Replace vek with glam.

This fixes #179.

Author: LegNeato

crates/cuda_std/Cargo.toml

@@ -8,6 +8,7 @@ repository = "https://github.com/Rust-GPU/Rust-CUDA"
 readme = "../../README.md"
 
 [dependencies]
+glam = { version = ">=0.22", default-features = false, features = ["libm", "cuda", "bytemuck"] }
 vek = { version = "0.17.1", default-features = false, features = ["libm"] }
 cuda_std_macros = { version = "0.2", path = "../cuda_std_macros" }
 half = "2.4.1"

crates/cuda_std/src/lib.rs

@@ -49,7 +49,11 @@ mod float_ext;
 pub use cuda_std_macros::*;
 pub use float::GpuFloat;
 pub use float_ext::*;
+pub use glam;
 pub use half;
+#[deprecated(
+    note = "The `vek` module is deprecated, use `glam` instead."
+)]
 pub use vek;
 
 pub use half::{bf16, f16};

crates/cuda_std/src/rt/mod.rs

@@ -152,23 +152,23 @@ impl<'a> From<&'a GridSize> for GridSize {
         other.clone()
     }
 }
-impl From<vek::Vec2<u32>> for GridSize {
-    fn from(vec: vek::Vec2<u32>) -> Self {
+impl From<glam::UVec2> for GridSize {
+    fn from(vec: glam::UVec2) -> Self {
         GridSize::xy(vec.x, vec.y)
     }
 }
-impl From<vek::Vec3<u32>> for GridSize {
-    fn from(vec: vek::Vec3<u32>) -> Self {
+impl From<glam::UVec3> for GridSize {
+    fn from(vec: glam::UVec3) -> Self {
         GridSize::xyz(vec.x, vec.y, vec.z)
     }
 }
-impl From<vek::Vec2<usize>> for GridSize {
-    fn from(vec: vek::Vec2<usize>) -> Self {
+impl From<glam::USizeVec2> for GridSize {
+    fn from(vec: glam::USizeVec2) -> Self {
         GridSize::xy(vec.x as u32, vec.y as u32)
     }
 }
-impl From<vek::Vec3<usize>> for GridSize {
-    fn from(vec: vek::Vec3<usize>) -> Self {
+impl From<glam::USizeVec3> for GridSize {
+    fn from(vec: glam::USizeVec3) -> Self {
         GridSize::xyz(vec.x as u32, vec.y as u32, vec.z as u32)
     }
 }
@@ -228,23 +228,23 @@ impl<'a> From<&'a BlockSize> for BlockSize {
         other.clone()
     }
 }
-impl From<vek::Vec2<u32>> for BlockSize {
-    fn from(vec: vek::Vec2<u32>) -> Self {
+impl From<glam::UVec2> for BlockSize {
+    fn from(vec: glam::UVec2) -> Self {
         BlockSize::xy(vec.x, vec.y)
     }
 }
-impl From<vek::Vec3<u32>> for BlockSize {
-    fn from(vec: vek::Vec3<u32>) -> Self {
+impl From<glam::UVec3> for BlockSize {
+    fn from(vec: glam::UVec3) -> Self {
         BlockSize::xyz(vec.x, vec.y, vec.z)
     }
 }
-impl From<vek::Vec2<usize>> for BlockSize {
-    fn from(vec: vek::Vec2<usize>) -> Self {
+impl From<glam::USizeVec2> for BlockSize {
+    fn from(vec: glam::USizeVec2) -> Self {
         BlockSize::xy(vec.x as u32, vec.y as u32)
     }
 }
-impl From<vek::Vec3<usize>> for BlockSize {
-    fn from(vec: vek::Vec3<usize>) -> Self {
+impl From<glam::USizeVec3> for BlockSize {
+    fn from(vec: glam::USizeVec3) -> Self {
         BlockSize::xyz(vec.x as u32, vec.y as u32, vec.z as u32)
     }
 }

crates/cuda_std/src/thread.rs

@@ -19,7 +19,7 @@
 // TODO: write some docs about the terms used in this module.
 
 use cuda_std_macros::gpu_only;
-use vek::{Vec2, Vec3};
+use glam::{UVec2, UVec3};
 
 // different calling conventions dont exist in nvptx, so we just use C as a placeholder.
 extern "C" {
@@ -152,7 +152,7 @@ pub fn grid_dim_z() -> u32 {
 /// Gets the 3d index of the thread currently executing the kernel.
 #[gpu_only]
 #[inline(always)]
-pub fn thread_idx() -> Vec3<u32> {
+pub fn thread_idx() -> UVec3 {
     unsafe {
         Vec3::new(
             __nvvm_thread_idx_x(),
@@ -165,7 +165,7 @@ pub fn thread_idx() -> Vec3<u32> {
 /// Gets the 3d index of the block that the thread currently executing the kernel is located in.
 #[gpu_only]
 #[inline(always)]
-pub fn block_idx() -> Vec3<u32> {
+pub fn block_idx() -> UVec3 {
     unsafe {
         Vec3::new(
             __nvvm_block_idx_x(),
@@ -179,7 +179,7 @@ pub fn block_idx() -> Vec3<u32> {
 /// how many threads exist in each thread block in every direction.
 #[gpu_only]
 #[inline(always)]
-pub fn block_dim() -> Vec3<u32> {
+pub fn block_dim() -> UVec3 {
     unsafe {
         Vec3::new(
             __nvvm_block_dim_x(),
@@ -193,7 +193,7 @@ pub fn block_dim() -> Vec3<u32> {
 /// how many thread blocks exist in each grid in every direction.
 #[gpu_only]
 #[inline(always)]
-pub fn grid_dim() -> Vec3<u32> {
+pub fn grid_dim() -> UVec3 {
     unsafe {
         Vec3::new(
             __nvvm_grid_dim_x(),
@@ -232,26 +232,26 @@ pub fn index_1d() -> u32 {
 }
 
 #[inline(always)]
-pub fn index_2d() -> Vec2<u32> {
+pub fn index_2d() -> UVec2 {
     let i = thread_idx_x() + block_idx_x() * block_dim_x();
     let j = thread_idx_y() + block_idx_y() * block_dim_y();
-    Vec2::new(i, j)
+    UVec2::new(i, j)
 }
 
 #[inline(always)]
-pub fn index_3d() -> Vec3<u32> {
+pub fn index_3d() -> UVec3 {
     let i = thread_idx_x() + block_idx_x() * block_dim_x();
     let j = thread_idx_y() + block_idx_y() * block_dim_y();
     let k = thread_idx_z() + block_idx_z() * block_dim_z();
-    Vec3::new(i, j, k)
+    UVec3::new(i, j, k)
 }
 
 /// Whether this is the first thread (not the first thread to be executing). This function is guaranteed
 /// to only return true in a single thread that is invoking it. This is useful for only doing something
 /// once.
 #[inline(always)]
 pub fn first() -> bool {
-    block_idx() == Vec3::zero() && thread_idx() == Vec3::zero()
+    block_idx() == UVec3::ZERO && thread_idx() == UVec3::ZERO
 }
 
 /// Gets the number of threads inside of a warp. Currently 32 threads on every GPU architecture.

crates/cust/CHANGELOG.md

@@ -4,6 +4,7 @@ Notable changes to this project will be documented in this file.
 
 ## Unreleased
 
+- `cuda_std::vek` is now deprecated. Use `cuda_std::glam`.
 - Add `memory::memcpy_dtoh` to allow copying from device to host.
 - `DeviceSlice` is represented as a slice again, but as `[()]` instead of `[T]`.
 - Reimplemented `Index` and `IndexMut` for `DeviceSlice` and removed `DeviceSlice::index`.

crates/optix_device/Cargo.toml

@@ -2,12 +2,17 @@
 name = "optix_device"
 version = "0.1.0"
 edition = "2021"
-authors = ["Anders Langlands <[email protected]>", "Riccardo D'Ambrosio <[email protected]>"]
+authors = [
+    "Anders Langlands <[email protected]>",
+    "Riccardo D'Ambrosio <[email protected]>"
+]
 
 [dependencies]
 bitflags = "2.8"
 cuda_std = { version = "0.2", path = "../cuda_std" }
-glam = { version = "0.29", features=["cuda", "libm"], default-features=false }
 paste = "1.0.15"
 seq-macro = "0.3.5"
 cust_core = { version = "0.1", path = "../cust_core" }
+
+[target.'cfg(not(target_os = "cuda"))'.dependencies]
+glam = { version = "0.29", features = ["cuda"], default-features = false }

crates/optix_device/src/hit.rs

@@ -1,6 +1,6 @@
 #[cfg(target_os = "cuda")]
 use core::arch::asm;
-use cuda_std::gpu_only;
+use cuda_std::{gpu_only, glam};
 use glam::Vec3;
 /// The type of primitive that a ray hit.
 #[repr(u32)]

crates/optix_device/src/lib.rs

@@ -14,8 +14,7 @@ pub mod trace;
 pub mod transform;
 pub mod util;
 
-use cuda_std::*;
-pub use glam;
+use cuda_std::{glam, *};
 use glam::UVec3;
 pub use misc::*;
 

crates/optix_device/src/ray.rs

@@ -1,7 +1,7 @@
 use crate::trace::*;
 #[cfg(target_os = "cuda")]
 use core::arch::asm;
-use cuda_std::gpu_only;
+use cuda_std::{glam, gpu_only};
 use glam::Vec3;
 
 /// Returns the ray origin that was passed into [`trace`] in world-space.

crates/optix_device/src/sys.rs

@@ -3,7 +3,7 @@
 use crate::trace::{RayFlags, TraversableHandle};
 #[cfg(target_os = "cuda")]
 use core::arch::asm;
-use cuda_std::gpu_only;
+use cuda_std::{glam, gpu_only};
 use glam::Vec3;
 use paste::paste;