Fix optix examples (Rust-GPU#160)

* Remove missing and unused symbols when compiling cuda.

See this [issue](ingowald/optix7course#35), from what I think is the source repo for this header.

Author: jorge-ortega

crates/optix/examples/common/gdt/gdt/gdt.h

@@ -107,7 +107,7 @@
 #define GDT_TERMINAL_RESET "\033[0m"
 #define GDT_TERMINAL_DEFAULT GDT_TERMINAL_RESET
 #define GDT_TERMINAL_BOLD "\033[1;1m"
-     
+
 
 
 
@@ -118,15 +118,7 @@ namespace gdt {
 #ifdef __CUDACC__
   using ::min;
   using ::max;
-  // inline __both__ float abs(float f)      { return fabsf(f); }
-  // inline __both__ double abs(double f)    { return fabs(f); }
   using std::abs;
-  // inline __both__ float sin(float f) { return ::sinf(f); }
-  // inline __both__ double sin(double f) { return ::sin(f); }
-  // inline __both__ float cos(float f) { return ::cosf(f); }
-  // inline __both__ double cos(double f) { return ::cos(f); }
-
-  using ::saturate;
 #else
   using std::min;
   using std::max;
@@ -139,12 +131,12 @@ namespace gdt {
   // inline __both__ double abs(double f)    { return fabs(f); }
   inline __both__ float rcp(float f)      { return 1.f/f; }
   inline __both__ double rcp(double d)    { return 1./d; }
-  
+
   inline __both__ int32_t divRoundUp(int32_t a, int32_t b) { return (a+b-1)/b; }
   inline __both__ uint32_t divRoundUp(uint32_t a, uint32_t b) { return (a+b-1)/b; }
   inline __both__ int64_t divRoundUp(int64_t a, int64_t b) { return (a+b-1)/b; }
   inline __both__ uint64_t divRoundUp(uint64_t a, uint64_t b) { return (a+b-1)/b; }
-  
+
 #ifdef __CUDACC__
   using ::sin; // this is the double version
   // inline __both__ float sin(float f) { return ::sinf(f); }
@@ -174,7 +166,7 @@ namespace gdt {
 #else
 #  define osp_snprintf snprintf
 #endif
-  
+
   /*! added pretty-print function for large numbers, printing 10000000 as "10M" instead */
   inline std::string prettyDouble(const double val) {
     const double absVal = abs(val);
@@ -195,7 +187,7 @@ namespace gdt {
 
     return result;
   }
-  
+
 
 
   inline std::string prettyNumber(const size_t s)
@@ -214,7 +206,7 @@ namespace gdt {
     }
     return buf;
   }
-  
+
   inline double getCurrentTime()
   {
 #ifdef _WIN32

crates/optix/examples/ex02_pipeline/device/src/lib.rs

@@ -1,15 +1,8 @@
-#![cfg_attr(
-    target_os = "cuda",
-    no_std,
-    feature(register_attr),
-    register_attr(nvvm_internal)
-)]
 // #![deny(warnings)]
 #![allow(clippy::missing_safety_doc)]
 
 use cuda_std::*;
 use cust_core::DeviceCopy;
-
 use optix_device as optix;
 
 extern crate alloc;
@@ -54,7 +47,7 @@ pub unsafe fn __raygen__renderFrame() {
     if idx[0] == 3 && idx[1] == 4 {
         vprintf(
             c"Hello from Rust kernel!\n".as_ptr().cast(),
-            std::ptr::null::<core::ffi::c_void>(),
+            core::ptr::null::<core::ffi::c_void>(),
         );
 
         #[repr(C)]
@@ -63,7 +56,7 @@ pub unsafe fn __raygen__renderFrame() {
         vprintf(
             c"frame id is %d\n".as_ptr().cast(),
             &PrintArgs(core::ptr::read_volatile(&PARAMS.frame_id)) as *const PrintArgs
-                as *const std::ffi::c_void,
+                as *const core::ffi::c_void,
         );
     }
 }

crates/optix/examples/ex02_pipeline/src/renderer.rs

@@ -1,5 +1,5 @@
 use anyhow::{Context, Result};
-use cust::context::{Context as CuContext, ContextFlags};
+use cust::context::Context as CuContext;
 use cust::device::{Device, DeviceAttribute};
 use cust::memory::{CopyDestination, DeviceBox, DeviceBuffer, DevicePointer, DeviceVariable};
 use cust::stream::{Stream, StreamFlags};
@@ -40,8 +40,7 @@ impl Renderer {
         let srf_align = device.get_attribute(DeviceAttribute::SurfaceAlignment)?;
         println!("tex align: {}\nsrf align: {}", tex_align, srf_align);
 
-        let cuda_context =
-            CuContext::create_and_push(ContextFlags::SCHED_AUTO | ContextFlags::MAP_HOST, device)?;
+        let cuda_context = CuContext::new(device)?;
         let stream = Stream::new(StreamFlags::DEFAULT, None)?;
 
         let mut ctx = DeviceContext::new(&cuda_context, false)?;
@@ -149,7 +148,7 @@ impl Renderer {
         let launch_params = DeviceVariable::new(LaunchParams {
             frame_id: 17,
             fb_size: [width as u32, height as u32],
-            color_buffer: color_buffer.as_device_ptr(),
+            color_buffer: color_buffer.as_device_ptr().as_raw(),
         })?;
 
         Ok(Renderer {
@@ -174,7 +173,7 @@ impl Renderer {
         self.color_buffer = unsafe { DeviceBuffer::uninitialized(width * height)? };
         self.launch_params.fb_size[0] = width as u32;
         self.launch_params.fb_size[1] = height as u32;
-        self.launch_params.color_buffer = self.color_buffer.as_device_ptr();
+        self.launch_params.color_buffer = self.color_buffer.as_device_ptr().as_raw();
         Ok(())
     }
 

crates/optix/examples/ex04_mesh/Cargo.toml

@@ -12,7 +12,7 @@ anyhow = "1.0.44"
 glfw = "0.42.0"
 gl = "0.14.0"
 num-traits = "0.2.14"
-glam = { version = "0.20", features=["cuda"] }
+glam = { version = "0.29.2", features=["cuda"] }
 
 [build-dependencies]
 find_cuda_helper = { version = "0.2", path = "../../../find_cuda_helper" }

crates/optix/examples/rust/ex04_mesh_gpu/src/lib.rs

@@ -1,4 +1,3 @@
-#![cfg_attr(target_os = "cuda", no_std, register_attr(nvvm_internal))]
 #![allow(non_snake_case, clippy::missing_safety_doc)]
 
 use cuda_std::kernel;
@@ -76,7 +75,7 @@ pub unsafe fn __miss__radiance() {
 }
 
 extern "C" {
-    #[cfg_attr(target_os = "cuda", nvvm_internal(addrspace(4)))]
+    #[cfg_attr(target_os = "cuda", nvvm_internal::addrspace(4))]
     static PARAMS: LaunchParams;
 }
 

crates/optix/src/acceleration.rs

@@ -1687,7 +1687,7 @@ pub struct InstancePointerArray<'i> {
 }
 
 impl<'i> InstancePointerArray<'i> {
-    pub fn new(instances: &'i DeviceSlice<CUdeviceptr>) -> InstancePointerArray {
+    pub fn new(instances: &'i DeviceSlice<CUdeviceptr>) -> InstancePointerArray<'i> {
         InstancePointerArray { instances }
     }
 }

crates/optix_device/src/lib.rs

@@ -1,3 +1,4 @@
+#![cfg_attr(target_os = "cuda", feature(asm_experimental_arch))]
 #[cfg(target_os = "cuda")]
 use core::arch::asm;