Rename pow to powf, remove saturate from MathExt, and update to latest glam (#248)

* remove saturate from spirv_std MathExt and rename pow to powf

* update to latest glam

* update glam again and add `std` feature to `spirv-std` to support cpu runner

* re-remove spirv-tools-sys submodule

* fixup! Merge branch 'main' into powf-saturate

* Use libm instead of MathExt

* Fix lint for unused import when compiling with std

Co-authored-by: khyperia <github@khyperia.com>

Cargo.lock

@@ -983,9 +983,10 @@ dependencies = [
 
 [[package]]
 name = "glam"
-version = "0.10.0"
+version = "0.10.2"
-source = "git+https://github.com/EmbarkStudios/glam-rs?rev=c9561e4dfd55fa5a9d6838cae3c9e90c8edafaf9#c9561e4dfd55fa5a9d6838cae3c9e90c8edafaf9"
+source = "git+https://github.com/EmbarkStudios/glam-rs?rev=ea8fe2766ca27f481656d3507cc129537e262182#ea8fe2766ca27f481656d3507cc129537e262182"
 dependencies = [
+ "num-traits",
  "version_check",
 ]
 
@@ -1124,6 +1125,12 @@ dependencies = [
  "winapi 0.3.9",
 ]
 
+[[package]]
+name = "libm"
+version = "0.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c7d73b3f436185384286bd8098d17ec07c9a7d2388a6599f824d8502b529702a"
+
 [[package]]
 name = "lock_api"
 version = "0.4.2"
@@ -1424,6 +1431,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "9a64b1ec5cda2586e284722486d802acf1f7dbdc623e2bfc57e65ca1cd099290"
 dependencies = [
  "autocfg 1.0.1",
+ "libm",
 ]
 
 [[package]]
@@ -2182,6 +2190,7 @@ name = "spirv-std"
 version = "0.1.0"
 dependencies = [
  "glam",
+ "num-traits",
 ]
 
 [[package]]

crates/spirv-std/Cargo.toml

@@ -8,4 +8,9 @@ repository = "https://github.com/EmbarkStudios/rust-gpu"
 
 [dependencies]
 # https://github.com/bitshifter/glam-rs/pull/85
-glam = { git = "https://github.com/EmbarkStudios/glam-rs", rev = "c9561e4dfd55fa5a9d6838cae3c9e90c8edafaf9" }
+glam = { git = "https://github.com/EmbarkStudios/glam-rs", rev = "ea8fe2766ca27f481656d3507cc129537e262182", default-features = false, features = ["libm"] }
+num-traits = { version = "0.2.14", default-features = false }
+
+[features]
+default = []
+std = ["glam/std"]

crates/spirv-std/src/lib.rs

@@ -37,10 +37,8 @@
     nonstandard_style
 )]
 
-mod math_ext;
-pub use math_ext::MathExt;
-
 pub use glam;
+pub use num_traits;
 
 macro_rules! pointer_addrspace_write {
     (false) => {};

crates/spirv-std/src/math_ext.rs

@@ -1,95 +0,0 @@
-pub trait MathExt {
-    fn pow(self, factor: Self) -> Self;
-    fn sqrt(self) -> Self;
-    fn log2(self) -> Self;
-    fn abs(self) -> Self;
-    fn cos(self) -> Self;
-    fn sin(self) -> Self;
-    fn sin_cos(self) -> (f32, f32);
-    fn round(self) -> Self;
-    fn floor(self) -> Self;
-    fn ceil(self) -> Self;
-    fn exp(self) -> Self;
-    fn saturate(self) -> Self;
-    fn trunc(self) -> Self;
-    fn fract(self) -> Self;
-
-    fn signum(self) -> Self;
-    fn copysign(self, sign: Self) -> Self;
-
-    // TODO: implement but need new intrinsic in core?
-    //fn tan(self) -> Self;
-}
-
-impl MathExt for f32 {
-    fn pow(self, factor: f32) -> f32 {
-        unsafe { core::intrinsics::powf32(self, factor) }
-    }
-
-    fn sqrt(self) -> f32 {
-        unsafe { core::intrinsics::sqrtf32(self) }
-    }
-
-    fn log2(self) -> f32 {
-        unsafe { core::intrinsics::log2f32(self) }
-    }
-
-    fn abs(self) -> f32 {
-        unsafe { core::intrinsics::fabsf32(self) }
-    }
-
-    fn cos(self) -> f32 {
-        unsafe { core::intrinsics::cosf32(self) }
-    }
-
-    fn sin(self) -> f32 {
-        unsafe { core::intrinsics::sinf32(self) }
-    }
-
-    // TODO: implement but need new intrinsic in core?
-    //fn tan(self) -> f32 {
-
-    fn sin_cos(self) -> (f32, f32) {
-        (self.sin(), self.cos())
-    }
-
-    fn round(self) -> f32 {
-        unsafe { core::intrinsics::roundf32(self) }
-    }
-
-    fn floor(self) -> f32 {
-        unsafe { core::intrinsics::floorf32(self) }
-    }
-
-    fn ceil(self) -> f32 {
-        unsafe { core::intrinsics::ceilf32(self) }
-    }
-
-    fn exp(self) -> f32 {
-        unsafe { core::intrinsics::expf32(self) }
-    }
-
-    fn saturate(self) -> f32 {
-        self.max(0.0).min(1.0)
-    }
-
-    fn trunc(self) -> f32 {
-        unsafe { core::intrinsics::truncf32(self) }
-    }
-
-    fn fract(self) -> f32 {
-        self - self.trunc()
-    }
-
-    fn signum(self) -> f32 {
-        if self.is_nan() {
-            Self::NAN
-        } else {
-            1.0_f32.copysign(self)
-        }
-    }
-
-    fn copysign(self, sign: f32) -> f32 {
-        unsafe { core::intrinsics::copysignf32(self, sign) }
-    }
-}

examples/runners/cpu/Cargo.toml

@@ -11,7 +11,7 @@ minifb = "0.19.1"
 # bring in the shader as natively compiled code
 shared = { path = "../../shaders/shared" }
 sky-shader = { path = "../../shaders/sky-shader" }
-spirv-std = { path = "../../../crates/spirv-std" }
+spirv-std = { path = "../../../crates/spirv-std", features = ["std"] }
 
 # for parallelism, not really needed though
 rayon = "1.5"

examples/runners/cpu/src/main.rs

@@ -18,10 +18,10 @@ fn srgb_oetf(x: f32) -> f32 {
 fn color_u32_from_vec4(v: Vec4) -> u32 {
     let convert = |f: f32| -> u32 { (f.min(1.0).max(0.0) * 255.0).round() as u32 };
 
-    convert(srgb_oetf(v.z()))
+    convert(srgb_oetf(v.z))
-        | convert(srgb_oetf(v.y())) << 8
+        | convert(srgb_oetf(v.y)) << 8
-        | convert(srgb_oetf(v.x())) << 16
+        | convert(srgb_oetf(v.x)) << 16
-        | convert(v.w()) << 24
+        | convert(v.w) << 24
 }
 
 fn main() {
@@ -55,8 +55,7 @@ fn main() {
                 -((i / WIDTH) as f32 / HEIGHT as f32 * 2.0 - 1.0),
             );
 
-            let frag_coord = (vec2(screen_pos.x(), -screen_pos.y()) + Vec2::one())
+            let frag_coord = (vec2(screen_pos.x, -screen_pos.y) + Vec2::one()) / Vec2::splat(2.0)
-                / Vec2::splat(2.0)
                 * vec2(WIDTH as f32, HEIGHT as f32);
 
             // evaluate the fragment shader for the specific pixel

examples/shaders/shared/src/lib.rs

@@ -6,8 +6,12 @@
 #![register_attr(spirv)]
 
 use core::f32::consts::PI;
-use spirv_std::glam::{Vec2, Vec3};
+use spirv_std::glam::Vec3;
-use spirv_std::MathExt;
+
+// Note: This cfg is incorrect on its surface, it really should be "are we compiling with std", but
+// we tie #[no_std] above to the same condition, so it's fine.
+#[cfg(target_arch = "spirv")]
+use spirv_std::num_traits::Float;
 
 #[derive(Copy, Clone)]
 pub struct ShaderConstants {
@@ -16,14 +20,16 @@ pub struct ShaderConstants {
     pub time: f32,
 }
 
-// TODO: add this to glam? Rust std has it on f32/f64
+pub fn saturate(x: f32) -> f32 {
-pub fn pow(v: Vec3, power: f32) -> Vec3 {
+    x.max(0.0).min(1.0)
-    Vec3::new(v.x().pow(power), v.y().pow(power), v.z().pow(power))
+}
+
+pub fn pow(v: Vec3, power: f32) -> Vec3 {
+    Vec3::new(v.x.powf(power), v.y.powf(power), v.z.powf(power))
 }
 
-// TODO: add this to glam? Rust std has it on f32/f64
 pub fn exp(v: Vec3) -> Vec3 {
-    Vec3::new(v.x().exp(), v.y().exp(), v.z().exp())
+    Vec3::new(v.x.exp(), v.y.exp(), v.z.exp())
 }
 
 /// Based on: https://seblagarde.wordpress.com/2014/12/01/inverse-trigonometric-functions-gpu-optimization-for-amd-gcn-architecture/
@@ -41,25 +47,7 @@ pub fn acos_approx(v: f32) -> f32 {
 
 pub fn smoothstep(edge0: f32, edge1: f32, x: f32) -> f32 {
     // Scale, bias and saturate x to 0..1 range
-    let x = ((x - edge0) / (edge1 - edge0)).saturate();
+    let x = saturate((x - edge0) / (edge1 - edge0));
     // Evaluate polynomial
     x * x * (3.0 - 2.0 * x)
 }
-
-pub fn tonemap(col: Vec3) -> Vec3 {
-    // see https://www.desmos.com/calculator/0eo9pzo1at
-    const A: f32 = 2.35;
-    const B: f32 = 2.8826666;
-    const C: f32 = 789.7459;
-    const D: f32 = 0.935;
-
-    let z = pow(col, A);
-    z / (pow(z, D) * B + Vec3::splat(C))
-}
-
-pub fn get_ray_dir(uv: Vec2, pos: Vec3, look_at_pos: Vec3) -> Vec3 {
-    let forward = (look_at_pos - pos).normalize();
-    let right = Vec3::new(0.0, 1.0, 0.0).cross(forward).normalize();
-    let up = forward.cross(right);
-    (forward + uv.x() * right + uv.y() * up).normalize()
-}

examples/shaders/sky-shader/src/lib.rs

@@ -8,7 +8,12 @@
 use core::f32::consts::PI;
 use shared::*;
 use spirv_std::glam::{const_vec3, Vec2, Vec3, Vec4};
-use spirv_std::{Input, MathExt, Output, PushConstant};
+use spirv_std::{Input, Output, PushConstant};
+
+// Note: This cfg is incorrect on its surface, it really should be "are we compiling with std", but
+// we tie #[no_std] above to the same condition, so it's fine.
+#[cfg(target_arch = "spirv")]
+use spirv_std::num_traits::Float;
 
 const DEPOLARIZATION_FACTOR: f32 = 0.035;
 const MIE_COEFFICIENT: f32 = 0.005;
@@ -26,9 +31,20 @@ const SUN_INTENSITY_FACTOR: f32 = 1000.0;
 const SUN_INTENSITY_FALLOFF_STEEPNESS: f32 = 1.5;
 const TURBIDITY: f32 = 2.0;
 
+pub fn tonemap(col: Vec3) -> Vec3 {
+    // see https://www.desmos.com/calculator/0eo9pzo1at
+    const A: f32 = 2.35;
+    const B: f32 = 2.8826666;
+    const C: f32 = 789.7459;
+    const D: f32 = 0.935;
+
+    let z = pow(col, A);
+    z / (pow(z, D) * B + Vec3::splat(C))
+}
+
 fn total_rayleigh(lambda: Vec3) -> Vec3 {
-    (8.0 * PI.pow(3.0)
+    (8.0 * PI.powf(3.0)
-        * (REFRACTIVE_INDEX.pow(2.0) - 1.0).pow(2.0)
+        * (REFRACTIVE_INDEX.powf(2.0) - 1.0).powf(2.0)
         * (6.0 + 3.0 * DEPOLARIZATION_FACTOR))
         / (3.0 * NUM_MOLECULES * pow(lambda, 4.0) * (6.0 - 7.0 * DEPOLARIZATION_FACTOR))
 }
@@ -39,11 +55,11 @@ fn total_mie(lambda: Vec3, k: Vec3, t: f32) -> Vec3 {
 }
 
 fn rayleigh_phase(cos_theta: f32) -> f32 {
-    (3.0 / (16.0 * PI)) * (1.0 + cos_theta.pow(2.0))
+    (3.0 / (16.0 * PI)) * (1.0 + cos_theta.powf(2.0))
 }
 
 fn henyey_greenstein_phase(cos_theta: f32, g: f32) -> f32 {
-    (1.0 / (4.0 * PI)) * ((1.0 - g.pow(2.0)) / (1.0 - 2.0 * g * cos_theta + g.pow(2.0)).pow(1.5))
+    (1.0 / (4.0 * PI)) * ((1.0 - g.powf(2.0)) / (1.0 - 2.0 * g * cos_theta + g.powf(2.0)).powf(1.5))
 }
 
 fn sun_intensity(zenith_angle_cos: f32) -> f32 {
@@ -58,7 +74,7 @@ fn sun_intensity(zenith_angle_cos: f32) -> f32 {
 
 fn sky(dir: Vec3, sun_position: Vec3) -> Vec3 {
     let up = Vec3::new(0.0, 1.0, 0.0);
-    let sunfade = 1.0 - (1.0 - (sun_position.y() / 450000.0).exp()).saturate();
+    let sunfade = 1.0 - (1.0 - saturate(sun_position.y / 450000.0).exp());
     let rayleigh_coefficient = RAYLEIGH - (1.0 * (1.0 - sunfade));
     let beta_r = total_rayleigh(PRIMARIES) * rayleigh_coefficient;
 
@@ -67,7 +83,7 @@ fn sky(dir: Vec3, sun_position: Vec3) -> Vec3 {
 
     // Optical length, cutoff angle at 90 to avoid singularity
     let zenith_angle = acos_approx(up.dot(dir).max(0.0));
-    let denom = (zenith_angle).cos() + 0.15 * (93.885 - ((zenith_angle * 180.0) / PI)).pow(-1.253);
+    let denom = (zenith_angle).cos() + 0.15 * (93.885 - ((zenith_angle * 180.0) / PI)).powf(-1.253);
 
     let s_r = RAYLEIGH_ZENITH_LENGTH / denom;
     let s_m = MIE_ZENITH_LENGTH / denom;
@@ -92,7 +108,7 @@ fn sky(dir: Vec3, sun_position: Vec3) -> Vec3 {
             sun_e * ((beta_r_theta + beta_m_theta) / (beta_r + beta_m)) * fex,
             0.5,
         ),
-        ((1.0 - up.dot(sun_direction)).pow(5.0)).saturate(),
+        saturate((1.0 - up.dot(sun_direction)).powf(5.0)),
     );
 
     // Composition + solar disc
@@ -108,10 +124,17 @@ fn sky(dir: Vec3, sun_position: Vec3) -> Vec3 {
     lin + l0
 }
 
+fn get_ray_dir(uv: Vec2, pos: Vec3, look_at_pos: Vec3) -> Vec3 {
+    let forward = (look_at_pos - pos).normalize();
+    let right = Vec3::new(0.0, 1.0, 0.0).cross(forward).normalize();
+    let up = forward.cross(right);
+    (forward + uv.x * right + uv.y * up).normalize()
+}
+
 pub fn fs(constants: &ShaderConstants, frag_coord: Vec2) -> Vec4 {
     let mut uv = (frag_coord - 0.5 * Vec2::new(constants.width as f32, constants.height as f32))
         / constants.height as f32;
-    uv.set_y(-uv.y());
+    uv.y = -uv.y;
 
     // hard-code information because we can't bind buffers at the moment
     let eye_pos = Vec3::new(0.0, 0.0997, 0.2);
@@ -136,7 +159,7 @@ pub fn main_fs(
 ) {
     let constants = constants.load();
 
-    let frag_coord = Vec2::new(in_frag_coord.load().x(), in_frag_coord.load().y());
+    let frag_coord = Vec2::new(in_frag_coord.load().x, in_frag_coord.load().y);
     let color = fs(&constants, frag_coord);
     output.store(color);
 }

setup.bat

@@ -1,6 +1,3 @@
 setlocal
 
 rustup toolchain install nightly-2020-11-15 --component rust-src rustc-dev llvm-tools-preview
-
-git submodule init
-git submodule update

setup.sh

@@ -1,6 +1,3 @@
 #!/usr/bin/env bash
 
 rustup toolchain install nightly-2020-11-15 --component rust-src rustc-dev llvm-tools-preview
-
-git submodule init
-git submodule update