molive/tape-drive
4
0
Fork 0
Code Issues 1 Pull Requests Actions Packages Projects Releases Wiki Activity

--wip-- [skip ci]

Browse Source
This commit is contained in:
Molive
2025-07-08 23:20:20 +01:00
parent e4eb599744
commit 52cc5c7385
11 changed files with 2338 additions and 803 deletions
Download Patch File Download Diff File Expand all files Collapse all files
opt.spv
BIN
View File
Binary file not shown.
src/implicit_include.glsl
+8
View File
@@ -8,4 +8,12 @@ layout(set = 1, binding = 0) uniform Object {
uint material;
} object_uniforms;
float scene(vec4 p) {
return 0.0;
}
vec2 interval_scene(vec4 pl, vec4 ph) {
return vec2(0.0);
}
#endif
src/instruction_set.rs
-4
View File
@@ -198,10 +198,6 @@ pub(crate) enum InstructionSet {
OPSmoothMaxMaterial = ((1 * 64) + 62),
#[allow(non_snake_case)]
#[allow(dead_code)]
/// Returns the cross product of two Vec3s.
OPCross = ((0 * 64) + 62),
#[allow(non_snake_case)]
#[allow(dead_code)]
/// Returns the distance to a sphere.
OPSDFSphere = ((3 * 64) + 61),
#[allow(non_snake_case)]
src/interpreter.rs
+2 -22
View File
@@ -20,6 +20,8 @@ pub const VALUE_05: Value = Value::splat(0.5);
pub const VALUE_M1: Value = Value::splat(-1.0);
pub const VALUE_2: Value = Value::splat(2.0);
pub const VALUE_PI: Value = Value::splat(core::f32::consts::PI);
pub const VALUE_PI_2: Value = Value::splat(core::f32::consts::FRAC_PI_2);
pub const VALUE_TAU: Value = Value::splat(core::f32::consts::TAU);
pub fn glsign(f: Value) -> Value {
if BYTECODE_IMITATES_GLSL {
@@ -159,17 +161,6 @@ impl<'csg> PointInterpreter<'csg> {
self.param_two(instruction, |val_a, val_b| val_a.simd_max(val_b));
},
SSAMaxMaterial => todo!(),
SSACross => {
let a_x = self.load(instruction.inputs[0]);
let a_y = self.load(instruction.inputs[1]);
let a_z = self.load(instruction.inputs[2]);
let b_x = self.load(instruction.inputs[3]);
let b_y = self.load(instruction.inputs[4]);
let b_z = self.load(instruction.inputs[5]);
self.store(instruction.outputs[0], (a_y * b_z) + (a_z * b_y));
self.store(instruction.outputs[1], (a_z * b_x) + (a_x * b_z));
self.store(instruction.outputs[2], (a_x * b_y) + (a_y * b_x));
},
SSADot => {
let val_a = (0..instruction.opcode.size)
.map(|i| self.load(instruction.inputs[i as usize]))
@@ -536,17 +527,6 @@ impl<'csg> IntervalInterpreter<'csg> {
self.param_two(instruction, |val_a, val_b| val_a.max_choice(val_b).0);
},
SSAMaxMaterial => todo!(),
SSACross => {
let a_x = self.load(instruction.inputs[0]);
let a_y = self.load(instruction.inputs[1]);
let a_z = self.load(instruction.inputs[2]);
let b_x = self.load(instruction.inputs[3]);
let b_y = self.load(instruction.inputs[4]);
let b_z = self.load(instruction.inputs[5]);
self.store(instruction.outputs[0], (a_y * b_z) + (a_z * b_y));
self.store(instruction.outputs[1], (a_z * b_x) + (a_x * b_z));
self.store(instruction.outputs[2], (a_x * b_y) + (a_y * b_x));
},
SSADot => {
let val_a = (0..instruction.opcode.size)
.map(|i| self.load(instruction.inputs[i as usize]))
src/lighting.frag.glsl
-7
View File
@@ -13,13 +13,6 @@ layout(location = 0) out vec4 f_color;
vec3 calculate_view_position(float depth)
{
vec4 view_position = camera_uniforms.inv_proj_view * vec4(v_pos, depth, 1.0);
//vec4 view_position = vec4(
// fma(camera_uniforms.inv_proj_view[0][0], v_pos.x, camera_uniforms.inv_proj_view[3][0]),
// fma(camera_uniforms.inv_proj_view[1][1], v_pos.y, camera_uniforms.inv_proj_view[3][1]),
// fma(camera_uniforms.inv_proj_view[2][2], depth, camera_uniforms.inv_proj_view[3][2]),
// fma(camera_uniforms.inv_proj_view[2][3], depth, camera_uniforms.inv_proj_view[3][3])
// );
return (view_position.xyz / view_position.w);
}
src/main.rs
+45 -5
View File
@@ -43,7 +43,7 @@ use vulkano::{
Device, DeviceCreateInfo, DeviceExtensions, DeviceFeatures, DeviceOwned, Queue,
QueueCreateInfo, QueueFlags, physical::PhysicalDeviceType,
},
format::{ClearValue, Format},
format::{ClearValue, Format, FormatFeatures},
image::{
Image, ImageAspects, ImageCreateInfo, ImageSubresourceRange, ImageType, ImageUsage,
SampleCount,
@@ -186,6 +186,7 @@ const CUBE_VERTEX: [IVertex; VERTEX_COUNT] = [
struct App {
instance: Arc<Instance>,
device: Arc<Device>,
depth_format: Format,
graphics_queue: Arc<Queue>,
transfer_queue: Arc<Queue>,
memory_allocator: Arc<StandardMemoryAllocator>,
@@ -321,6 +322,33 @@ impl App {
physical_device.properties().device_type,
);
let depth_format = if physical_device
.format_properties(Format::D24_UNORM_S8_UINT)
.unwrap()
.optimal_tiling_features
.contains(FormatFeatures::DEPTH_STENCIL_ATTACHMENT)
{
Format::D24_UNORM_S8_UINT
} else if physical_device
.format_properties(Format::D16_UNORM_S8_UINT)
.unwrap()
.optimal_tiling_features
.contains(FormatFeatures::DEPTH_STENCIL_ATTACHMENT)
{
Format::D16_UNORM_S8_UINT
} else if physical_device
.format_properties(Format::D32_SFLOAT_S8_UINT)
.unwrap()
.optimal_tiling_features
.contains(FormatFeatures::DEPTH_STENCIL_ATTACHMENT)
{
Format::D32_SFLOAT_S8_UINT
} else {
panic!("Could not find suitable depth format!");
};
info!("Using depth format {:?}", depth_format);
let (device, mut queues) = Device::new(
physical_device,
DeviceCreateInfo {
@@ -448,6 +476,7 @@ impl App {
App {
instance,
device,
depth_format,
graphics_queue,
transfer_queue,
memory_allocator,
@@ -659,7 +688,7 @@ impl ApplicationHandler for App {
store_op: DontCare,
},
depth: {
format: Format::D24_UNORM_S8_UINT,
format: self.depth_format,
samples: 1,
load_op: Clear,
store_op: DontCare,
@@ -701,7 +730,12 @@ impl ApplicationHandler for App {
ao_r_m_buffer,
normal_buffer,
depth_buffer,
) = framebuffer_generation(&images, &render_pass, &self.memory_allocator);
) = framebuffer_generation(
&images,
&render_pass,
&self.memory_allocator,
self.depth_format,
);
let lighting_pipeline = pipeline_recompile(
&render_pass,
@@ -1301,7 +1335,12 @@ impl App {
rcx.ao_r_m_buffer,
rcx.normal_buffer,
rcx.depth_buffer,
) = framebuffer_generation(&new_images, &rcx.render_pass, &self.memory_allocator);
) = framebuffer_generation(
&new_images,
&rcx.render_pass,
&self.memory_allocator,
self.depth_format,
);
rcx.viewport.extent[0] = window_size.width as f32;
rcx.viewport.extent[1] = window_size.height as f32 * -1.;
@@ -1486,6 +1525,7 @@ fn framebuffer_generation(
images: &[Arc<Image>],
render_pass: &Arc<RenderPass>,
allocator: &Arc<StandardMemoryAllocator>,
depth_format: Format,
) -> (
Vec<Arc<Framebuffer>>,
Arc<ImageView>,
@@ -1498,7 +1538,7 @@ fn framebuffer_generation(
allocator.clone(),
ImageCreateInfo {
image_type: ImageType::Dim2d,
format: Format::D24_UNORM_S8_UINT,
format: depth_format,
extent: images[0].extent(),
usage: ImageUsage::DEPTH_STENCIL_ATTACHMENT
| ImageUsage::INPUT_ATTACHMENT
src/normals.frag.glsl
-11
View File
@@ -14,20 +14,9 @@ layout(location = 0) out vec4 f_color;
layout(location = 1) out vec4 f_normal;
layout(location = 2) out vec4 f_ao_r_m;
float scene(vec4 p) {
return length(p.xyz) - 1.0;
}
vec3 calculate_view_position(float depth)
{
vec4 view_position = camera_uniforms.inv_proj_view * vec4(v_pos, depth, 1.0);
//vec4 view_position = vec4(
// fma(camera_uniforms.inv_proj_view[0][0], v_pos.x, camera_uniforms.inv_proj_view[3][0]),
// fma(camera_uniforms.inv_proj_view[1][1], v_pos.y, camera_uniforms.inv_proj_view[3][1]),
// fma(camera_uniforms.inv_proj_view[2][2], depth, camera_uniforms.inv_proj_view[3][2]),
// fma(camera_uniforms.inv_proj_view[2][3], depth, camera_uniforms.inv_proj_view[3][3])
// );
return (view_position.xyz / view_position.w);
}
src/ssa.rs
+2264 -709
View File
File diff suppressed because it is too large Load Diff
src/threads.rs
+4 -7
View File
@@ -174,11 +174,6 @@ pub fn thread_loop(recv: mpmc::Receiver<WorkItem>, send: mpsc::SyncSender<WorkCo
);
},
WorkItem::GetPushConstants(csg, time, frame_index, index) => {
let mut push_constants = PushConstantData {
world: Mat4::IDENTITY.to_cols_array_2d(),
inv_world: Mat4::IDENTITY.to_cols_array_2d(),
};
let csg = csg.read().unwrap();
let world = Mat4::from_translation(csg.pos * 0.01)
@@ -190,8 +185,10 @@ pub fn thread_loop(recv: mpmc::Receiver<WorkItem>, send: mpsc::SyncSender<WorkCo
)
* Mat4::from_scale(csg.scale * 2.0);
push_constants.world = world.to_cols_array_2d();
push_constants.inv_world = world.inverse().to_cols_array_2d();
let push_constants = PushConstantData {
world: world.to_cols_array_2d(),
inv_world: world.inverse().to_cols_array_2d(),
};
interval_check(&csg, index as u32 + 1, time, frame_index);
src/trace.frag.glsl
+1 -5
View File
@@ -11,10 +11,6 @@ layout(depth_greater) out float gl_FragDepth;
layout(location = 0) out uint material;
float scene(vec4 p) {
return length(p.xyz) - 1.0;
}
const float EPSILON = .001;
const uint MAX_STEPS = 50;
const float NEARPLANE = 0.;
@@ -26,7 +22,7 @@ vec2 spheretracing(vec3 ori, vec3 dir, out vec3 p) {
td.y = scene(vec4(p, camera_uniforms.campos_and_time.w)) * .98;
td.x += td.y;
p = fma(dir, vec3(td.x), ori);
for (int i = 0; i < MAX_STEPS && td.y > EPSILON && td.x < FARPLANE; i++) {
for (int i = 0; td.y > EPSILON && td.x < FARPLANE; i++) {
td.y = scene(vec4(p, camera_uniforms.campos_and_time.w)) * .98;
td.x += td.y;
p = fma(dir, vec3(td.x), ori);
src/types/interval.rs
+14 -33
View File
@@ -7,8 +7,8 @@ use std::simd::{
use crate::{
interpreter::{
Mask, VALUE_0, VALUE_1, VALUE_2, VALUE_05, VALUE_M1, VALUE_NAN, VALUE_PI, Value, glfract,
glsign,
Mask, VALUE_0, VALUE_1, VALUE_2, VALUE_05, VALUE_M1, VALUE_NAN, VALUE_PI, VALUE_PI_2,
VALUE_TAU, Value, glfract, glsign,
},
vm::choice::{Choice, VChoice},
};
@@ -40,6 +40,7 @@ impl Interval {
pub const NAN: Self = Self::const_splat(core::f32::NAN);
pub const ONE: Self = Self::const_splat(1.0);
pub const PI: Self = Self::const_splat(core::f32::consts::PI);
pub const PI_2: Self = Self::const_splat(core::f32::consts::FRAC_PI_2);
pub const ZERO: Self = Self::const_splat(0.0);
/// Builds a new interval
@@ -164,47 +165,27 @@ impl Interval {
/// Computes the sine of the interval
#[inline]
pub fn sin(self) -> Self {
let same_cycle = ((self.lower / VALUE_PI) + VALUE_05)
.floor()
.simd_eq(((self.upper / VALUE_PI) + VALUE_05).floor());
let up = (((self.upper / VALUE_PI) + VALUE_05).floor()) % (VALUE_2);
let whole_cycle = (((self.upper / VALUE_PI) + VALUE_05).floor()
- ((self.lower / VALUE_PI) + VALUE_05).floor())
.simd_gt(VALUE_1);
let temp0 = self.lower.sin();
let temp1 = self.upper.sin();
let lower = self.has_nan().select(
VALUE_NAN,
((!whole_cycle & (up.simd_eq(VALUE_1))) | same_cycle)
.select(temp0.simd_min(temp1), VALUE_M1),
);
let upper = self.has_nan().select(
VALUE_NAN,
((!whole_cycle & (up.simd_eq(VALUE_0))) | same_cycle)
.select(temp0.simd_max(temp1), VALUE_1),
);
Interval::new(lower, upper)
(self - Self::PI_2).cos()
}
/// Computes the cosine of the interval
#[inline]
pub fn cos(self) -> Self {
let same_cycle = (self.lower / VALUE_PI)
.floor()
.simd_eq((self.upper / VALUE_PI).floor());
let up = ((self.upper / VALUE_PI).floor()) % (VALUE_2);
let whole_cycle =
((self.upper / VALUE_PI).floor() - (self.lower / VALUE_PI).floor()).simd_gt(VALUE_1);
let lower_cycle = (self.lower / VALUE_PI).floor();
let upper_cycle = (self.upper / VALUE_PI).floor();
let same_cycle = lower_cycle.simd_eq(upper_cycle);
let cycle = upper_cycle % VALUE_2;
let within_one_cycle = (upper_cycle - lower_cycle).simd_eq(VALUE_1);
let temp0 = self.lower.cos();
let temp1 = self.upper.cos();
let lower = self.has_nan().select(
VALUE_NAN,
((!whole_cycle & (up.simd_eq(VALUE_0))) | same_cycle)
(same_cycle | (within_one_cycle & (cycle.simd_eq(VALUE_0))))
.select(temp0.simd_min(temp1), VALUE_M1),
);
let upper = self.has_nan().select(
VALUE_NAN,
((!whole_cycle & (up.simd_eq(VALUE_1))) | same_cycle)
(same_cycle | (within_one_cycle & (cycle.simd_eq(VALUE_1))))
.select(temp0.simd_max(temp1), VALUE_1),
);
Interval::new(lower, upper)
@@ -350,8 +331,8 @@ impl Interval {
);
(
Interval::new(
has_nan.select(VALUE_NAN, self.lower.simd_min(rhs.lower)),
has_nan.select(VALUE_NAN, self.upper.simd_min(rhs.upper)),
has_nan.select(VALUE_NAN, self.lower.simd_max(rhs.lower)),
has_nan.select(VALUE_NAN, self.upper.simd_max(rhs.upper)),
),
VChoice(choice),
)
@@ -629,7 +610,7 @@ impl std::ops::Rem<Interval> for Interval {
);
let upper = has_nan.select(
VALUE_NAN,
(other_constant & floors).select(self.upper % rhs.lower, rhs.upper.abs()),
(other_constant & floors).select(self.upper % rhs.lower, rhs.abs().upper),
);
Interval::new(lower, upper)