A GameCube/Wii emulator and debugger written in Rust.
more spec
Layle702f233bb16c
diff --git a/crates/backend-wgpu/Cargo.toml b/crates/backend-wgpu/Cargo.tomlindex c0b5cac..ea9cfcc 100644--- a/crates/backend-wgpu/Cargo.toml+++ b/crates/backend-wgpu/Cargo.toml@@ -18,6 +18,3 @@ tracing = "0.1.44"rustc-hash = "2.1.2"png = "0.18.1"renderdoc = { version = "0.12.1", optional = true }--[build-dependencies]-wesl = "0.3.2"diff --git a/crates/backend-wgpu/build.rs b/crates/backend-wgpu/build.rsdeleted file mode 100644index dc0fdea..0000000--- a/crates/backend-wgpu/build.rs+++ /dev/null@@ -1,3 +0,0 @@-fn main() {- wesl::Wesl::new("src/shaders").build_artifact(&"package::main".parse().unwrap(), "gx_shader");-}diff --git a/crates/backend-wgpu/src/action.rs b/crates/backend-wgpu/src/action.rsindex ee4eb94..e895216 100644--- a/crates/backend-wgpu/src/action.rs+++ b/crates/backend-wgpu/src/action.rs@@ -175,7 +175,10 @@ impl GxRenderer {tracing::info!(?shader_key, "compiled specialized shader variant");}let pipeline_key = self.current_pipeline_key();- let full_key = FullPipelineKey { shader: shader_key, fixed: pipeline_key };+ let full_key = FullPipelineKey {+ shader: shader_key,+ fixed: pipeline_key,+ };if !self.pipeline_cache.contains_key(&full_key) {let module = &self.shader_cache[&shader_key];let pipeline = self.create_pipeline(device, module, &pipeline_key);diff --git a/crates/backend-wgpu/src/shader_specialization.rs b/crates/backend-wgpu/src/shader_specialization.rsindex 51015f6..d425a24 100644--- a/crates/backend-wgpu/src/shader_specialization.rs+++ b/crates/backend-wgpu/src/shader_specialization.rs@@ -1,5 +1,6 @@use gecko::flipper::gx::regs::{AlphaCompare, AlphaOp, CompareFunc};use gecko::host::DrawData;+use std::fmt::Write;use wesl::{VirtualResolver, Wesl};const COMMON_WESL: &str = include_str!("shaders/common.wesl");@@ -10,33 +11,51 @@ const ALPHA_TEST_WESL: &str = include_str!("shaders/alpha_test.wesl");const LIGHTING_WESL: &str = include_str!("shaders/lighting.wesl");const MAIN_WESL: &str = include_str!("shaders/main.wesl");+#[derive(Clone, Copy, Hash, PartialEq, Eq, Debug, Default)]+pub(crate) struct StageKey {+ pub order: u32,+ pub color_env: u32,+ pub alpha_env: u32,+ pub ind_cmd: u32,+}+#[derive(Clone, Copy, Hash, PartialEq, Eq, Debug)]pub(crate) struct ShaderKey {pub num_tev_stages: u8,pub num_indirect_stages: u8,- pub has_lighting: bool,+ pub has_lighting_c0: bool,+ pub has_lighting_c1: bool,pub alpha_test_enabled: bool,+ pub stages: [StageKey; 16],}impl ShaderKey {pub(crate) fn from_draw(draw: &DrawData, alpha_cmp: AlphaCompare) -> Self {let num_tev_stages = draw.num_tev_stages.clamp(1, 16);let num_indirect_stages = draw.num_indirect_stages.min(4);- let has_lighting = draw.color_ctrl[0].enable()- || draw.alpha_ctrl[0].enable()- || draw.color_ctrl[1].enable()- || draw.alpha_ctrl[1].enable();+ let has_lighting_c0 = draw.color_ctrl[0].enable() || draw.alpha_ctrl[0].enable();+ let has_lighting_c1 = draw.color_ctrl[1].enable() || draw.alpha_ctrl[1].enable();let comp0 = alpha_cmp.comp0();let comp1 = alpha_cmp.comp1();let op = alpha_cmp.op();- let always_pass = comp0 == CompareFunc::Always- && comp1 == CompareFunc::Always- && matches!(op, AlphaOp::And | AlphaOp::Or);+ let always_pass =+ comp0 == CompareFunc::Always && comp1 == CompareFunc::Always && matches!(op, AlphaOp::And | AlphaOp::Or);+ let mut stages = [StageKey::default(); 16];+ for i in 0..num_tev_stages as usize {+ stages[i] = StageKey {+ order: draw.tev_orders[i],+ color_env: draw.tev_color_env[i],+ alpha_env: draw.tev_alpha_env[i],+ ind_cmd: draw.tev_indirect[i],+ };+ }Self {num_tev_stages,num_indirect_stages,- has_lighting,+ has_lighting_c0,+ has_lighting_c1,alpha_test_enabled: !always_pass,+ stages,}}}@@ -49,19 +68,34 @@ fn make_resolver() -> VirtualResolver<'static> {r.add_module("package::tev_indirect".parse().unwrap(), TEV_INDIRECT_WESL.into());r.add_module("package::alpha_test".parse().unwrap(), ALPHA_TEST_WESL.into());r.add_module("package::lighting".parse().unwrap(), LIGHTING_WESL.into());- r.add_module("package::main".parse().unwrap(), MAIN_WESL.into());r}+fn generate_main(key: &ShaderKey) -> String {+ let mut consts = String::with_capacity(2048);+ for i in 0..16 {+ let s = &key.stages[i];+ let _ = writeln!(consts, "const STAGE_{i}_ORDER: u32 = {}u;", s.order);+ let _ = writeln!(consts, "const STAGE_{i}_COLOR_ENV: u32 = {}u;", s.color_env);+ let _ = writeln!(consts, "const STAGE_{i}_ALPHA_ENV: u32 = {}u;", s.alpha_env);+ let _ = writeln!(consts, "const STAGE_{i}_IND_CMD: u32 = {}u;", s.ind_cmd);+ }+ MAIN_WESL.replace("// __STAGE_CONSTS__", &consts)+}+pub(crate) fn compile_variant(key: ShaderKey) -> String {- let mut compiler = Wesl::new("").set_custom_resolver(make_resolver());+ let mut resolver = make_resolver();+ let main_src = generate_main(&key);+ resolver.add_module("package::main".parse().unwrap(), main_src.into());+ let mut compiler = Wesl::new("").set_custom_resolver(resolver);for i in 1..=16u8 {compiler.set_feature(&format!("TEV_STAGE_{i}_ENABLED"), i <= key.num_tev_stages);}for i in 0..4u8 {compiler.set_feature(&format!("IND_STAGE_{i}_ENABLED"), i < key.num_indirect_stages);}- compiler.set_feature("HAS_LIGHTING", key.has_lighting);+ compiler.set_feature("HAS_LIGHTING_C0", key.has_lighting_c0);+ compiler.set_feature("HAS_LIGHTING_C1", key.has_lighting_c1);compiler.set_feature("ALPHA_TEST_ENABLED", key.alpha_test_enabled);let entry = "package::main".parse().expect("valid module path");compilerdiff --git a/crates/backend-wgpu/src/shaders/lighting.wesl b/crates/backend-wgpu/src/shaders/lighting.weslindex 55ddc13..aa02419 100644--- a/crates/backend-wgpu/src/shaders/lighting.wesl+++ b/crates/backend-wgpu/src/shaders/lighting.wesl@@ -1,16 +1,6 @@import super::common::{FrameUniforms, frame};import super::tev_helpers::extract_bits;-// ChanCtrl bit layout (lsb0):-// [0] mat_src-// [1] enable-// [2..5] lit_mask_lo (4 bits)-// [6] amb_src-// [7..8] diff_fn (0=None, 1=Signed, 2=Clamp)-// [9] attn_enable-// [10] attn_select (1=Spot, 0=Spec, only when attn_enable)-// [11..14] lit_mask_hi (4 bits)-fn safe_div(n: f32, d: f32) -> f32 {if abs(d) < 1e-10 {return 0.0;@@ -32,18 +22,15 @@ fn compute_light_factor(ctrl: u32, light_idx: u32, normal: vec3<f32>, pos: vec3<var attn: f32;if !attn_enable {- // AttnFn::Noneldir = normalize(ldir);attn = 1.0;} else if attn_select {- // AttnFn::Spotlet dist = length(ldir);ldir = normalize(ldir);let cos_angle = dot(light_dir, ldir);let angle_attn = max(cosatt.x + cosatt.y * cos_angle + cosatt.z * cos_angle * cos_angle, 0.0);attn = safe_div(angle_attn, distatt.x + distatt.y * dist + distatt.z * dist * dist);} else {- // AttnFn::Speclet half_angle = normalize(light_pos);let norm_dot_half = dot(normal, half_angle);if norm_dot_half <= 0.0 {@@ -58,102 +45,125 @@ fn compute_light_factor(ctrl: u32, light_idx: u32, normal: vec3<f32>, pos: vec3<let dif_attn = dot(ldir, normal);var diffuse: f32;switch diff_fn {- case 0u: { diffuse = 1.0; } // None- case 1u: { diffuse = dif_attn; } // Signed- case 2u: { diffuse = max(dif_attn, 0.0); } // Clamp+ case 0u: { diffuse = 1.0; }+ case 1u: { diffuse = dif_attn; }+ case 2u: { diffuse = max(dif_attn, 0.0); }default: { diffuse = 1.0; }}return attn * diffuse;}-fn compute_channel_lighting(- color_ctrl: u32,- alpha_ctrl: u32,- ambient: vec4<f32>,- material: vec4<f32>,+fn material_select(+ color_ctrl: u32, alpha_ctrl: u32,+ ambient: vec4<f32>, material: vec4<f32>,vertex_color: vec4<f32>,- normal: vec3<f32>,- pos: vec3<f32>,) -> vec4<f32> {- // RGB material/ambient source selectionvar mat_rgb: vec3<f32>;- if extract_bits(color_ctrl, 0u, 1u) != 0u {- mat_rgb = vertex_color.rgb;- } else {- mat_rgb = material.rgb;- }+ if extract_bits(color_ctrl, 0u, 1u) != 0u { mat_rgb = vertex_color.rgb; }+ else { mat_rgb = material.rgb; }+ var mat_a: f32;+ if extract_bits(alpha_ctrl, 0u, 1u) != 0u { mat_a = vertex_color.a; }+ else { mat_a = material.a; }+ return vec4(mat_rgb, mat_a);+}++fn compute_lighting_body(+ color_ctrl: u32, alpha_ctrl: u32,+ ambient: vec4<f32>, material: vec4<f32>,+ vertex_color: vec4<f32>, normal: vec3<f32>, pos: vec3<f32>,+) -> vec4<f32> {+ var mat_rgb: vec3<f32>;+ if extract_bits(color_ctrl, 0u, 1u) != 0u { mat_rgb = vertex_color.rgb; }+ else { mat_rgb = material.rgb; }var amb_rgb: vec3<f32>;- if extract_bits(color_ctrl, 6u, 1u) != 0u {- amb_rgb = vertex_color.rgb;- } else {- amb_rgb = ambient.rgb;- }+ if extract_bits(color_ctrl, 6u, 1u) != 0u { amb_rgb = vertex_color.rgb; }+ else { amb_rgb = ambient.rgb; }- // Alpha material/ambient source selectionvar mat_a: f32;- if extract_bits(alpha_ctrl, 0u, 1u) != 0u {- mat_a = vertex_color.a;- } else {- mat_a = material.a;- }+ if extract_bits(alpha_ctrl, 0u, 1u) != 0u { mat_a = vertex_color.a; }+ else { mat_a = material.a; }var amb_a: f32;- if extract_bits(alpha_ctrl, 6u, 1u) != 0u {- amb_a = vertex_color.a;- } else {- amb_a = ambient.a;- }+ if extract_bits(alpha_ctrl, 6u, 1u) != 0u { amb_a = vertex_color.a; }+ else { amb_a = ambient.a; }- @if(!HAS_LIGHTING) {+ let color_lit_off = extract_bits(color_ctrl, 1u, 1u) == 0u;+ let alpha_lit_off = extract_bits(alpha_ctrl, 1u, 1u) == 0u;+ if color_lit_off && alpha_lit_off {return vec4(mat_rgb, mat_a);}- @if(HAS_LIGHTING) {- let color_lit_off = extract_bits(color_ctrl, 1u, 1u) == 0u;- let alpha_lit_off = extract_bits(alpha_ctrl, 1u, 1u) == 0u;-- if color_lit_off && alpha_lit_off {- return vec4(mat_rgb, mat_a);- }+ var rgb_out = mat_rgb;+ if !color_lit_off {+ let mask_lo = extract_bits(color_ctrl, 2u, 4u);+ let mask_hi = extract_bits(color_ctrl, 11u, 4u);+ let light_mask = mask_lo | (mask_hi << 4u);- var rgb_out = mat_rgb;- if !color_lit_off {- let mask_lo = extract_bits(color_ctrl, 2u, 4u);- let mask_hi = extract_bits(color_ctrl, 11u, 4u);- let light_mask = mask_lo | (mask_hi << 4u);-- var acc = amb_rgb;- for (var i = 0u; i < 8u; i++) {- if ((light_mask >> i) & 1u) == 0u {- continue;- }- let lc = frame.light_colors[i];- let factor = compute_light_factor(color_ctrl, i, normal, pos);- acc += lc.rgb * factor;+ var acc = amb_rgb;+ for (var i = 0u; i < 8u; i++) {+ if ((light_mask >> i) & 1u) == 0u {+ continue;}- rgb_out = mat_rgb * clamp(acc, vec3(0.0), vec3(1.0));+ let lc = frame.light_colors[i];+ let factor = compute_light_factor(color_ctrl, i, normal, pos);+ acc += lc.rgb * factor;}+ rgb_out = mat_rgb * clamp(acc, vec3(0.0), vec3(1.0));+ }- var a_out = mat_a;- if !alpha_lit_off {- let mask_lo = extract_bits(alpha_ctrl, 2u, 4u);- let mask_hi = extract_bits(alpha_ctrl, 11u, 4u);- let light_mask = mask_lo | (mask_hi << 4u);-- var acc_a = amb_a;- for (var i = 0u; i < 8u; i++) {- if ((light_mask >> i) & 1u) == 0u {- continue;- }- let lc = frame.light_colors[i];- let factor = compute_light_factor(alpha_ctrl, i, normal, pos);- acc_a += lc.a * factor;+ var a_out = mat_a;+ if !alpha_lit_off {+ let mask_lo = extract_bits(alpha_ctrl, 2u, 4u);+ let mask_hi = extract_bits(alpha_ctrl, 11u, 4u);+ let light_mask = mask_lo | (mask_hi << 4u);++ var acc_a = amb_a;+ for (var i = 0u; i < 8u; i++) {+ if ((light_mask >> i) & 1u) == 0u {+ continue;}- a_out = mat_a * clamp(acc_a, 0.0, 1.0);+ let lc = frame.light_colors[i];+ let factor = compute_light_factor(alpha_ctrl, i, normal, pos);+ acc_a += lc.a * factor;}-- return vec4(rgb_out, a_out);+ a_out = mat_a * clamp(acc_a, 0.0, 1.0);}++ return vec4(rgb_out, a_out);+}++@if(HAS_LIGHTING_C0)+fn compute_lighting_c0(+ color_ctrl: u32, alpha_ctrl: u32,+ ambient: vec4<f32>, material: vec4<f32>,+ vertex_color: vec4<f32>, normal: vec3<f32>, pos: vec3<f32>,+) -> vec4<f32> {+ return compute_lighting_body(color_ctrl, alpha_ctrl, ambient, material, vertex_color, normal, pos);+}+@else+fn compute_lighting_c0(+ color_ctrl: u32, alpha_ctrl: u32,+ ambient: vec4<f32>, material: vec4<f32>,+ vertex_color: vec4<f32>, normal: vec3<f32>, pos: vec3<f32>,+) -> vec4<f32> {+ return material_select(color_ctrl, alpha_ctrl, ambient, material, vertex_color);+}++@if(HAS_LIGHTING_C1)+fn compute_lighting_c1(+ color_ctrl: u32, alpha_ctrl: u32,+ ambient: vec4<f32>, material: vec4<f32>,+ vertex_color: vec4<f32>, normal: vec3<f32>, pos: vec3<f32>,+) -> vec4<f32> {+ return compute_lighting_body(color_ctrl, alpha_ctrl, ambient, material, vertex_color, normal, pos);+}+@else+fn compute_lighting_c1(+ color_ctrl: u32, alpha_ctrl: u32,+ ambient: vec4<f32>, material: vec4<f32>,+ vertex_color: vec4<f32>, normal: vec3<f32>, pos: vec3<f32>,+) -> vec4<f32> {+ return material_select(color_ctrl, alpha_ctrl, ambient, material, vertex_color);}diff --git a/crates/backend-wgpu/src/shaders/main.wesl b/crates/backend-wgpu/src/shaders/main.weslindex 4db79f2..8f7813e 100644--- a/crates/backend-wgpu/src/shaders/main.wesl+++ b/crates/backend-wgpu/src/shaders/main.wesl@@ -1,21 +1,23 @@import super::common::{FrameUniforms, DrawUniforms, VsIn, VsOut, frame, draw, tex0, tex1, tex2, tex3, tex4, tex5, tex6, tex7, samp0, samp1, samp2, samp3, samp4, samp5, samp6, samp7};-import super::tev_helpers::{read_packed, extract_bits, select_texcoord, stq_to_uv, sample_texmap, swizzle_tev_color, tev_color_in, tev_alpha_in};+import super::tev_helpers::{extract_bits, select_texcoord, stq_to_uv, sample_texmap, swizzle_tev_color, tev_color_in, tev_alpha_in};import super::tev_indirect::{IndirectSamples, IndirectResult, sample_indirect_textures, apply_indirect, texmap_size};import super::tev_combiners::{tev_combine_color, tev_combine_alpha};import super::alpha_test::{alpha_compare, alpha_combine};-import super::lighting::compute_channel_lighting;+import super::lighting::{compute_lighting_c0, compute_lighting_c1};++// __STAGE_CONSTS__@vertexfn vs_main(in: VsIn) -> VsOut {var out: VsOut;out.clip_pos = draw.mvp * vec4<f32>(in.pos_view, 1.0);out.clip_pos.z = out.clip_pos.z + out.clip_pos.w;- out.color = compute_channel_lighting(+ out.color = compute_lighting_c0(frame.color_ctrl0, frame.alpha_ctrl0,frame.ambient_color0, frame.material_color0,in.color, in.normal, in.pos_view,);- out.color1 = compute_channel_lighting(+ out.color1 = compute_lighting_c1(frame.color_ctrl1, frame.alpha_ctrl1,frame.ambient_color1, frame.material_color1,in.color1, in.normal, in.pos_view,@@ -40,22 +42,24 @@ struct TevState {fn run_stage(stage: u32,+ order: u32,+ cenv: u32,+ aenv: u32,+ ind_cmd: u32,in: VsOut,color0: vec4<f32>,presampled: IndirectSamples,state: ptr<function, TevState>,) {- let order = read_packed(frame.tev_orders, stage);let texmap_idx = order & 7u;let texcoord_idx = (order >> 3u) & 7u;let tex_enabled = ((order >> 6u) & 1u) != 0u;let ras_channel = (order >> 7u) & 7u;- let aenv = read_packed(frame.tev_alpha_env, stage);let stq = select_texcoord(in, texcoord_idx);let direct_uv = stq_to_uv(stq);let direct_size = texmap_size(texmap_idx);- let indirect = apply_indirect(stage, direct_uv, direct_size, presampled, (*state).running_uv);+ let indirect = apply_indirect(ind_cmd, direct_uv, direct_size, presampled, (*state).running_uv);(*state).running_uv = indirect.uv;var ras_color: vec4<f32>;@@ -80,7 +84,6 @@ fn run_stage(ras_color = swizzle_tev_color(extract_bits(aenv, 0u, 2u), ras_color);tex_color = swizzle_tev_color(extract_bits(aenv, 2u, 2u), tex_color);- let cenv = read_packed(frame.tev_color_env, stage);let c_d = extract_bits(cenv, 0u, 4u);let c_c = extract_bits(cenv, 4u, 4u);let c_b = extract_bits(cenv, 8u, 4u);@@ -137,22 +140,22 @@ fn fs_main(in: VsOut) -> @location(0) vec4<f32> {let color0 = in.color;let presampled = sample_indirect_textures(in);- @if(TEV_STAGE_1_ENABLED) { run_stage( 0u, in, color0, presampled, &state); }- @if(TEV_STAGE_2_ENABLED) { run_stage( 1u, in, color0, presampled, &state); }- @if(TEV_STAGE_3_ENABLED) { run_stage( 2u, in, color0, presampled, &state); }- @if(TEV_STAGE_4_ENABLED) { run_stage( 3u, in, color0, presampled, &state); }- @if(TEV_STAGE_5_ENABLED) { run_stage( 4u, in, color0, presampled, &state); }- @if(TEV_STAGE_6_ENABLED) { run_stage( 5u, in, color0, presampled, &state); }- @if(TEV_STAGE_7_ENABLED) { run_stage( 6u, in, color0, presampled, &state); }- @if(TEV_STAGE_8_ENABLED) { run_stage( 7u, in, color0, presampled, &state); }- @if(TEV_STAGE_9_ENABLED) { run_stage( 8u, in, color0, presampled, &state); }- @if(TEV_STAGE_10_ENABLED) { run_stage( 9u, in, color0, presampled, &state); }- @if(TEV_STAGE_11_ENABLED) { run_stage(10u, in, color0, presampled, &state); }- @if(TEV_STAGE_12_ENABLED) { run_stage(11u, in, color0, presampled, &state); }- @if(TEV_STAGE_13_ENABLED) { run_stage(12u, in, color0, presampled, &state); }- @if(TEV_STAGE_14_ENABLED) { run_stage(13u, in, color0, presampled, &state); }- @if(TEV_STAGE_15_ENABLED) { run_stage(14u, in, color0, presampled, &state); }- @if(TEV_STAGE_16_ENABLED) { run_stage(15u, in, color0, presampled, &state); }+ @if(TEV_STAGE_1_ENABLED) { run_stage( 0u, STAGE_0_ORDER, STAGE_0_COLOR_ENV, STAGE_0_ALPHA_ENV, STAGE_0_IND_CMD, in, color0, presampled, &state); }+ @if(TEV_STAGE_2_ENABLED) { run_stage( 1u, STAGE_1_ORDER, STAGE_1_COLOR_ENV, STAGE_1_ALPHA_ENV, STAGE_1_IND_CMD, in, color0, presampled, &state); }+ @if(TEV_STAGE_3_ENABLED) { run_stage( 2u, STAGE_2_ORDER, STAGE_2_COLOR_ENV, STAGE_2_ALPHA_ENV, STAGE_2_IND_CMD, in, color0, presampled, &state); }+ @if(TEV_STAGE_4_ENABLED) { run_stage( 3u, STAGE_3_ORDER, STAGE_3_COLOR_ENV, STAGE_3_ALPHA_ENV, STAGE_3_IND_CMD, in, color0, presampled, &state); }+ @if(TEV_STAGE_5_ENABLED) { run_stage( 4u, STAGE_4_ORDER, STAGE_4_COLOR_ENV, STAGE_4_ALPHA_ENV, STAGE_4_IND_CMD, in, color0, presampled, &state); }+ @if(TEV_STAGE_6_ENABLED) { run_stage( 5u, STAGE_5_ORDER, STAGE_5_COLOR_ENV, STAGE_5_ALPHA_ENV, STAGE_5_IND_CMD, in, color0, presampled, &state); }+ @if(TEV_STAGE_7_ENABLED) { run_stage( 6u, STAGE_6_ORDER, STAGE_6_COLOR_ENV, STAGE_6_ALPHA_ENV, STAGE_6_IND_CMD, in, color0, presampled, &state); }+ @if(TEV_STAGE_8_ENABLED) { run_stage( 7u, STAGE_7_ORDER, STAGE_7_COLOR_ENV, STAGE_7_ALPHA_ENV, STAGE_7_IND_CMD, in, color0, presampled, &state); }+ @if(TEV_STAGE_9_ENABLED) { run_stage( 8u, STAGE_8_ORDER, STAGE_8_COLOR_ENV, STAGE_8_ALPHA_ENV, STAGE_8_IND_CMD, in, color0, presampled, &state); }+ @if(TEV_STAGE_10_ENABLED) { run_stage( 9u, STAGE_9_ORDER, STAGE_9_COLOR_ENV, STAGE_9_ALPHA_ENV, STAGE_9_IND_CMD, in, color0, presampled, &state); }+ @if(TEV_STAGE_11_ENABLED) { run_stage(10u, STAGE_10_ORDER, STAGE_10_COLOR_ENV, STAGE_10_ALPHA_ENV, STAGE_10_IND_CMD, in, color0, presampled, &state); }+ @if(TEV_STAGE_12_ENABLED) { run_stage(11u, STAGE_11_ORDER, STAGE_11_COLOR_ENV, STAGE_11_ALPHA_ENV, STAGE_11_IND_CMD, in, color0, presampled, &state); }+ @if(TEV_STAGE_13_ENABLED) { run_stage(12u, STAGE_12_ORDER, STAGE_12_COLOR_ENV, STAGE_12_ALPHA_ENV, STAGE_12_IND_CMD, in, color0, presampled, &state); }+ @if(TEV_STAGE_14_ENABLED) { run_stage(13u, STAGE_13_ORDER, STAGE_13_COLOR_ENV, STAGE_13_ALPHA_ENV, STAGE_13_IND_CMD, in, color0, presampled, &state); }+ @if(TEV_STAGE_15_ENABLED) { run_stage(14u, STAGE_14_ORDER, STAGE_14_COLOR_ENV, STAGE_14_ALPHA_ENV, STAGE_14_IND_CMD, in, color0, presampled, &state); }+ @if(TEV_STAGE_16_ENABLED) { run_stage(15u, STAGE_15_ORDER, STAGE_15_COLOR_ENV, STAGE_15_ALPHA_ENV, STAGE_15_IND_CMD, in, color0, presampled, &state); }let color = vec4<f32>(state.regs[state.last_color_dest].rgb, state.regs[state.last_alpha_dest].a);diff --git a/crates/backend-wgpu/src/shaders/tev_indirect.wesl b/crates/backend-wgpu/src/shaders/tev_indirect.weslindex cf1a995..fde827e 100644--- a/crates/backend-wgpu/src/shaders/tev_indirect.wesl+++ b/crates/backend-wgpu/src/shaders/tev_indirect.wesl@@ -110,7 +110,7 @@ struct IndirectResult {}fn apply_indirect(- stage: u32,+ cmd: u32,direct_uv: vec2<f32>,direct_tex_size: vec2<f32>,presampled: IndirectSamples,@@ -120,72 +120,76 @@ fn apply_indirect(result.uv = direct_uv;result.bump_alpha = 0.0;- let cmd = read_packed(frame.tev_indirect, stage);- let source_stage = cmd & 3u;- let format = (cmd >> 2u) & 3u;- let bias_s = (cmd >> 4u) & 1u;- let bias_t = (cmd >> 5u) & 1u;- let bias_u = (cmd >> 6u) & 1u;- let bump_source = (cmd >> 7u) & 3u;- let matrix_index = (cmd >> 9u) & 3u;- let matrix_mode = (cmd >> 11u) & 3u;- let wrap_s = (cmd >> 13u) & 7u;- let wrap_t = (cmd >> 16u) & 7u;- let add_previous = ((cmd >> 20u) & 1u) != 0u;- // TODO: bit 19?-- let sample = presampled.samples[source_stage];-- if bump_source != 0u {- var raw_byte: f32 = 0.0;- switch bump_source {- case 1u: { raw_byte = sample.x; }- case 2u: { raw_byte = sample.y; }- case 3u: { raw_byte = sample.z; }- default: {}- }- let byte_u8 = u32(floor(raw_byte * 255.0));- result.bump_alpha = f32(ind_bump_alpha_byte(byte_u8, format)) / 255.0;- }-- let has_offset = matrix_index != 0u || wrap_s != 0u || wrap_t != 0u || add_previous;- if !has_offset {+ @if(!IND_STAGE_0_ENABLED) {return result;}- let quantized = ind_quantize_and_bias(sample, format, bias_s, bias_t, bias_u);-- var offset_uv = vec2<f32>(0.0);- if matrix_index != 0u {- let idx = matrix_index - 1u;- let row0 = vec3<f32>(frame.indirect_matrices[idx * 2u].xyz) * (1.0 / 1024.0);- let row1 = vec3<f32>(frame.indirect_matrices[idx * 2u + 1u].xyz) * (1.0 / 1024.0);- let scale_exp = frame.indirect_matrices[idx * 2u].w;- let inv_scale = exp2(-f32(scale_exp));- switch matrix_mode {- case 0u: {- offset_uv = vec2<f32>(dot(row0, quantized), dot(row1, quantized))- * inv_scale- / direct_tex_size;- }- case 1u: {- offset_uv = direct_uv * (quantized.x * (1.0 / 256.0) * inv_scale);+ @if(IND_STAGE_0_ENABLED) {+ let source_stage = cmd & 3u;+ let format = (cmd >> 2u) & 3u;+ let bias_s = (cmd >> 4u) & 1u;+ let bias_t = (cmd >> 5u) & 1u;+ let bias_u = (cmd >> 6u) & 1u;+ let bump_source = (cmd >> 7u) & 3u;+ let matrix_index = (cmd >> 9u) & 3u;+ let matrix_mode = (cmd >> 11u) & 3u;+ let wrap_s = (cmd >> 13u) & 7u;+ let wrap_t = (cmd >> 16u) & 7u;+ let add_previous = ((cmd >> 20u) & 1u) != 0u;++ let sample = presampled.samples[source_stage];++ if bump_source != 0u {+ var raw_byte: f32 = 0.0;+ switch bump_source {+ case 1u: { raw_byte = sample.x; }+ case 2u: { raw_byte = sample.y; }+ case 3u: { raw_byte = sample.z; }+ default: {}}- case 2u: {- offset_uv = direct_uv * (quantized.y * (1.0 / 256.0) * inv_scale);+ let byte_u8 = u32(floor(raw_byte * 255.0));+ result.bump_alpha = f32(ind_bump_alpha_byte(byte_u8, format)) / 255.0;+ }++ let has_offset = matrix_index != 0u || wrap_s != 0u || wrap_t != 0u || add_previous;+ if !has_offset {+ return result;+ }++ let quantized = ind_quantize_and_bias(sample, format, bias_s, bias_t, bias_u);++ var offset_uv = vec2<f32>(0.0);+ if matrix_index != 0u {+ let idx = matrix_index - 1u;+ let row0 = vec3<f32>(frame.indirect_matrices[idx * 2u].xyz) * (1.0 / 1024.0);+ let row1 = vec3<f32>(frame.indirect_matrices[idx * 2u + 1u].xyz) * (1.0 / 1024.0);+ let scale_exp = frame.indirect_matrices[idx * 2u].w;+ let inv_scale = exp2(-f32(scale_exp));+ switch matrix_mode {+ case 0u: {+ offset_uv = vec2<f32>(dot(row0, quantized), dot(row1, quantized))+ * inv_scale+ / direct_tex_size;+ }+ case 1u: {+ offset_uv = direct_uv * (quantized.x * (1.0 / 256.0) * inv_scale);+ }+ case 2u: {+ offset_uv = direct_uv * (quantized.y * (1.0 / 256.0) * inv_scale);+ }+ default: {}}- default: {}}- }- let wrapped = vec2<f32>(- ind_wrap_axis(direct_uv.x, wrap_s, direct_tex_size.x),- ind_wrap_axis(direct_uv.y, wrap_t, direct_tex_size.y),- );- var final_uv = wrapped + offset_uv;- if add_previous {- final_uv = final_uv + previous_uv;+ let wrapped = vec2<f32>(+ ind_wrap_axis(direct_uv.x, wrap_s, direct_tex_size.x),+ ind_wrap_axis(direct_uv.y, wrap_t, direct_tex_size.y),+ );+ var final_uv = wrapped + offset_uv;+ if add_previous {+ final_uv = final_uv + previous_uv;+ }+ result.uv = final_uv;+ return result;}- result.uv = final_uv;- return result;}diff --git a/crates/gecko/src/flipper/gx/vertex.rs b/crates/gecko/src/flipper/gx/vertex.rsindex b096936..4c2d91d 100644--- a/crates/gecko/src/flipper/gx/vertex.rs+++ b/crates/gecko/src/flipper/gx/vertex.rs@@ -82,10 +82,11 @@ impl GraphicsProcessor {}}let mut boxed = self.draw_box_pool.pop().unwrap_or_default();- std::mem::swap(&mut boxed.vertices, &mut self.draw_vertices_scratch);self.draw_vertices_scratch.clear();- self.draw_vertices_scratch.reserve(vertex_count);+ if self.draw_vertices_scratch.capacity() < vertex_count {+ self.draw_vertices_scratch.reserve(vertex_count);+ }self::dispatch_decode(self, mmio, cmd, data, vertex_count, &vf);@@ -101,7 +102,8 @@ impl GraphicsProcessor {self.konst_dirty = false;}- let draw_vertices = std::mem::take(&mut self.draw_vertices_scratch);+ boxed.vertices.clear();+ boxed.vertices.append(&mut self.draw_vertices_scratch);// Flatten the three indirect matrices to 6 rows. Row 2M is row// 0 of matrix M, row 2M+1 is row 1. The .w lane carries the@@ -133,7 +135,6 @@ impl GraphicsProcessor {];boxed.primitive = primitive;- boxed.vertices = draw_vertices;boxed.modelview = modelview.0;boxed.tev_color_env = std::array::from_fn(|i| self.cur_tev_color_env[i].raw());boxed.tev_alpha_env = std::array::from_fn(|i| self.cur_tev_alpha_env[i].raw());@@ -725,9 +726,6 @@ fn dispatch_decode<const SYSTEM: SystemId>(};if let Some(parser) = parser {- gp.draw_vertices_scratch.clear();- gp.draw_vertices_scratch.reserve(vertex_count);-let gp_raw = gp as *mut GraphicsProcessor as *mut std::ffi::c_void;let xf_mem_ptr = gp.xf_mem.as_ptr();let arrays_ptr = gp.jit_vtx_arrays.0.as_ptr();diff --git a/crates/tinyapp/src/app.rs b/crates/tinyapp/src/app.rsindex 6da233f..a77ac3d 100644--- a/crates/tinyapp/src/app.rs+++ b/crates/tinyapp/src/app.rs@@ -90,7 +90,11 @@ impl State {let view = frame.texture.create_view(&Default::default());#[cfg(feature = "fps-counter")]- let (fps, native_pct) = if waiting { (0.0, 0.0) } else { fps::read(&self.fps_shared) };+ let (fps, native_pct) = if waiting {+ (0.0, 0.0)+ } else {+ fps::read(&self.fps_shared)+ };#[cfg(not(feature = "fps-counter"))]let fps = if waiting {0.0