Fix examples #162

trigpolynom · 2025-03-15T22:39:53Z

built successfully based on changes made on top of the work @jorge-ortega has already done in #160. Below are some config params I had to set for my Ubuntu 22.04 env:

Set CUDA Root and Paths

export CUDA_ROOT=/usr/lib/cuda
export CUDA_PATH=$CUDA_ROOT
export CUDA_NVVM_ROOT=$CUDA_ROOT/nvvm

CUDA Binary and Library Paths

export PATH=$CUDA_ROOT/bin:$CUDA_ROOT/nvvm/lib64:/root/.cargo/bin:$PATH
export LD_LIBRARY_PATH=$CUDA_ROOT/lib64:$CUDA_ROOT/nvvm/lib64:$CUDA_ROOT/nvvm/libdevice:$LD_LIBRARY_PATH
export LIBRARY_PATH=$CUDA_ROOT/lib64:$CUDA_ROOT/nvvm/libdevice:$LIBRARY_PATH

Rust and OptiX Configuration

export LLVM_LINK_STATIC=1
export RUST_LOG=info
export OPTIX_ROOT=$HOME/optix/
export OPTIX_ROOT_DIR=$OPTIX_ROOT

Add Rust Toolchain Path

export PATH=$HOME/.cargo/bin:$PATH
export LD_LIBRARY_PATH=$HOME/.rustup/toolchains/nightly-2025-03-02-x86_64-unknown-linux-gnu/lib:$LD_LIBRARY_PATH
export RUSTFLAGS="-C link-args=-ltinfo"

LegNeato

Thanks a ton for the PR! 👏

Some small changes and questions. You can also rebase on main as I landed the PR this depends on.

LegNeato · 2025-03-16T01:42:02Z

crates/find_cuda_helper/src/lib.rs

@@ -55,6 +55,10 @@ pub fn find_cuda_root() -> Option<PathBuf> {
    None
 }

+// pub fn find_cuda_root() -> Option<PathBuf> {


Why is this commented out? We use this in other places I believe.

LegNeato · 2025-03-16T01:49:04Z

examples/cuda/cpu/path_tracer/src/viewer.rs

@@ -55,9 +62,9 @@ pub fn run(camera: &Camera, scene: &Scene) -> ! {
    let event_loop = EventLoop::new();
    let wb = WindowBuilder::new()
        .with_title("Render")
-        .with_inner_size(PhysicalSize::new(WIDTH, HEIGHT));
+        .with_inner_size(PhysicalSize::new(WIDTH as f64, HEIGHT as f64)); // Explicitly specify f64


Suggested change

.with_inner_size(PhysicalSize::new(WIDTH as f64, HEIGHT as f64)); // Explicitly specify f64

.with_inner_size(PhysicalSize::new(WIDTH as f64, HEIGHT as f64));

LegNeato · 2025-03-16T01:49:28Z

rustc-ice-2025-03-15T03_46_45-55918.txt

@@ -0,0 +1,45 @@
+thread 'opt 4qj5v7pmzdsa87mgb0rgitfl0' panicked at /rustc/8c392966a013fd8a09e6b78b3c8d6e442bc278e1/compiler/rustc_codegen_ssa/src/back/write.rs:916:21:


Nit: remove.

LegNeato · 2025-03-16T01:50:04Z

examples/cuda/cpu/path_tracer/src/viewer.rs

-    window::WindowBuilder,
-    ContextBuilder,
-};
+// use glutin::{


Nit: remove commented code.

LegNeato · 2025-03-16T01:51:57Z

Also be sure to run cargo rustfmt --all and cargo clippy!

trigpolynom · 2025-03-16T03:27:07Z

Thank you for these! I got it to build right before hopping on a flight and forgot to clean some stuff up. Will clean up ASAP and handle formatting

jorge-ortega · 2025-03-15T23:46:53Z

examples/cuda/gpu/path_tracer_gpu/src/optix.rs

@@ -21,7 +21,6 @@ use optix_device::{

 extern "C" {
    #[cfg(target_os = "cuda")]
-    #[cfg_attr(target_os = "cuda", nvvm_internal(addrspace(4)))]


Update this to nvvm_interanal::addrspace(4) on the cuda target.

We might need to add back L23 to keep this gated only for cuda.

examples/cuda/gpu/path_tracer_gpu/src/lib.rs

trigpolynom · 2025-03-16T04:19:12Z

Ahhh im afraid its failing on cargo clippy. I spoke prematurely! All comments should be addressed though and I ran rustfmt. @LegNeato and @jorge-ortega . Still succeeding on cargo build though. Will address issues found by clippy.

LegNeato

CI is failing.

trigpolynom · 2025-03-16T15:06:13Z

CI is failing.

Yeah windows build fails and Linux clippy task fails but build succeeds. Will move this in progress. Sorry! Got too excited haha.

trigpolynom · 2025-03-16T21:06:02Z

have clippy running successfully now and rebased on main @jorge-ortega

LegNeato

Awesome, getting close! Looks like formatting is still off so CI is failing.

LegNeato · 2025-03-17T01:51:47Z

crates/optix/optix_wrapper.rs

@@ -212,8 +212,8 @@ pub type OptixLogCallback = ::std::option::Option<
 pub const OptixDeviceContextValidationMode_OPTIX_DEVICE_CONTEXT_VALIDATION_MODE_OFF:
    OptixDeviceContextValidationMode = 0;
 pub const OptixDeviceContextValidationMode_OPTIX_DEVICE_CONTEXT_VALIDATION_MODE_ALL:
-    OptixDeviceContextValidationMode = -1;
-pub type OptixDeviceContextValidationMode = ::std::os::raw::c_int;
+    OptixDeviceContextValidationMode = 4294967295;


Still haven't explained why this is better.

hmm this seems to be something that changed on build. looking into

This whole file is generated through bindgen, and its values are going to differ when building on linux vs windows. The build seems to suggest this file was only written there for debugging, as it also writes it to the output directory, and that one is ultimately included in sys.rs.

yeah it automatically converts -1 to 4294967295 on my machine on compilation.

tl;dr is it should be safe to ignore this file in your changes.

Should we add to gitignore?

I'd remove it and update build.rs so it only writes to the output dir. We can do that in another PR and ignore the file for now.

LegNeato · 2025-03-17T01:54:03Z

crates/rustc_codegen_nvvm/src/nvvm.rs

@@ -155,14 +155,20 @@ pub fn codegen_bitcode_modules(
 /// linked when building the nvvm program.
 pub fn find_libdevice() -> Option<Vec<u8>> {
    if let Some(base_path) = find_cuda_root() {
-        let libdevice_file = fs::read_dir(Path::new(&base_path).join("nvvm").join("libdevice"))
+        let libdevice_dir = Path::new(&base_path).join("nvvm").join("libdevice");
+        println!("Checking libdevice directory: {:?}", libdevice_dir);


Just printing feels wrong here. Can we use the context (I think there is a dcx() somewhere) or make this a trace!() or a debug!() ? Pretty sure rustc has tracing, but I don't know which version it was added in.

This was for debugging and I forgot to remove. My bad! Should be fixed now.

jorge-ortega · 2025-03-17T03:42:28Z

crates/optix/src/acceleration.rs

@@ -1139,13 +1134,13 @@ impl From<CurveType> for sys::OptixPrimitiveType {
 #[repr(u32)]
 #[derive(Copy, Clone, PartialEq)]
 pub enum VertexFormat {
-    None = sys::OptixVertexFormat_OPTIX_VERTEX_FORMAT_NONE as u32,


Removing the cast will break on windows. The enum type in the generated bindings are c_int on windows (thus the u32 cast) but c_uint on linux. See rust-lang/rust-bindgen#1361

If clippy is complaining about this, we'll need to ignore them here.

Yeah that was a clippy change. I'll make these changes tomorrow.

trigpolynom · 2025-03-18T03:13:28Z

im getting rustfmt and cargo build to pass using the .github/rust.yml commands but clippy is failing on issues like the following:

help: if you import `c_int`, refer to it directly
    |
204 -             .and_then(|millis| c::c_int::try_from(millis).ok())
204 +             .and_then(|millis| c_int::try_from(millis).ok())
    |

error[E0433]: failed to resolve: use of undeclared type `IFlags`
   --> /home/nonpl/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/rustix-1.0.2/src/fs/ioctl.rs:149:35
    |
149 |         ioctl::ioctl(fd, ctl).map(IFlags::from_bits_retain)
    |                                   ^^^^^^ use of undeclared type `IFlags`

How is it failing on imported code? @LegNeato and @jorge-ortega, please excuse any non-rust lingo or ignorance on my behalf.

jorge-ortega · 2025-03-18T22:19:26Z

Not sure your error is coming from clippy, but rather trying to compile something before clippy runs. Try running cargo tree -i rustix to see what it pulling in the failing crate. My guess it's going to be sysinfo. On windows, I'm not able to compile ntapi, which gets pulled in through sysinfo 0.24.7. We might need a newer version then this to get it to compile on windows.

jorge-ortega · 2025-03-18T22:29:51Z

CI passes on windows because we don't build the path_tracer example atm, but the changes don't compile for me on windows.

trigpolynom · 2025-03-18T22:42:59Z

Not sure your error is coming from clippy, but rather trying to compile something before clippy runs. Try running cargo tree -i rustix to see what it pulling in the failing crate. My guess it's going to be sysinfo. On windows, I'm not able to compile ntapi, which gets pulled in through sysinfo 0.24.7. We might need a newer version then this to get it to compile on windows.

hmm that makes sense but cargo build completes successfully for me not just when I run CI checks.

LegNeato · 2025-03-19T01:44:59Z

@trigpolynom I'd prefer to not use discord at this point and stick to GitHub (I turned on discussions). I'd prefer the server was shut down or not advertised as associated with the project.

Discord is not nearly as searchable. Over and over again I've seen it drag maintainers down with the same questions. Information and questions are better in GitHub so it is searchable and can be referenced from tasks and issues. I've also seen discord encourage drive by questions. It's easier to just ask than learn, search, read docs, read code, or solve their own issues. Answers almost never make it back to the docs. For whatever reason answers from GitHub discussions more often than not make it back into code and docs in my experience...maybe people are in a different mindset in the GitHub UI 🤷‍♂️.

trigpolynom · 2025-03-19T01:58:51Z

@trigpolynom I'd prefer to not use discord at this point and stick to GitHub (I turned on discussions). I'd prefer the server was shut down or not advertised as associated with the project.

Discord is not nearly as searchable. Over and over again I've seen it drag maintainers down with the same questions. Information and questions are better in GitHub so it is searchable and can be referenced from tasks and issues. I've also seen discord encourage drive by questions. It's easier to just ask than learn, search, read docs, read code, or solve their own issues. Answers almost never make it back to the docs. For whatever reason answers from GitHub discussions more often than not make it back into code and docs in my experience...maybe people are in a different mindset in the GitHub UI 🤷‍♂️.

Understood! Makes complete sense!

jorge-ortega · 2025-03-19T02:30:47Z

Don't think this merged cleanly with my changes. Sysinfo is still outdated and the optix wrapper should have been removed. Can you try updating your branch again?

trigpolynom · 2025-03-19T02:56:02Z

Hmmm my branch is borked after messing with a reset. Do you want to open up a pr with your branch since you got path tracer working? I can close this one.

jorge-ortega · 2025-03-19T04:02:10Z

Closing this one and continuing in #171.

LegNeato requested changes Mar 16, 2025

View reviewed changes

jorge-ortega reviewed Mar 16, 2025

View reviewed changes

trigpolynom force-pushed the fix-examples branch from 4f014f8 to 9ccc405 Compare March 16, 2025 04:06

LegNeato requested changes Mar 16, 2025

View reviewed changes

trigpolynom force-pushed the fix-examples branch from a7f106c to 45ab416 Compare March 16, 2025 20:28

LegNeato requested changes Mar 17, 2025

View reviewed changes

jorge-ortega requested changes Mar 17, 2025

View reviewed changes

trigpolynom added 12 commits March 18, 2025 15:55

fixing

b935463

path_tracer

fdcea86

changed

8377048

merged

ddd2598

merged

d8c9807

addressing comments

512174d

removing log

105d451

added cfg back

2fd24f6

fixed cfg on launch params

16344c0

clippy ran successfully

22e66a0

addressed

31a1b03

format works and build works for linux

76e60a2

This comment was marked as resolved.

Sign in to view

trigpolynom force-pushed the fix-examples branch from 9f10e96 to 76e60a2 Compare March 19, 2025 02:51

jorge-ortega mentioned this pull request Mar 19, 2025

Fix path_tracer example #171

Merged

jorge-ortega closed this Mar 19, 2025

	.with_inner_size(PhysicalSize::new(WIDTH as f64, HEIGHT as f64)); // Explicitly specify f64
	.with_inner_size(PhysicalSize::new(WIDTH as f64, HEIGHT as f64));

		@@ -0,0 +1,45 @@
		thread 'opt 4qj5v7pmzdsa87mgb0rgitfl0' panicked at /rustc/8c392966a013fd8a09e6b78b3c8d6e442bc278e1/compiler/rustc_codegen_ssa/src/back/write.rs:916:21:

Fix examples #162

Fix examples #162

Conversation

trigpolynom commented Mar 15, 2025 • edited Loading

Set CUDA Root and Paths

CUDA Binary and Library Paths

Rust and OptiX Configuration

Add Rust Toolchain Path

LegNeato left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

LegNeato commented Mar 16, 2025

trigpolynom commented Mar 16, 2025

Choose a reason for hiding this comment

jorge-ortega Mar 18, 2025 • edited Loading

Choose a reason for hiding this comment

trigpolynom commented Mar 16, 2025 • edited Loading

LegNeato left a comment

Choose a reason for hiding this comment

trigpolynom commented Mar 16, 2025

trigpolynom commented Mar 16, 2025

LegNeato left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

trigpolynom commented Mar 18, 2025 • edited Loading

jorge-ortega commented Mar 18, 2025

jorge-ortega commented Mar 18, 2025

trigpolynom commented Mar 18, 2025

This comment was marked as resolved.

LegNeato commented Mar 19, 2025 • edited Loading

trigpolynom commented Mar 19, 2025

jorge-ortega commented Mar 19, 2025

trigpolynom commented Mar 19, 2025

jorge-ortega commented Mar 19, 2025

trigpolynom commented Mar 15, 2025 •

edited

Loading

jorge-ortega Mar 18, 2025 •

edited

Loading

trigpolynom commented Mar 16, 2025 •

edited

Loading

trigpolynom commented Mar 18, 2025 •

edited

Loading

LegNeato commented Mar 19, 2025 •

edited

Loading