gen;
system_constants_.ps_param_gen = ps_param_gen;
// Texture signedness / gamma.
bool gamma_render_target_as_srgb =
render_target_cache_->gamma_render_target_as_srgb();