summaryrefslogtreecommitdiff
path: root/cnn_v3/src
diff options
context:
space:
mode:
Diffstat (limited to 'cnn_v3/src')
-rw-r--r--cnn_v3/src/cnn_v3_effect.cc185
-rw-r--r--cnn_v3/src/cnn_v3_effect.h61
-rw-r--r--cnn_v3/src/gbuf_deferred_effect.cc91
-rw-r--r--cnn_v3/src/gbuf_deferred_effect.h10
-rw-r--r--cnn_v3/src/gbuf_view_effect.cc90
-rw-r--r--cnn_v3/src/gbuf_view_effect.h17
-rw-r--r--cnn_v3/src/gbuffer_effect.cc222
-rw-r--r--cnn_v3/src/gbuffer_effect.h42
8 files changed, 377 insertions, 341 deletions
diff --git a/cnn_v3/src/cnn_v3_effect.cc b/cnn_v3/src/cnn_v3_effect.cc
index e576ceb..fa1716f 100644
--- a/cnn_v3/src/cnn_v3_effect.cc
+++ b/cnn_v3/src/cnn_v3_effect.cc
@@ -22,18 +22,23 @@
// Format: Conv(IN→OUT, KxK) has OUT*IN*K*K weights + OUT biases
// Layout: OIHW order (out × in × kH × kW), biases appended
// ---------------------------------------------------------------------------
-static const uint32_t kEnc0Weights = 20 * 8 * 9 + 8; // Conv(20→8, 3×3)+bias = 1448
-static const uint32_t kEnc1Weights = 8 * 16 * 9 + 16; // Conv(8→16, 3×3)+bias = 1168
-static const uint32_t kBnWeights = 16 * 16 * 9 + 16; // Conv(16→16, 3×3,dil=2)+bias = 2320
-static const uint32_t kDec1Weights = 32 * 8 * 9 + 8; // Conv(32→8, 3×3)+bias = 2312
-static const uint32_t kDec0Weights = 16 * 4 * 9 + 4; // Conv(16→4, 3×3)+bias = 580
+static const uint32_t kEnc0Weights =
+ 20 * 8 * 9 + 8; // Conv(20→8, 3×3)+bias = 1448
+static const uint32_t kEnc1Weights =
+ 8 * 16 * 9 + 16; // Conv(8→16, 3×3)+bias = 1168
+static const uint32_t kBnWeights =
+ 16 * 16 * 9 + 16; // Conv(16→16, 3×3,dil=2)+bias = 2320
+static const uint32_t kDec1Weights =
+ 32 * 8 * 9 + 8; // Conv(32→8, 3×3)+bias = 2312
+static const uint32_t kDec0Weights =
+ 16 * 4 * 9 + 4; // Conv(16→4, 3×3)+bias = 580
-static const uint32_t kEnc0Offset = 0;
-static const uint32_t kEnc1Offset = kEnc0Offset + kEnc0Weights;
-static const uint32_t kBnOffset = kEnc1Offset + kEnc1Weights;
-static const uint32_t kDec1Offset = kBnOffset + kBnWeights;
-static const uint32_t kDec0Offset = kDec1Offset + kDec1Weights;
-static const uint32_t kTotalF16 = kDec0Offset + kDec0Weights;
+static const uint32_t kEnc0Offset = 0;
+static const uint32_t kEnc1Offset = kEnc0Offset + kEnc0Weights;
+static const uint32_t kBnOffset = kEnc1Offset + kEnc1Weights;
+static const uint32_t kDec1Offset = kBnOffset + kBnWeights;
+static const uint32_t kDec0Offset = kDec1Offset + kDec1Weights;
+static const uint32_t kTotalF16 = kDec0Offset + kDec0Weights;
// = 1448 + 1168 + 2320 + 2312 + 580 = 7828 f16
static const uint32_t kWeightsBufBytes = ((kTotalF16 + 1) / 2) * 4;
@@ -57,7 +62,7 @@ static WGPUShaderModule make_shader(WGPUDevice device, const char* wgsl) {
WGPUShaderSourceWGSL src = {};
src.chain.sType = WGPUSType_ShaderSourceWGSL;
- src.code = str_view(composed.c_str());
+ src.code = str_view(composed.c_str());
WGPUShaderModuleDescriptor desc = {};
desc.nextInChain = &src.chain;
@@ -69,7 +74,7 @@ static WGPUBindGroupLayout make_bgl(WGPUDevice device,
uint32_t count) {
WGPUBindGroupLayoutDescriptor desc = {};
desc.entryCount = count;
- desc.entries = entries;
+ desc.entries = entries;
return wgpuDeviceCreateBindGroupLayout(device, &desc);
}
@@ -79,14 +84,15 @@ static WGPUComputePipeline make_compute_pipeline(WGPUDevice device,
WGPUBindGroupLayout bgl) {
WGPUPipelineLayoutDescriptor pl_desc = {};
pl_desc.bindGroupLayoutCount = 1;
- pl_desc.bindGroupLayouts = &bgl;
+ pl_desc.bindGroupLayouts = &bgl;
WGPUPipelineLayout pl = wgpuDeviceCreatePipelineLayout(device, &pl_desc);
WGPUComputePipelineDescriptor pipe_desc = {};
- pipe_desc.layout = pl;
- pipe_desc.compute.module = shader;
- pipe_desc.compute.entryPoint = str_view(entry);
- WGPUComputePipeline pipe = wgpuDeviceCreateComputePipeline(device, &pipe_desc);
+ pipe_desc.layout = pl;
+ pipe_desc.compute.module = shader;
+ pipe_desc.compute.entryPoint = str_view(entry);
+ WGPUComputePipeline pipe =
+ wgpuDeviceCreateComputePipeline(device, &pipe_desc);
wgpuPipelineLayoutRelease(pl);
return pipe;
@@ -95,36 +101,36 @@ static WGPUComputePipeline make_compute_pipeline(WGPUDevice device,
// BGL entry helpers
static WGPUBindGroupLayoutEntry bgl_uint_tex(uint32_t binding) {
WGPUBindGroupLayoutEntry e = {};
- e.binding = binding;
- e.visibility = WGPUShaderStage_Compute;
- e.texture.sampleType = WGPUTextureSampleType_Uint;
- e.texture.viewDimension = WGPUTextureViewDimension_2D;
+ e.binding = binding;
+ e.visibility = WGPUShaderStage_Compute;
+ e.texture.sampleType = WGPUTextureSampleType_Uint;
+ e.texture.viewDimension = WGPUTextureViewDimension_2D;
return e;
}
static WGPUBindGroupLayoutEntry bgl_storage_buf(uint32_t binding) {
WGPUBindGroupLayoutEntry e = {};
- e.binding = binding;
- e.visibility = WGPUShaderStage_Compute;
- e.buffer.type = WGPUBufferBindingType_ReadOnlyStorage;
+ e.binding = binding;
+ e.visibility = WGPUShaderStage_Compute;
+ e.buffer.type = WGPUBufferBindingType_ReadOnlyStorage;
return e;
}
static WGPUBindGroupLayoutEntry bgl_uniform_buf(uint32_t binding,
uint64_t min_size) {
WGPUBindGroupLayoutEntry e = {};
- e.binding = binding;
- e.visibility = WGPUShaderStage_Compute;
- e.buffer.type = WGPUBufferBindingType_Uniform;
- e.buffer.minBindingSize = min_size;
+ e.binding = binding;
+ e.visibility = WGPUShaderStage_Compute;
+ e.buffer.type = WGPUBufferBindingType_Uniform;
+ e.buffer.minBindingSize = min_size;
return e;
}
-static WGPUBindGroupLayoutEntry bgl_storage_tex_write(
- uint32_t binding, WGPUTextureFormat fmt) {
+static WGPUBindGroupLayoutEntry bgl_storage_tex_write(uint32_t binding,
+ WGPUTextureFormat fmt) {
WGPUBindGroupLayoutEntry e = {};
- e.binding = binding;
- e.visibility = WGPUShaderStage_Compute;
- e.storageTexture.access = WGPUStorageTextureAccess_WriteOnly;
- e.storageTexture.format = fmt;
- e.storageTexture.viewDimension = WGPUTextureViewDimension_2D;
+ e.binding = binding;
+ e.visibility = WGPUShaderStage_Compute;
+ e.storageTexture.access = WGPUStorageTextureAccess_WriteOnly;
+ e.storageTexture.format = fmt;
+ e.storageTexture.viewDimension = WGPUTextureViewDimension_2D;
return e;
}
@@ -141,16 +147,16 @@ CNNv3Effect::CNNv3Effect(const GpuContext& ctx,
const std::string& prefix =
outputs.empty() ? std::string("cnn_v3") : outputs[0];
- node_enc0_ = prefix + "_enc0";
+ node_enc0_ = prefix + "_enc0";
node_enc1_lo_ = prefix + "_enc1_lo";
node_enc1_hi_ = prefix + "_enc1_hi";
- node_bn_lo_ = prefix + "_bn_lo";
- node_bn_hi_ = prefix + "_bn_hi";
- node_dec1_ = prefix + "_dec1";
+ node_bn_lo_ = prefix + "_bn_lo";
+ node_bn_hi_ = prefix + "_bn_hi";
+ node_dec1_ = prefix + "_dec1";
- weights_buf_ = gpu_create_buffer(
- ctx_.device, kWeightsBufBytes,
- WGPUBufferUsage_Storage | WGPUBufferUsage_CopyDst);
+ weights_buf_ =
+ gpu_create_buffer(ctx_.device, kWeightsBufBytes,
+ WGPUBufferUsage_Storage | WGPUBufferUsage_CopyDst);
enc0_params_buf_.init(ctx_.device);
enc1_params_buf_.init(ctx_.device);
@@ -166,7 +172,9 @@ CNNv3Effect::CNNv3Effect(const GpuContext& ctx,
}
enc1_params_.weight_offset = kEnc1Offset;
- for (int i = 0; i < 16; ++i) { enc1_params_.gamma[i] = 1.0f; }
+ for (int i = 0; i < 16; ++i) {
+ enc1_params_.gamma[i] = 1.0f;
+ }
bn_params_.weight_offset = kBnOffset;
@@ -177,7 +185,9 @@ CNNv3Effect::CNNv3Effect(const GpuContext& ctx,
}
dec0_params_.weight_offset = kDec0Offset;
- for (int i = 0; i < 4; ++i) { dec0_params_.gamma[i] = 1.0f; }
+ for (int i = 0; i < 4; ++i) {
+ dec0_params_.gamma[i] = 1.0f;
+ }
create_pipelines();
@@ -205,15 +215,15 @@ void CNNv3Effect::declare_nodes(NodeRegistry& registry) {
const int H = registry.default_height();
// enc0: rgba32uint full-res (8ch packed f16)
- registry.declare_node(node_enc0_, NodeType::GBUF_RGBA32UINT, W, H);
+ registry.declare_node(node_enc0_, NodeType::GBUF_RGBA32UINT, W, H);
// enc1: two rgba32uint half-res (8ch each = 16ch total)
registry.declare_node(node_enc1_lo_, NodeType::GBUF_RGBA32UINT, W / 2, H / 2);
registry.declare_node(node_enc1_hi_, NodeType::GBUF_RGBA32UINT, W / 2, H / 2);
// bottleneck: two rgba32uint quarter-res (8ch each = 16ch total)
- registry.declare_node(node_bn_lo_, NodeType::GBUF_RGBA32UINT, W / 4, H / 4);
- registry.declare_node(node_bn_hi_, NodeType::GBUF_RGBA32UINT, W / 4, H / 4);
+ registry.declare_node(node_bn_lo_, NodeType::GBUF_RGBA32UINT, W / 4, H / 4);
+ registry.declare_node(node_bn_hi_, NodeType::GBUF_RGBA32UINT, W / 4, H / 4);
// dec1: rgba32uint half-res (8ch packed f16)
- registry.declare_node(node_dec1_, NodeType::GBUF_RGBA32UINT, W / 2, H / 2);
+ registry.declare_node(node_dec1_, NodeType::GBUF_RGBA32UINT, W / 2, H / 2);
// output_nodes_[0]: rgba16float full-res — declared externally by caller
}
@@ -222,12 +232,13 @@ void CNNv3Effect::declare_nodes(NodeRegistry& registry) {
// ---------------------------------------------------------------------------
void CNNv3Effect::upload_weights(WGPUQueue queue, const void* data,
- uint32_t size_bytes) {
+ uint32_t size_bytes) {
wgpuQueueWriteBuffer(queue, weights_buf_.buffer, 0, data, size_bytes);
}
void CNNv3Effect::load_film_mlp(const void* data, uint32_t size_bytes) {
- if (size_bytes != sizeof(CNNv3FilmMlp)) return;
+ if (size_bytes != sizeof(CNNv3FilmMlp))
+ return;
memcpy(&mlp_, data, sizeof(CNNv3FilmMlp));
mlp_loaded_ = true;
}
@@ -246,16 +257,19 @@ void CNNv3Effect::set_film_params(const CNNv3FiLMParams& fp) {
float h[16];
for (int j = 0; j < 16; ++j) {
float s = mlp_.l0_b[j];
- for (int i = 0; i < 5; ++i) s += mlp_.l0_w[j * 5 + i] * cond[i];
+ for (int i = 0; i < 5; ++i)
+ s += mlp_.l0_w[j * 5 + i] * cond[i];
h[j] = s > 0.f ? s : 0.f;
}
// Layer 1: Linear(16→72)
- // Output split: g_enc0(8)|b_enc0(8)|g_enc1(16)|b_enc1(16)|g_dec1(8)|b_dec1(8)|g_dec0(4)|b_dec0(4)
+ // Output split:
+ // g_enc0(8)|b_enc0(8)|g_enc1(16)|b_enc1(16)|g_dec1(8)|b_dec1(8)|g_dec0(4)|b_dec0(4)
float film[72];
for (int j = 0; j < 72; ++j) {
float s = mlp_.l1_b[j];
- for (int i = 0; i < 16; ++i) s += mlp_.l1_w[j * 16 + i] * h[i];
+ for (int i = 0; i < 16; ++i)
+ s += mlp_.l1_w[j * 16 + i] * h[i];
film[j] = s;
}
@@ -270,9 +284,11 @@ void CNNv3Effect::set_film_params(const CNNv3FiLMParams& fp) {
enc0_params_.beta_hi[i] = p[i + 4];
}
p += 8;
- for (int i = 0; i < 16; ++i) enc1_params_.gamma[i] = p[i];
+ for (int i = 0; i < 16; ++i)
+ enc1_params_.gamma[i] = p[i];
p += 16;
- for (int i = 0; i < 16; ++i) enc1_params_.beta[i] = p[i];
+ for (int i = 0; i < 16; ++i)
+ enc1_params_.beta[i] = p[i];
p += 16;
for (int i = 0; i < 4; ++i) {
dec1_params_.gamma_lo[i] = p[i];
@@ -284,9 +300,11 @@ void CNNv3Effect::set_film_params(const CNNv3FiLMParams& fp) {
dec1_params_.beta_hi[i] = p[i + 4];
}
p += 8;
- for (int i = 0; i < 4; ++i) dec0_params_.gamma[i] = p[i];
+ for (int i = 0; i < 4; ++i)
+ dec0_params_.gamma[i] = p[i];
p += 4;
- for (int i = 0; i < 4; ++i) dec0_params_.beta[i] = p[i];
+ for (int i = 0; i < 4; ++i)
+ dec0_params_.beta[i] = p[i];
}
// ---------------------------------------------------------------------------
@@ -307,27 +325,24 @@ void CNNv3Effect::render(WGPUCommandEncoder encoder,
const int W = (int)params.resolution.x;
const int H = (int)params.resolution.y;
- auto dispatch = [&](WGPUComputePipeline pipe, WGPUBindGroup bg,
- int w, int h) {
+ auto dispatch = [&](WGPUComputePipeline pipe, WGPUBindGroup bg, int w,
+ int h) {
WGPUComputePassDescriptor pass_desc = {};
WGPUComputePassEncoder pass =
wgpuCommandEncoderBeginComputePass(encoder, &pass_desc);
wgpuComputePassEncoderSetPipeline(pass, pipe);
wgpuComputePassEncoderSetBindGroup(pass, 0, bg, 0, nullptr);
- wgpuComputePassEncoderDispatchWorkgroups(
- pass,
- (uint32_t)((w + 7) / 8),
- (uint32_t)((h + 7) / 8),
- 1);
+ wgpuComputePassEncoderDispatchWorkgroups(pass, (uint32_t)((w + 7) / 8),
+ (uint32_t)((h + 7) / 8), 1);
wgpuComputePassEncoderEnd(pass);
wgpuComputePassEncoderRelease(pass);
};
- dispatch(enc0_pipeline_.get(), enc0_bg_.get(), W, H);
- dispatch(enc1_pipeline_.get(), enc1_bg_.get(), W / 2, H / 2);
- dispatch(bn_pipeline_.get(), bn_bg_.get(), W / 4, H / 4);
- dispatch(dec1_pipeline_.get(), dec1_bg_.get(), W / 2, H / 2);
- dispatch(dec0_pipeline_.get(), dec0_bg_.get(), W, H);
+ dispatch(enc0_pipeline_.get(), enc0_bg_.get(), W, H);
+ dispatch(enc1_pipeline_.get(), enc1_bg_.get(), W / 2, H / 2);
+ dispatch(bn_pipeline_.get(), bn_bg_.get(), W / 4, H / 4);
+ dispatch(dec1_pipeline_.get(), dec1_bg_.get(), W / 2, H / 2);
+ dispatch(dec0_pipeline_.get(), dec0_bg_.get(), W, H);
}
// ---------------------------------------------------------------------------
@@ -443,40 +458,39 @@ void CNNv3Effect::create_pipelines() {
static void bg_tex(WGPUBindGroupEntry& e, uint32_t binding,
WGPUTextureView view) {
e = {};
- e.binding = binding;
+ e.binding = binding;
e.textureView = view;
}
static void bg_buf(WGPUBindGroupEntry& e, uint32_t binding, WGPUBuffer buf,
uint64_t size) {
e = {};
e.binding = binding;
- e.buffer = buf;
- e.size = size;
+ e.buffer = buf;
+ e.size = size;
}
void CNNv3Effect::update_bind_groups(NodeRegistry& nodes) {
WGPUDevice dev = ctx_.device;
- WGPUTextureView feat0_view = nodes.get_view(input_nodes_[0]);
- WGPUTextureView feat1_view = nodes.get_view(input_nodes_[1]);
- WGPUTextureView enc0_view = nodes.get_view(node_enc0_);
+ WGPUTextureView feat0_view = nodes.get_view(input_nodes_[0]);
+ WGPUTextureView feat1_view = nodes.get_view(input_nodes_[1]);
+ WGPUTextureView enc0_view = nodes.get_view(node_enc0_);
WGPUTextureView enc1_lo_view = nodes.get_view(node_enc1_lo_);
WGPUTextureView enc1_hi_view = nodes.get_view(node_enc1_hi_);
- WGPUTextureView bn_lo_view = nodes.get_view(node_bn_lo_);
- WGPUTextureView bn_hi_view = nodes.get_view(node_bn_hi_);
- WGPUTextureView dec1_view = nodes.get_view(node_dec1_);
- WGPUTextureView out_view = nodes.get_view(output_nodes_[0]);
+ WGPUTextureView bn_lo_view = nodes.get_view(node_bn_lo_);
+ WGPUTextureView bn_hi_view = nodes.get_view(node_bn_hi_);
+ WGPUTextureView dec1_view = nodes.get_view(node_dec1_);
+ WGPUTextureView out_view = nodes.get_view(output_nodes_[0]);
WGPUBuffer wb = weights_buf_.buffer;
auto make_bg = [&](WGPUComputePipeline pipe, WGPUBindGroupEntry* e,
uint32_t count) -> WGPUBindGroup {
- WGPUBindGroupLayout bgl =
- wgpuComputePipelineGetBindGroupLayout(pipe, 0);
+ WGPUBindGroupLayout bgl = wgpuComputePipelineGetBindGroupLayout(pipe, 0);
WGPUBindGroupDescriptor desc = {};
- desc.layout = bgl;
+ desc.layout = bgl;
desc.entryCount = count;
- desc.entries = e;
+ desc.entries = e;
WGPUBindGroup bg = wgpuDeviceCreateBindGroup(dev, &desc);
wgpuBindGroupLayoutRelease(bgl);
return bg;
@@ -504,7 +518,8 @@ void CNNv3Effect::update_bind_groups(NodeRegistry& nodes) {
enc1_bg_.replace(make_bg(enc1_pipeline_.get(), e, 5));
}
- // bottleneck: enc1_lo(B0), enc1_hi(B1), weights(B2), params(B3), bn_lo(B4), bn_hi(B5)
+ // bottleneck: enc1_lo(B0), enc1_hi(B1), weights(B2), params(B3), bn_lo(B4),
+ // bn_hi(B5)
{
WGPUBindGroupEntry e[6] = {};
bg_tex(e[0], 0, enc1_lo_view);
diff --git a/cnn_v3/src/cnn_v3_effect.h b/cnn_v3/src/cnn_v3_effect.h
index 589680c..ac0166f 100644
--- a/cnn_v3/src/cnn_v3_effect.h
+++ b/cnn_v3/src/cnn_v3_effect.h
@@ -38,12 +38,12 @@
// offset 80: beta_hi (vec4f)
// total: 96 bytes
struct CnnV3Params8ch {
- uint32_t weight_offset; // offset 0
- uint32_t _pad[7]; // offsets 4-31
- float gamma_lo[4]; // offset 32
- float gamma_hi[4]; // offset 48
- float beta_lo[4]; // offset 64
- float beta_hi[4]; // offset 80
+ uint32_t weight_offset; // offset 0
+ uint32_t _pad[7]; // offsets 4-31
+ float gamma_lo[4]; // offset 32
+ float gamma_hi[4]; // offset 48
+ float beta_lo[4]; // offset 64
+ float beta_hi[4]; // offset 80
};
static_assert(sizeof(CnnV3Params8ch) == 96, "CnnV3Params8ch must be 96 bytes");
@@ -56,12 +56,13 @@ static_assert(sizeof(CnnV3Params8ch) == 96, "CnnV3Params8ch must be 96 bytes");
// offset 96: beta_0..3 (4x vec4f = 64 bytes)
// total: 160 bytes
struct CnnV3Params16ch {
- uint32_t weight_offset; // offset 0
- uint32_t _pad[7]; // offsets 4-31
- float gamma[16]; // offsets 32-95
- float beta[16]; // offsets 96-159
+ uint32_t weight_offset; // offset 0
+ uint32_t _pad[7]; // offsets 4-31
+ float gamma[16]; // offsets 32-95
+ float beta[16]; // offsets 96-159
};
-static_assert(sizeof(CnnV3Params16ch) == 160, "CnnV3Params16ch must be 160 bytes");
+static_assert(sizeof(CnnV3Params16ch) == 160,
+ "CnnV3Params16ch must be 160 bytes");
// dec0: 4-channel FiLM
//
@@ -72,10 +73,10 @@ static_assert(sizeof(CnnV3Params16ch) == 160, "CnnV3Params16ch must be 160 bytes
// offset 48: beta (vec4f)
// total: 64 bytes
struct CnnV3Params4ch {
- uint32_t weight_offset; // offset 0
- uint32_t _pad[7]; // offsets 4-31
- float gamma[4]; // offset 32
- float beta[4]; // offset 48
+ uint32_t weight_offset; // offset 0
+ uint32_t _pad[7]; // offsets 4-31
+ float gamma[4]; // offset 32
+ float beta[4]; // offset 48
};
static_assert(sizeof(CnnV3Params4ch) == 64, "CnnV3Params4ch must be 64 bytes");
@@ -90,20 +91,20 @@ static_assert(sizeof(CnnV3ParamsBn) == 16, "CnnV3ParamsBn must be 16 bytes");
// FiLM conditioning inputs (CPU-side, uploaded via set_film_params each frame)
// ---------------------------------------------------------------------------
struct CNNv3FiLMParams {
- float beat_phase = 0.0f; // 0-1 within current beat
- float beat_norm = 0.0f; // beat_time / 8.0, normalized 8-beat cycle
- float audio_intensity = 0.0f; // peak audio level 0-1
- float style_p0 = 0.0f; // user-defined style param
- float style_p1 = 0.0f; // user-defined style param
+ float beat_phase = 0.0f; // 0-1 within current beat
+ float beat_norm = 0.0f; // beat_time / 8.0, normalized 8-beat cycle
+ float audio_intensity = 0.0f; // peak audio level 0-1
+ float style_p0 = 0.0f; // user-defined style param
+ float style_p1 = 0.0f; // user-defined style param
};
// FiLM MLP weights: Linear(5→16)→ReLU→Linear(16→72).
// Loaded from cnn_v3_film_mlp.bin (1320 f32 = 5280 bytes).
// Layout: l0_w(80) | l0_b(16) | l1_w(1152) | l1_b(72), all row-major f32.
struct CNNv3FilmMlp {
- float l0_w[16 * 5]; // (16, 5) row-major
+ float l0_w[16 * 5]; // (16, 5) row-major
float l0_b[16];
- float l1_w[72 * 16]; // (72, 16) row-major
+ float l1_w[72 * 16]; // (72, 16) row-major
float l1_b[72];
};
static_assert(sizeof(CNNv3FilmMlp) == 1320 * 4, "CNNv3FilmMlp size mismatch");
@@ -153,21 +154,21 @@ class CNNv3Effect : public Effect {
BindGroup dec0_bg_;
// Params uniform buffers (one per pass)
- UniformBuffer<CnnV3Params8ch> enc0_params_buf_;
+ UniformBuffer<CnnV3Params8ch> enc0_params_buf_;
UniformBuffer<CnnV3Params16ch> enc1_params_buf_;
- UniformBuffer<CnnV3ParamsBn> bn_params_buf_;
- UniformBuffer<CnnV3Params8ch> dec1_params_buf_;
- UniformBuffer<CnnV3Params4ch> dec0_params_buf_;
+ UniformBuffer<CnnV3ParamsBn> bn_params_buf_;
+ UniformBuffer<CnnV3Params8ch> dec1_params_buf_;
+ UniformBuffer<CnnV3Params4ch> dec0_params_buf_;
// Shared packed-f16 weights (storage buffer, read-only in all shaders)
GpuBuffer weights_buf_;
// Per-pass params shadow (updated by set_film_params, uploaded in render)
- CnnV3Params8ch enc0_params_{};
+ CnnV3Params8ch enc0_params_{};
CnnV3Params16ch enc1_params_{};
- CnnV3ParamsBn bn_params_{};
- CnnV3Params8ch dec1_params_{};
- CnnV3Params4ch dec0_params_{};
+ CnnV3ParamsBn bn_params_{};
+ CnnV3Params8ch dec1_params_{};
+ CnnV3Params4ch dec0_params_{};
void create_pipelines();
void update_bind_groups(NodeRegistry& nodes);
diff --git a/cnn_v3/src/gbuf_deferred_effect.cc b/cnn_v3/src/gbuf_deferred_effect.cc
index de6bd29..561f660 100644
--- a/cnn_v3/src/gbuf_deferred_effect.cc
+++ b/cnn_v3/src/gbuf_deferred_effect.cc
@@ -1,4 +1,5 @@
-// GBufDeferredEffect — simple deferred render: albedo * shadow from packed G-buffer.
+// GBufDeferredEffect — simple deferred render: albedo * shadow from packed
+// G-buffer.
#include "gbuf_deferred_effect.h"
#include "gpu/gpu.h"
@@ -10,22 +11,24 @@ extern const char* gbuf_deferred_wgsl;
struct GBufDeferredUniforms {
float resolution[2];
};
-static_assert(sizeof(GBufDeferredUniforms) == 8, "GBufDeferredUniforms must be 8 bytes");
+static_assert(sizeof(GBufDeferredUniforms) == 8,
+ "GBufDeferredUniforms must be 8 bytes");
static WGPUBindGroupLayoutEntry bgl_uint_tex(uint32_t binding) {
WGPUBindGroupLayoutEntry e = {};
- e.binding = binding;
- e.visibility = WGPUShaderStage_Fragment;
- e.texture.sampleType = WGPUTextureSampleType_Uint;
+ e.binding = binding;
+ e.visibility = WGPUShaderStage_Fragment;
+ e.texture.sampleType = WGPUTextureSampleType_Uint;
e.texture.viewDimension = WGPUTextureViewDimension_2D;
return e;
}
-static WGPUBindGroupLayoutEntry bgl_uniform(uint32_t binding, uint64_t min_size) {
+static WGPUBindGroupLayoutEntry bgl_uniform(uint32_t binding,
+ uint64_t min_size) {
WGPUBindGroupLayoutEntry e = {};
- e.binding = binding;
- e.visibility = WGPUShaderStage_Fragment;
- e.buffer.type = WGPUBufferBindingType_Uniform;
+ e.binding = binding;
+ e.visibility = WGPUShaderStage_Fragment;
+ e.buffer.type = WGPUBufferBindingType_Uniform;
e.buffer.minBindingSize = min_size;
return e;
}
@@ -44,40 +47,43 @@ GBufDeferredEffect::GBufDeferredEffect(const GpuContext& ctx,
};
WGPUBindGroupLayoutDescriptor bgl_desc = {};
bgl_desc.entryCount = 3;
- bgl_desc.entries = entries;
- WGPUBindGroupLayout bgl = wgpuDeviceCreateBindGroupLayout(ctx_.device, &bgl_desc);
+ bgl_desc.entries = entries;
+ WGPUBindGroupLayout bgl =
+ wgpuDeviceCreateBindGroupLayout(ctx_.device, &bgl_desc);
WGPUPipelineLayoutDescriptor pl_desc = {};
pl_desc.bindGroupLayoutCount = 1;
- pl_desc.bindGroupLayouts = &bgl;
+ pl_desc.bindGroupLayouts = &bgl;
WGPUPipelineLayout pl = wgpuDeviceCreatePipelineLayout(ctx_.device, &pl_desc);
WGPUShaderSourceWGSL wgsl_src = {};
wgsl_src.chain.sType = WGPUSType_ShaderSourceWGSL;
- const std::string composed = ShaderComposer::Get().Compose({}, gbuf_deferred_wgsl);
- wgsl_src.code = str_view(composed.c_str());
+ const std::string composed =
+ ShaderComposer::Get().Compose({}, gbuf_deferred_wgsl);
+ wgsl_src.code = str_view(composed.c_str());
WGPUShaderModuleDescriptor shader_desc = {};
shader_desc.nextInChain = &wgsl_src.chain;
- WGPUShaderModule shader = wgpuDeviceCreateShaderModule(ctx_.device, &shader_desc);
+ WGPUShaderModule shader =
+ wgpuDeviceCreateShaderModule(ctx_.device, &shader_desc);
WGPUColorTargetState target = {};
- target.format = WGPUTextureFormat_RGBA8Unorm;
+ target.format = WGPUTextureFormat_RGBA8Unorm;
target.writeMask = WGPUColorWriteMask_All;
WGPUFragmentState frag = {};
- frag.module = shader;
- frag.entryPoint = str_view("fs_main");
+ frag.module = shader;
+ frag.entryPoint = str_view("fs_main");
frag.targetCount = 1;
- frag.targets = &target;
+ frag.targets = &target;
WGPURenderPipelineDescriptor pipe_desc = {};
- pipe_desc.layout = pl;
- pipe_desc.vertex.module = shader;
- pipe_desc.vertex.entryPoint = str_view("vs_main");
- pipe_desc.fragment = &frag;
- pipe_desc.primitive.topology = WGPUPrimitiveTopology_TriangleList;
- pipe_desc.multisample.count = 1;
- pipe_desc.multisample.mask = UINT32_MAX;
+ pipe_desc.layout = pl;
+ pipe_desc.vertex.module = shader;
+ pipe_desc.vertex.entryPoint = str_view("vs_main");
+ pipe_desc.fragment = &frag;
+ pipe_desc.primitive.topology = WGPUPrimitiveTopology_TriangleList;
+ pipe_desc.multisample.count = 1;
+ pipe_desc.multisample.mask = UINT32_MAX;
pipeline_.set(wgpuDeviceCreateRenderPipeline(ctx_.device, &pipe_desc));
@@ -89,46 +95,47 @@ GBufDeferredEffect::GBufDeferredEffect(const GpuContext& ctx,
void GBufDeferredEffect::render(WGPUCommandEncoder encoder,
const UniformsSequenceParams& params,
NodeRegistry& nodes) {
- WGPUTextureView feat0_view = nodes.get_view(input_nodes_[0]);
- WGPUTextureView feat1_view = nodes.get_view(input_nodes_[1]);
+ WGPUTextureView feat0_view = nodes.get_view(input_nodes_[0]);
+ WGPUTextureView feat1_view = nodes.get_view(input_nodes_[1]);
WGPUTextureView output_view = nodes.get_view(output_nodes_[0]);
- // Upload resolution uniform into the base class uniforms buffer (first 8 bytes).
+ // Upload resolution uniform into the base class uniforms buffer (first 8
+ // bytes).
GBufDeferredUniforms u;
u.resolution[0] = params.resolution.x;
u.resolution[1] = params.resolution.y;
- wgpuQueueWriteBuffer(ctx_.queue, uniforms_buffer_.get().buffer, 0,
- &u, sizeof(u));
+ wgpuQueueWriteBuffer(ctx_.queue, uniforms_buffer_.get().buffer, 0, &u,
+ sizeof(u));
WGPUBindGroupLayout bgl =
wgpuRenderPipelineGetBindGroupLayout(pipeline_.get(), 0);
WGPUBindGroupEntry bg_entries[3] = {};
- bg_entries[0].binding = 0;
+ bg_entries[0].binding = 0;
bg_entries[0].textureView = feat0_view;
- bg_entries[1].binding = 1;
+ bg_entries[1].binding = 1;
bg_entries[1].textureView = feat1_view;
- bg_entries[2].binding = 2;
- bg_entries[2].buffer = uniforms_buffer_.get().buffer;
- bg_entries[2].size = sizeof(GBufDeferredUniforms);
+ bg_entries[2].binding = 2;
+ bg_entries[2].buffer = uniforms_buffer_.get().buffer;
+ bg_entries[2].size = sizeof(GBufDeferredUniforms);
WGPUBindGroupDescriptor bg_desc = {};
- bg_desc.layout = bgl;
+ bg_desc.layout = bgl;
bg_desc.entryCount = 3;
- bg_desc.entries = bg_entries;
+ bg_desc.entries = bg_entries;
bind_group_.replace(wgpuDeviceCreateBindGroup(ctx_.device, &bg_desc));
wgpuBindGroupLayoutRelease(bgl);
WGPURenderPassColorAttachment color_att = {};
- color_att.view = output_view;
- color_att.loadOp = WGPULoadOp_Clear;
- color_att.storeOp = WGPUStoreOp_Store;
+ color_att.view = output_view;
+ color_att.loadOp = WGPULoadOp_Clear;
+ color_att.storeOp = WGPUStoreOp_Store;
color_att.clearValue = {0.0f, 0.0f, 0.0f, 1.0f};
color_att.depthSlice = WGPU_DEPTH_SLICE_UNDEFINED;
WGPURenderPassDescriptor pass_desc = {};
pass_desc.colorAttachmentCount = 1;
- pass_desc.colorAttachments = &color_att;
+ pass_desc.colorAttachments = &color_att;
WGPURenderPassEncoder pass =
wgpuCommandEncoderBeginRenderPass(encoder, &pass_desc);
diff --git a/cnn_v3/src/gbuf_deferred_effect.h b/cnn_v3/src/gbuf_deferred_effect.h
index 4daf13d..d0368ff 100644
--- a/cnn_v3/src/gbuf_deferred_effect.h
+++ b/cnn_v3/src/gbuf_deferred_effect.h
@@ -1,5 +1,6 @@
// GBufDeferredEffect — simple deferred render from packed G-buffer.
-// Inputs: feat_tex0, feat_tex1 (rgba32uint). Output: albedo * shadow (rgba8unorm).
+// Inputs: feat_tex0, feat_tex1 (rgba32uint). Output: albedo * shadow
+// (rgba8unorm).
#pragma once
#include "gpu/effect.h"
@@ -10,11 +11,10 @@ class GBufDeferredEffect : public Effect {
public:
GBufDeferredEffect(const GpuContext& ctx,
const std::vector<std::string>& inputs,
- const std::vector<std::string>& outputs,
- float start_time, float end_time);
+ const std::vector<std::string>& outputs, float start_time,
+ float end_time);
- void render(WGPUCommandEncoder encoder,
- const UniformsSequenceParams& params,
+ void render(WGPUCommandEncoder encoder, const UniformsSequenceParams& params,
NodeRegistry& nodes) override;
private:
diff --git a/cnn_v3/src/gbuf_view_effect.cc b/cnn_v3/src/gbuf_view_effect.cc
index ccf80b0..4c23322 100644
--- a/cnn_v3/src/gbuf_view_effect.cc
+++ b/cnn_v3/src/gbuf_view_effect.cc
@@ -19,28 +19,28 @@ extern const char* gbuf_view_wgsl;
// BGL entry: texture_2d<u32> read binding (fragment stage)
static WGPUBindGroupLayoutEntry bgl_uint_tex_frag(uint32_t binding) {
WGPUBindGroupLayoutEntry e = {};
- e.binding = binding;
- e.visibility = WGPUShaderStage_Fragment;
- e.texture.sampleType = WGPUTextureSampleType_Uint;
+ e.binding = binding;
+ e.visibility = WGPUShaderStage_Fragment;
+ e.texture.sampleType = WGPUTextureSampleType_Uint;
e.texture.viewDimension = WGPUTextureViewDimension_2D;
return e;
}
// BGL entry: uniform buffer (fragment stage)
static WGPUBindGroupLayoutEntry bgl_uniform_frag(uint32_t binding,
- uint64_t min_size) {
+ uint64_t min_size) {
WGPUBindGroupLayoutEntry e = {};
- e.binding = binding;
- e.visibility = WGPUShaderStage_Fragment;
- e.buffer.type = WGPUBufferBindingType_Uniform;
+ e.binding = binding;
+ e.visibility = WGPUShaderStage_Fragment;
+ e.buffer.type = WGPUBufferBindingType_Uniform;
e.buffer.minBindingSize = min_size;
return e;
}
GBufViewEffect::GBufViewEffect(const GpuContext& ctx,
- const std::vector<std::string>& inputs,
- const std::vector<std::string>& outputs,
- float start_time, float end_time)
+ const std::vector<std::string>& inputs,
+ const std::vector<std::string>& outputs,
+ float start_time, float end_time)
: Effect(ctx, inputs, outputs, start_time, end_time) {
HEADLESS_RETURN_IF_NULL(ctx_.device);
@@ -48,24 +48,26 @@ GBufViewEffect::GBufViewEffect(const GpuContext& ctx,
WGPUBindGroupLayoutEntry entries[3] = {
bgl_uint_tex_frag(0),
bgl_uint_tex_frag(1),
- bgl_uniform_frag(2, 8), // only resolution (vec2f = 8 bytes) is read
+ bgl_uniform_frag(2, 8), // only resolution (vec2f = 8 bytes) is read
};
WGPUBindGroupLayoutDescriptor bgl_desc = {};
bgl_desc.entryCount = 3;
- bgl_desc.entries = entries;
- WGPUBindGroupLayout bgl = wgpuDeviceCreateBindGroupLayout(ctx_.device, &bgl_desc);
+ bgl_desc.entries = entries;
+ WGPUBindGroupLayout bgl =
+ wgpuDeviceCreateBindGroupLayout(ctx_.device, &bgl_desc);
// Pipeline layout
WGPUPipelineLayoutDescriptor pl_desc = {};
pl_desc.bindGroupLayoutCount = 1;
- pl_desc.bindGroupLayouts = &bgl;
+ pl_desc.bindGroupLayouts = &bgl;
WGPUPipelineLayout pl = wgpuDeviceCreatePipelineLayout(ctx_.device, &pl_desc);
// Shader module
WGPUShaderSourceWGSL wgsl_src = {};
wgsl_src.chain.sType = WGPUSType_ShaderSourceWGSL;
- const std::string composed = ShaderComposer::Get().Compose({}, gbuf_view_wgsl);
- wgsl_src.code = str_view(composed.c_str());
+ const std::string composed =
+ ShaderComposer::Get().Compose({}, gbuf_view_wgsl);
+ wgsl_src.code = str_view(composed.c_str());
WGPUShaderModuleDescriptor shader_desc = {};
shader_desc.nextInChain = &wgsl_src.chain;
WGPUShaderModule shader =
@@ -73,23 +75,23 @@ GBufViewEffect::GBufViewEffect(const GpuContext& ctx,
// Render pipeline
WGPUColorTargetState target = {};
- target.format = WGPUTextureFormat_RGBA8Unorm;
+ target.format = WGPUTextureFormat_RGBA8Unorm;
target.writeMask = WGPUColorWriteMask_All;
WGPUFragmentState frag = {};
- frag.module = shader;
- frag.entryPoint = str_view("fs_main");
+ frag.module = shader;
+ frag.entryPoint = str_view("fs_main");
frag.targetCount = 1;
- frag.targets = &target;
+ frag.targets = &target;
WGPURenderPipelineDescriptor pipe_desc = {};
- pipe_desc.layout = pl;
- pipe_desc.vertex.module = shader;
- pipe_desc.vertex.entryPoint = str_view("vs_main");
- pipe_desc.fragment = &frag;
- pipe_desc.primitive.topology = WGPUPrimitiveTopology_TriangleList;
- pipe_desc.multisample.count = 1;
- pipe_desc.multisample.mask = UINT32_MAX;
+ pipe_desc.layout = pl;
+ pipe_desc.vertex.module = shader;
+ pipe_desc.vertex.entryPoint = str_view("vs_main");
+ pipe_desc.fragment = &frag;
+ pipe_desc.primitive.topology = WGPUPrimitiveTopology_TriangleList;
+ pipe_desc.multisample.count = 1;
+ pipe_desc.multisample.mask = UINT32_MAX;
pipeline_.set(wgpuDeviceCreateRenderPipeline(ctx_.device, &pipe_desc));
@@ -99,10 +101,10 @@ GBufViewEffect::GBufViewEffect(const GpuContext& ctx,
}
void GBufViewEffect::render(WGPUCommandEncoder encoder,
- const UniformsSequenceParams& params,
- NodeRegistry& nodes) {
- WGPUTextureView feat0_view = nodes.get_view(input_nodes_[0]);
- WGPUTextureView feat1_view = nodes.get_view(input_nodes_[1]);
+ const UniformsSequenceParams& params,
+ NodeRegistry& nodes) {
+ WGPUTextureView feat0_view = nodes.get_view(input_nodes_[0]);
+ WGPUTextureView feat1_view = nodes.get_view(input_nodes_[1]);
WGPUTextureView output_view = nodes.get_view(output_nodes_[0]);
// Rebuild bind group (views may change with ping-pong or resize)
@@ -110,31 +112,31 @@ void GBufViewEffect::render(WGPUCommandEncoder encoder,
wgpuRenderPipelineGetBindGroupLayout(pipeline_.get(), 0);
WGPUBindGroupEntry bg_entries[3] = {};
- bg_entries[0].binding = 0;
+ bg_entries[0].binding = 0;
bg_entries[0].textureView = feat0_view;
- bg_entries[1].binding = 1;
+ bg_entries[1].binding = 1;
bg_entries[1].textureView = feat1_view;
- bg_entries[2].binding = 2;
- bg_entries[2].buffer = uniforms_buffer_.get().buffer;
- bg_entries[2].size = sizeof(UniformsSequenceParams);
+ bg_entries[2].binding = 2;
+ bg_entries[2].buffer = uniforms_buffer_.get().buffer;
+ bg_entries[2].size = sizeof(UniformsSequenceParams);
WGPUBindGroupDescriptor bg_desc = {};
- bg_desc.layout = bgl;
+ bg_desc.layout = bgl;
bg_desc.entryCount = 3;
- bg_desc.entries = bg_entries;
+ bg_desc.entries = bg_entries;
bind_group_.replace(wgpuDeviceCreateBindGroup(ctx_.device, &bg_desc));
wgpuBindGroupLayoutRelease(bgl);
WGPURenderPassColorAttachment color_att = {};
- color_att.view = output_view;
- color_att.loadOp = WGPULoadOp_Clear;
- color_att.storeOp = WGPUStoreOp_Store;
- color_att.clearValue = {0.0f, 0.0f, 0.0f, 1.0f};
- color_att.depthSlice = WGPU_DEPTH_SLICE_UNDEFINED;
+ color_att.view = output_view;
+ color_att.loadOp = WGPULoadOp_Clear;
+ color_att.storeOp = WGPUStoreOp_Store;
+ color_att.clearValue = {0.0f, 0.0f, 0.0f, 1.0f};
+ color_att.depthSlice = WGPU_DEPTH_SLICE_UNDEFINED;
WGPURenderPassDescriptor pass_desc = {};
pass_desc.colorAttachmentCount = 1;
- pass_desc.colorAttachments = &color_att;
+ pass_desc.colorAttachments = &color_att;
WGPURenderPassEncoder pass =
wgpuCommandEncoderBeginRenderPass(encoder, &pass_desc);
diff --git a/cnn_v3/src/gbuf_view_effect.h b/cnn_v3/src/gbuf_view_effect.h
index d4d8139..a8fd5c9 100644
--- a/cnn_v3/src/gbuf_view_effect.h
+++ b/cnn_v3/src/gbuf_view_effect.h
@@ -1,6 +1,7 @@
// GBufViewEffect: Visualizes G-buffer feature textures as a 4×5 channel grid.
-// Inputs: feat_tex0 (rgba32uint, ch 0-7 f16), feat_tex1 (rgba32uint, ch 8-19 unorm8)
-// Output: rgba8unorm tiled channel visualization (downscaled 4× per channel)
+// Inputs: feat_tex0 (rgba32uint, ch 0-7 f16), feat_tex1 (rgba32uint, ch 8-19
+// unorm8) Output: rgba8unorm tiled channel visualization (downscaled 4× per
+// channel)
#pragma once
@@ -10,16 +11,14 @@
class GBufViewEffect : public Effect {
public:
- GBufViewEffect(const GpuContext& ctx,
- const std::vector<std::string>& inputs,
- const std::vector<std::string>& outputs,
- float start_time, float end_time);
+ GBufViewEffect(const GpuContext& ctx, const std::vector<std::string>& inputs,
+ const std::vector<std::string>& outputs, float start_time,
+ float end_time);
- void render(WGPUCommandEncoder encoder,
- const UniformsSequenceParams& params,
+ void render(WGPUCommandEncoder encoder, const UniformsSequenceParams& params,
NodeRegistry& nodes) override;
private:
RenderPipeline pipeline_;
- BindGroup bind_group_;
+ BindGroup bind_group_;
};
diff --git a/cnn_v3/src/gbuffer_effect.cc b/cnn_v3/src/gbuffer_effect.cc
index 82ad8b1..395c8bc 100644
--- a/cnn_v3/src/gbuffer_effect.cc
+++ b/cnn_v3/src/gbuffer_effect.cc
@@ -1,5 +1,6 @@
// GBufferEffect implementation
-// Rasterizes proxy geometry to MRT G-buffer, then packs into CNN v3 feature textures.
+// Rasterizes proxy geometry to MRT G-buffer, then packs into CNN v3 feature
+// textures.
#include "gbuffer_effect.h"
#include "3d/object.h"
@@ -10,9 +11,10 @@
#include <cstring>
#include <vector>
-// Shader source (loaded from asset at runtime — declared extern by the build system)
-// For standalone use outside the asset system, the caller must ensure the WGSL
-// source strings are available. They are declared here as weak-linkable externs.
+// Shader source (loaded from asset at runtime — declared extern by the build
+// system) For standalone use outside the asset system, the caller must ensure
+// the WGSL source strings are available. They are declared here as
+// weak-linkable externs.
extern const char* gbuf_raster_wgsl;
extern const char* gbuf_shadow_wgsl;
extern const char* gbuf_pack_wgsl;
@@ -20,7 +22,8 @@ extern const char* gbuf_pack_wgsl;
// Maximum number of objects the G-buffer supports per frame.
static const int kGBufMaxObjects = 256;
-// ObjectData struct that mirrors the WGSL layout in gbuf_raster.wgsl and renderer.h
+// ObjectData struct that mirrors the WGSL layout in gbuf_raster.wgsl and
+// renderer.h
struct GBufObjectData {
mat4 model;
mat4 inv_model;
@@ -38,7 +41,7 @@ struct GBufGlobalUniforms {
mat4 view_proj;
mat4 inv_view_proj;
vec4 camera_pos_time;
- vec4 params; // x = num_objects
+ vec4 params; // x = num_objects
vec2 resolution;
vec2 padding;
};
@@ -56,12 +59,12 @@ GBufferEffect::GBufferEffect(const GpuContext& ctx,
// Derive internal node name prefix from the first output name.
const std::string& prefix = outputs.empty() ? "gbuf" : outputs[0];
- node_albedo_ = prefix + "_albedo";
+ node_albedo_ = prefix + "_albedo";
node_normal_mat_ = prefix + "_normal_mat";
- node_depth_ = prefix + "_depth";
- node_shadow_ = prefix + "_shadow";
- node_transp_ = prefix + "_transp";
- node_prev_tex_ = prefix + "_prev";
+ node_depth_ = prefix + "_depth";
+ node_shadow_ = prefix + "_shadow";
+ node_transp_ = prefix + "_transp";
+ node_prev_tex_ = prefix + "_prev";
// Allocate GPU buffers for scene data.
global_uniforms_buf_ =
gpu_create_buffer(ctx_.device, sizeof(GBufGlobalUniforms),
@@ -83,11 +86,11 @@ GBufferEffect::GBufferEffect(const GpuContext& ctx,
}
void GBufferEffect::declare_nodes(NodeRegistry& registry) {
- registry.declare_node(node_albedo_, NodeType::GBUF_ALBEDO, -1, -1);
- registry.declare_node(node_normal_mat_, NodeType::GBUF_ALBEDO, -1, -1);
- registry.declare_node(node_depth_, NodeType::GBUF_DEPTH32, -1, -1);
- registry.declare_node(node_shadow_, NodeType::GBUF_R8, -1, -1);
- registry.declare_node(node_transp_, NodeType::GBUF_R8, -1, -1);
+ registry.declare_node(node_albedo_, NodeType::GBUF_ALBEDO, -1, -1);
+ registry.declare_node(node_normal_mat_, NodeType::GBUF_ALBEDO, -1, -1);
+ registry.declare_node(node_depth_, NodeType::GBUF_DEPTH32, -1, -1);
+ registry.declare_node(node_shadow_, NodeType::GBUF_R8, -1, -1);
+ registry.declare_node(node_transp_, NodeType::GBUF_R8, -1, -1);
// feat_tex0 / feat_tex1 are the declared output_nodes_ — they get registered
// by the sequence infrastructure; declare them here as well if not already.
if (!registry.has_node(output_nodes_[0])) {
@@ -96,7 +99,8 @@ void GBufferEffect::declare_nodes(NodeRegistry& registry) {
if (!registry.has_node(output_nodes_[1])) {
registry.declare_node(output_nodes_[1], NodeType::GBUF_RGBA32UINT, -1, -1);
}
- // F16X8 = Rgba16Float with CopySrc|CopyDst — matches CNNv3Effect output format.
+ // F16X8 = Rgba16Float with CopySrc|CopyDst — matches CNNv3Effect output
+ // format.
registry.declare_node(node_prev_tex_, NodeType::F16X8, -1, -1);
}
@@ -111,7 +115,7 @@ void GBufferEffect::set_scene() {
seed ^= seed << 13;
seed ^= seed >> 17;
seed ^= seed << 5;
- return (float)(seed >> 8) / 16777216.0f; // [0, 1)
+ return (float)(seed >> 8) / 16777216.0f; // [0, 1)
};
auto rrange = [&](float lo, float hi) { return lo + rnd() * (hi - lo); };
@@ -120,8 +124,8 @@ void GBufferEffect::set_scene() {
{
Object3D obj(ObjectType::BOX);
obj.position = vec3(1.0f, 0.0f, 0.0f);
- obj.scale = vec3(0.6f, 0.6f, 0.6f);
- obj.color = vec4(0.9f, 0.5f, 0.3f, 1.0f);
+ obj.scale = vec3(0.6f, 0.6f, 0.6f);
+ obj.color = vec4(0.9f, 0.5f, 0.3f, 1.0f);
scene_.add_object(obj);
cube_anims_.push_back({{0.0f, 1.0f, 0.0f}, 0.0f});
}
@@ -129,8 +133,8 @@ void GBufferEffect::set_scene() {
Object3D obj(ObjectType::SPHERE);
obj.position = vec3(-1.0f, 0.0f, 0.0f);
const float r = 0.9f;
- obj.scale = vec3(r, r, r);
- obj.color = vec4(0.3f, 0.6f, 0.9f, 1.0f);
+ obj.scale = vec3(r, r, r);
+ obj.color = vec4(0.3f, 0.6f, 0.9f, 1.0f);
const int idx = (int)scene_.objects.size();
scene_.add_object(obj);
sphere_anims_.push_back({idx, r});
@@ -141,36 +145,35 @@ void GBufferEffect::set_scene() {
Object3D obj(ObjectType::SPHERE);
obj.position = vec3(0.0f, 2.2f, 0.0f);
const float r = 0.6f;
- obj.scale = vec3(r, r, r);
- obj.color = vec4(0.9f, 0.8f, 0.2f, 1.0f);
+ obj.scale = vec3(r, r, r);
+ obj.color = vec4(0.9f, 0.8f, 0.2f, 1.0f);
const int idx = (int)scene_.objects.size();
scene_.add_object(obj);
sphere_anims_.push_back({idx, r});
}
// Camera: above and in front of the scene, looking at origin.
- camera_.set_look_at(vec3(0.0f, 2.5f, 6.0f),
- vec3(0.0f, 0.0f, 0.0f),
+ camera_.set_look_at(vec3(0.0f, 2.5f, 6.0f), vec3(0.0f, 0.0f, 0.0f),
vec3(0.0f, 1.0f, 0.0f));
- camera_.fov_y_rad = 0.7854f; // 45°
+ camera_.fov_y_rad = 0.7854f; // 45°
camera_.near_plane = 0.1f;
- camera_.far_plane = 20.0f;
+ camera_.far_plane = 20.0f;
// aspect_ratio is updated each frame from params.resolution.
scene_ready_ = true;
}
static void clear_r8_node(WGPUCommandEncoder encoder, WGPUTextureView view,
- float value) {
+ float value) {
WGPURenderPassColorAttachment att = {};
- att.view = view;
- att.loadOp = WGPULoadOp_Clear;
- att.storeOp = WGPUStoreOp_Store;
+ att.view = view;
+ att.loadOp = WGPULoadOp_Clear;
+ att.storeOp = WGPUStoreOp_Store;
att.clearValue = {value, value, value, value};
att.depthSlice = WGPU_DEPTH_SLICE_UNDEFINED;
WGPURenderPassDescriptor pd = {};
pd.colorAttachmentCount = 1;
- pd.colorAttachments = &att;
+ pd.colorAttachments = &att;
WGPURenderPassEncoder p = wgpuCommandEncoderBeginRenderPass(encoder, &pd);
wgpuRenderPassEncoderEnd(p);
wgpuRenderPassEncoderRelease(p);
@@ -190,8 +193,7 @@ void GBufferEffect::render(WGPUCommandEncoder encoder,
const float angle = params.time * 0.3f;
const float R = 6.0f;
camera_.set_look_at(vec3(R * sinf(angle), 2.5f, R * cosf(angle)),
- vec3(0.0f, 0.0f, 0.0f),
- vec3(0.0f, 1.0f, 0.0f));
+ vec3(0.0f, 0.0f, 0.0f), vec3(0.0f, 1.0f, 0.0f));
// Animate cubes: axis-angle rotation driven by physical time.
for (int i = 0; i < (int)cube_anims_.size(); ++i) {
@@ -210,8 +212,8 @@ void GBufferEffect::render(WGPUCommandEncoder encoder,
GBufLightsUniforms lu = {};
lu.params = vec4(1.0f, 0.0f, 0.0f, 0.0f);
// Key: warm sun, upper-right-front.
- lu.lights[0].direction = vec4(0.408f, 0.816f, 0.408f, 0.0f); // norm(1,2,1)
- lu.lights[0].color = vec4(1.00f, 0.92f, 0.78f, 1.0f);
+ lu.lights[0].direction = vec4(0.408f, 0.816f, 0.408f, 0.0f); // norm(1,2,1)
+ lu.lights[0].color = vec4(1.00f, 0.92f, 0.78f, 1.0f);
// Fill: cool sky, upper-left-back. (disabled for debugging)
// lu.lights[1].direction = vec4(-0.577f, 0.577f, -0.577f, 0.0f);
// lu.lights[1].color = vec4(0.40f, 0.45f, 0.80f, 0.4f);
@@ -227,11 +229,11 @@ void GBufferEffect::render(WGPUCommandEncoder encoder,
res_uni._pad1 = 0.0f;
pack_res_uniform_.update(ctx_.queue, res_uni);
- WGPUTextureView albedo_view = nodes.get_view(node_albedo_);
+ WGPUTextureView albedo_view = nodes.get_view(node_albedo_);
WGPUTextureView normal_mat_view = nodes.get_view(node_normal_mat_);
- WGPUTextureView depth_view = nodes.get_view(node_depth_);
- WGPUTextureView feat0_view = nodes.get_view(output_nodes_[0]);
- WGPUTextureView feat1_view = nodes.get_view(output_nodes_[1]);
+ WGPUTextureView depth_view = nodes.get_view(node_depth_);
+ WGPUTextureView feat0_view = nodes.get_view(output_nodes_[0]);
+ WGPUTextureView feat1_view = nodes.get_view(output_nodes_[1]);
// node_prev_tex_ is updated by post_render() at the end of each frame.
// On frame 0 it is zero (NodeRegistry zeroes new textures) — correct default.
@@ -266,17 +268,16 @@ void GBufferEffect::render(WGPUCommandEncoder encoder,
raster_pass_desc.colorAttachments = color_attachments;
raster_pass_desc.depthStencilAttachment = &depth_attachment;
- const int num_objects =
- (int)(scene_.objects.size() < (size_t)kGBufMaxObjects
- ? scene_.objects.size()
- : (size_t)kGBufMaxObjects);
+ const int num_objects = (int)(scene_.objects.size() < (size_t)kGBufMaxObjects
+ ? scene_.objects.size()
+ : (size_t)kGBufMaxObjects);
if (num_objects > 0 && raster_pipeline_.get() != nullptr) {
WGPURenderPassEncoder raster_pass =
wgpuCommandEncoderBeginRenderPass(encoder, &raster_pass_desc);
wgpuRenderPassEncoderSetPipeline(raster_pass, raster_pipeline_.get());
- wgpuRenderPassEncoderSetBindGroup(raster_pass, 0,
- raster_bind_group_.get(), 0, nullptr);
+ wgpuRenderPassEncoderSetBindGroup(raster_pass, 0, raster_bind_group_.get(),
+ 0, nullptr);
// Draw 36 vertices (proxy box) × num_objects instances.
wgpuRenderPassEncoderDraw(raster_pass, 36, (uint32_t)num_objects, 0, 0);
wgpuRenderPassEncoderEnd(raster_pass);
@@ -293,45 +294,46 @@ void GBufferEffect::render(WGPUCommandEncoder encoder,
if (shadow_pipeline_.get() != nullptr) {
WGPUBindGroupEntry shadow_entries[5] = {};
shadow_entries[0].binding = 0;
- shadow_entries[0].buffer = global_uniforms_buf_.buffer;
- shadow_entries[0].size = sizeof(GBufGlobalUniforms);
+ shadow_entries[0].buffer = global_uniforms_buf_.buffer;
+ shadow_entries[0].size = sizeof(GBufGlobalUniforms);
shadow_entries[1].binding = 1;
- shadow_entries[1].buffer = objects_buf_.buffer;
- shadow_entries[1].size = (size_t)objects_buf_capacity_ * sizeof(GBufObjectData);
+ shadow_entries[1].buffer = objects_buf_.buffer;
+ shadow_entries[1].size =
+ (size_t)objects_buf_capacity_ * sizeof(GBufObjectData);
- shadow_entries[2].binding = 2;
+ shadow_entries[2].binding = 2;
shadow_entries[2].textureView = depth_view;
shadow_entries[3].binding = 3;
- shadow_entries[3].buffer = lights_uniform_.get().buffer;
- shadow_entries[3].size = sizeof(GBufLightsUniforms);
+ shadow_entries[3].buffer = lights_uniform_.get().buffer;
+ shadow_entries[3].size = sizeof(GBufLightsUniforms);
- shadow_entries[4].binding = 4;
+ shadow_entries[4].binding = 4;
shadow_entries[4].textureView = normal_mat_view;
WGPUBindGroupLayout shadow_bgl =
wgpuRenderPipelineGetBindGroupLayout(shadow_pipeline_.get(), 0);
WGPUBindGroupDescriptor shadow_bg_desc = {};
- shadow_bg_desc.layout = shadow_bgl;
+ shadow_bg_desc.layout = shadow_bgl;
shadow_bg_desc.entryCount = 5;
- shadow_bg_desc.entries = shadow_entries;
+ shadow_bg_desc.entries = shadow_entries;
WGPUBindGroup shadow_bg =
wgpuDeviceCreateBindGroup(ctx_.device, &shadow_bg_desc);
wgpuBindGroupLayoutRelease(shadow_bgl);
WGPURenderPassColorAttachment shadow_att = {};
- shadow_att.view = nodes.get_view(node_shadow_);
- shadow_att.loadOp = WGPULoadOp_Clear;
- shadow_att.storeOp = WGPUStoreOp_Store;
+ shadow_att.view = nodes.get_view(node_shadow_);
+ shadow_att.loadOp = WGPULoadOp_Clear;
+ shadow_att.storeOp = WGPUStoreOp_Store;
shadow_att.clearValue = {1.0f, 1.0f, 1.0f, 1.0f};
shadow_att.depthSlice = WGPU_DEPTH_SLICE_UNDEFINED;
WGPURenderPassDescriptor shadow_pass_desc = {};
shadow_pass_desc.colorAttachmentCount = 1;
- shadow_pass_desc.colorAttachments = &shadow_att;
+ shadow_pass_desc.colorAttachments = &shadow_att;
WGPURenderPassEncoder shadow_pass =
wgpuCommandEncoderBeginRenderPass(encoder, &shadow_pass_desc);
@@ -408,7 +410,7 @@ void GBufferEffect::render(WGPUCommandEncoder encoder,
wgpuComputePassEncoderSetPipeline(compute_pass, pack_pipeline_.get());
wgpuComputePassEncoderSetBindGroup(compute_pass, 0, pack_bg, 0, nullptr);
- const uint32_t wg_x = ((uint32_t)width_ + 7u) / 8u;
+ const uint32_t wg_x = ((uint32_t)width_ + 7u) / 8u;
const uint32_t wg_y = ((uint32_t)height_ + 7u) / 8u;
wgpuComputePassEncoderDispatchWorkgroups(compute_pass, wg_x, wg_y, 1);
wgpuComputePassEncoderEnd(compute_pass);
@@ -433,29 +435,29 @@ void GBufferEffect::ensure_objects_buffer(int num_objects) {
objects_buf_capacity_ = num_objects;
}
-void GBufferEffect::upload_scene_data(const Scene& scene,
- const Camera& camera, float time) {
- const int num_objects =
- (int)(scene.objects.size() < (size_t)kGBufMaxObjects
- ? scene.objects.size()
- : (size_t)kGBufMaxObjects);
+void GBufferEffect::upload_scene_data(const Scene& scene, const Camera& camera,
+ float time) {
+ const int num_objects = (int)(scene.objects.size() < (size_t)kGBufMaxObjects
+ ? scene.objects.size()
+ : (size_t)kGBufMaxObjects);
const mat4 view = camera.get_view_matrix();
mat4 proj = camera.get_projection_matrix();
- proj.m[5] = -proj.m[5]; // undo post-process Y flip: G-buffer uses integer reads
- const mat4 vp = proj * view;
+ proj.m[5] =
+ -proj.m[5]; // undo post-process Y flip: G-buffer uses integer reads
+ const mat4 vp = proj * view;
GBufGlobalUniforms gu = {};
- gu.view_proj = vp;
- gu.inv_view_proj = vp.inverse();
- gu.camera_pos_time = vec4(camera.position.x, camera.position.y,
- camera.position.z, time);
- gu.params = vec4((float)num_objects, 0.0f, 0.0f, 0.0f);
+ gu.view_proj = vp;
+ gu.inv_view_proj = vp.inverse();
+ gu.camera_pos_time =
+ vec4(camera.position.x, camera.position.y, camera.position.z, time);
+ gu.params = vec4((float)num_objects, 0.0f, 0.0f, 0.0f);
gu.resolution = vec2((float)width_, (float)height_);
- gu.padding = vec2(0.0f, 0.0f);
+ gu.padding = vec2(0.0f, 0.0f);
- wgpuQueueWriteBuffer(ctx_.queue, global_uniforms_buf_.buffer, 0,
- &gu, sizeof(GBufGlobalUniforms));
+ wgpuQueueWriteBuffer(ctx_.queue, global_uniforms_buf_.buffer, 0, &gu,
+ sizeof(GBufGlobalUniforms));
// Upload object data (no per-frame heap alloc — reuse s_obj_staging).
if (num_objects > 0) {
@@ -463,13 +465,12 @@ void GBufferEffect::upload_scene_data(const Scene& scene,
for (int i = 0; i < num_objects; ++i) {
const Object3D& obj = scene.objects[(size_t)i];
const mat4 m = obj.get_model_matrix();
- s_obj_staging[i].model = m;
+ s_obj_staging[i].model = m;
s_obj_staging[i].inv_model = m.inverse();
- s_obj_staging[i].color = obj.color;
- s_obj_staging[i].params = vec4((float)(int)obj.type, 0.0f, 0.0f, 0.0f);
+ s_obj_staging[i].color = obj.color;
+ s_obj_staging[i].params = vec4((float)(int)obj.type, 0.0f, 0.0f, 0.0f);
}
- wgpuQueueWriteBuffer(ctx_.queue, objects_buf_.buffer, 0,
- s_obj_staging,
+ wgpuQueueWriteBuffer(ctx_.queue, objects_buf_.buffer, 0, s_obj_staging,
(size_t)num_objects * sizeof(GBufObjectData));
}
}
@@ -483,8 +484,7 @@ void GBufferEffect::create_raster_pipeline() {
return; // Asset not loaded yet; pipeline creation deferred.
}
- const std::string composed =
- ShaderComposer::Get().Compose({}, src);
+ const std::string composed = ShaderComposer::Get().Compose({}, src);
WGPUShaderSourceWGSL wgsl_src = {};
wgsl_src.chain.sType = WGPUSType_ShaderSourceWGSL;
@@ -492,7 +492,8 @@ void GBufferEffect::create_raster_pipeline() {
WGPUShaderModuleDescriptor shader_desc = {};
shader_desc.nextInChain = &wgsl_src.chain;
- WGPUShaderModule shader = wgpuDeviceCreateShaderModule(ctx_.device, &shader_desc);
+ WGPUShaderModule shader =
+ wgpuDeviceCreateShaderModule(ctx_.device, &shader_desc);
// Bind group layout: B0 = GlobalUniforms, B1 = ObjectsBuffer (storage read)
WGPUBindGroupLayoutEntry bgl_entries[2] = {};
@@ -511,7 +512,8 @@ void GBufferEffect::create_raster_pipeline() {
WGPUBindGroupLayoutDescriptor bgl_desc = {};
bgl_desc.entryCount = 2;
bgl_desc.entries = bgl_entries;
- WGPUBindGroupLayout bgl = wgpuDeviceCreateBindGroupLayout(ctx_.device, &bgl_desc);
+ WGPUBindGroupLayout bgl =
+ wgpuDeviceCreateBindGroupLayout(ctx_.device, &bgl_desc);
WGPUPipelineLayoutDescriptor pl_desc = {};
pl_desc.bindGroupLayoutCount = 1;
@@ -542,9 +544,9 @@ void GBufferEffect::create_raster_pipeline() {
pipe_desc.vertex.entryPoint = str_view("vs_main");
pipe_desc.fragment = &frag;
pipe_desc.depthStencil = &ds;
- pipe_desc.primitive.topology = WGPUPrimitiveTopology_TriangleList;
- pipe_desc.primitive.cullMode = WGPUCullMode_Back;
- pipe_desc.primitive.frontFace = WGPUFrontFace_CCW; // standard (no Y flip)
+ pipe_desc.primitive.topology = WGPUPrimitiveTopology_TriangleList;
+ pipe_desc.primitive.cullMode = WGPUCullMode_Back;
+ pipe_desc.primitive.frontFace = WGPUFrontFace_CCW; // standard (no Y flip)
pipe_desc.multisample.count = 1;
pipe_desc.multisample.mask = 0xFFFFFFFF;
@@ -571,9 +573,11 @@ void GBufferEffect::create_shadow_pipeline() {
WGPUShaderModuleDescriptor shader_desc = {};
shader_desc.nextInChain = &wgsl_src.chain;
- WGPUShaderModule shader = wgpuDeviceCreateShaderModule(ctx_.device, &shader_desc);
+ WGPUShaderModule shader =
+ wgpuDeviceCreateShaderModule(ctx_.device, &shader_desc);
- // BGL: B0=GlobalUniforms, B1=ObjectsBuffer, B2=texture_depth_2d, B3=GBufLightsUniforms
+ // BGL: B0=GlobalUniforms, B1=ObjectsBuffer, B2=texture_depth_2d,
+ // B3=GBufLightsUniforms
WGPUBindGroupLayoutEntry bgl_entries[5] = {};
bgl_entries[0].binding = 0;
@@ -605,7 +609,8 @@ void GBufferEffect::create_shadow_pipeline() {
WGPUBindGroupLayoutDescriptor bgl_desc = {};
bgl_desc.entryCount = 5;
bgl_desc.entries = bgl_entries;
- WGPUBindGroupLayout bgl = wgpuDeviceCreateBindGroupLayout(ctx_.device, &bgl_desc);
+ WGPUBindGroupLayout bgl =
+ wgpuDeviceCreateBindGroupLayout(ctx_.device, &bgl_desc);
WGPUPipelineLayoutDescriptor pl_desc = {};
pl_desc.bindGroupLayoutCount = 1;
@@ -656,7 +661,8 @@ void GBufferEffect::create_pack_pipeline() {
WGPUShaderModuleDescriptor shader_desc = {};
shader_desc.nextInChain = &wgsl_src.chain;
- WGPUShaderModule shader = wgpuDeviceCreateShaderModule(ctx_.device, &shader_desc);
+ WGPUShaderModule shader =
+ wgpuDeviceCreateShaderModule(ctx_.device, &shader_desc);
// Build explicit bind group layout for bindings 0-9.
WGPUBindGroupLayoutEntry bgl_entries[10] = {};
@@ -725,7 +731,8 @@ void GBufferEffect::create_pack_pipeline() {
WGPUBindGroupLayoutDescriptor bgl_desc = {};
bgl_desc.entryCount = 10;
bgl_desc.entries = bgl_entries;
- WGPUBindGroupLayout bgl = wgpuDeviceCreateBindGroupLayout(ctx_.device, &bgl_desc);
+ WGPUBindGroupLayout bgl =
+ wgpuDeviceCreateBindGroupLayout(ctx_.device, &bgl_desc);
WGPUPipelineLayoutDescriptor pl_desc = {};
pl_desc.bindGroupLayoutCount = 1;
@@ -755,12 +762,12 @@ void GBufferEffect::update_raster_bind_group(NodeRegistry& nodes) {
WGPUBindGroupEntry entries[2] = {};
entries[0].binding = 0;
- entries[0].buffer = global_uniforms_buf_.buffer;
- entries[0].size = sizeof(GBufGlobalUniforms);
+ entries[0].buffer = global_uniforms_buf_.buffer;
+ entries[0].size = sizeof(GBufGlobalUniforms);
entries[1].binding = 1;
- entries[1].buffer = objects_buf_.buffer;
- entries[1].size = (size_t)objects_buf_capacity_ * sizeof(GBufObjectData);
+ entries[1].buffer = objects_buf_.buffer;
+ entries[1].size = (size_t)objects_buf_capacity_ * sizeof(GBufObjectData);
WGPUBindGroupLayout bgl =
wgpuRenderPipelineGetBindGroupLayout(raster_pipeline_.get(), 0);
@@ -777,20 +784,23 @@ void GBufferEffect::update_raster_bind_group(NodeRegistry& nodes) {
void GBufferEffect::wire_dag(const std::vector<EffectDAGNode>& dag) {
const std::string out = find_downstream_output(dag);
// "sink" is an external view (no owned texture) — not a valid copy source.
- if (out != "sink") cnn_output_node_ = out;
+ if (out != "sink")
+ cnn_output_node_ = out;
}
-void GBufferEffect::post_render(WGPUCommandEncoder encoder, NodeRegistry& nodes) {
- if (cnn_output_node_.empty() || !nodes.has_node(cnn_output_node_)) return;
+void GBufferEffect::post_render(WGPUCommandEncoder encoder,
+ NodeRegistry& nodes) {
+ if (cnn_output_node_.empty() || !nodes.has_node(cnn_output_node_))
+ return;
WGPUTexture src_tex = nodes.get_texture(cnn_output_node_);
- if (!src_tex) return; // external view (e.g. sink) — no owned texture to copy
+ if (!src_tex)
+ return; // external view (e.g. sink) — no owned texture to copy
WGPUTexelCopyTextureInfo src = {};
- src.texture = src_tex;
+ src.texture = src_tex;
src.mipLevel = 0;
WGPUTexelCopyTextureInfo dst = {};
- dst.texture = nodes.get_texture(node_prev_tex_);
+ dst.texture = nodes.get_texture(node_prev_tex_);
dst.mipLevel = 0;
WGPUExtent3D extent = {(uint32_t)width_, (uint32_t)height_, 1};
wgpuCommandEncoderCopyTextureToTexture(encoder, &src, &dst, &extent);
}
-
diff --git a/cnn_v3/src/gbuffer_effect.h b/cnn_v3/src/gbuffer_effect.h
index 76d4347..260444a 100644
--- a/cnn_v3/src/gbuffer_effect.h
+++ b/cnn_v3/src/gbuffer_effect.h
@@ -1,5 +1,6 @@
// GBufferEffect: Multi-pass G-buffer rendering for CNN v3 input
-// Outputs: gbuf_feat0, gbuf_feat1 (packed rgba32uint feature textures, 32 bytes/pixel)
+// Outputs: gbuf_feat0, gbuf_feat1 (packed rgba32uint feature textures, 32
+// bytes/pixel)
#pragma once
@@ -21,16 +22,17 @@ struct GBufResUniforms {
static_assert(sizeof(GBufResUniforms) == 16,
"GBufResUniforms must be 16 bytes");
-// Single directional light: direction points *toward* the light source (world space).
+// Single directional light: direction points *toward* the light source (world
+// space).
struct GBufLight {
- vec4 direction; // xyz = normalized direction toward light, w = unused
- vec4 color; // rgb = color, a = intensity
+ vec4 direction; // xyz = normalized direction toward light, w = unused
+ vec4 color; // rgb = color, a = intensity
};
static_assert(sizeof(GBufLight) == 32, "GBufLight must be 32 bytes");
struct GBufLightsUniforms {
GBufLight lights[2];
- vec4 params; // x = num_lights
+ vec4 params; // x = num_lights
};
static_assert(sizeof(GBufLightsUniforms) == 80,
"GBufLightsUniforms must be 80 bytes");
@@ -60,12 +62,12 @@ class GBufferEffect : public Effect {
private:
// Per-cube animation state (axis-angle rotation)
struct CubeAnim {
- vec3 axis;
- float speed; // radians/second, may be negative
+ vec3 axis;
+ float speed; // radians/second, may be negative
};
// Per-sphere animation state (radius driven by audio_intensity)
struct SphereAnim {
- int obj_idx; // index into scene_.objects
+ int obj_idx; // index into scene_.objects
float base_radius;
};
@@ -75,41 +77,42 @@ class GBufferEffect : public Effect {
std::string node_depth_;
std::string node_shadow_;
std::string node_transp_;
- std::string node_prev_tex_; // persistent prev-frame CNN output (rgba8unorm)
+ std::string node_prev_tex_; // persistent prev-frame CNN output (rgba8unorm)
// Name of the CNN effect's output node; set by caller before first render.
// When non-empty, the CNN output is copied into node_prev_tex_ each frame.
std::string cnn_output_node_;
public:
- void set_cnn_output_node(const std::string& name) { cnn_output_node_ = name; }
+ void set_cnn_output_node(const std::string& name) {
+ cnn_output_node_ = name;
+ }
private:
-
// Owned scene and camera — populated by set_scene()
- Scene scene_;
+ Scene scene_;
Camera camera_;
- bool scene_ready_ = false;
+ bool scene_ready_ = false;
- std::vector<CubeAnim> cube_anims_;
+ std::vector<CubeAnim> cube_anims_;
std::vector<SphereAnim> sphere_anims_;
// Pass 1: MRT rasterization pipeline
RenderPipeline raster_pipeline_;
- BindGroup raster_bind_group_;
+ BindGroup raster_bind_group_;
// Pass 2: SDF shadow pipeline
RenderPipeline shadow_pipeline_;
// Pass 4: Pack compute pipeline
- ComputePipeline pack_pipeline_;
- UniformBuffer<GBufResUniforms> pack_res_uniform_;
+ ComputePipeline pack_pipeline_;
+ UniformBuffer<GBufResUniforms> pack_res_uniform_;
UniformBuffer<GBufLightsUniforms> lights_uniform_;
// GPU-side object data buffers (global uniforms + objects storage)
GpuBuffer global_uniforms_buf_;
GpuBuffer objects_buf_;
- int objects_buf_capacity_ = 0;
+ int objects_buf_capacity_ = 0;
void create_raster_pipeline();
void create_shadow_pipeline();
@@ -117,7 +120,6 @@ class GBufferEffect : public Effect {
void update_raster_bind_group(NodeRegistry& nodes);
- void upload_scene_data(const Scene& scene, const Camera& camera,
- float time);
+ void upload_scene_data(const Scene& scene, const Camera& camera, float time);
void ensure_objects_buffer(int num_objects);
};
generated by cgit v1.2.3 (git 2.51.0) at 2026-05-30 01:00:03 +0000