Experimentations around cuda graphs

candle-core/Cargo.toml

@@ -52,3 +52,7 @@ harness = false
 [[example]]
 name = "metal_basics"
 required-features = ["metal"]
+
+[[example]]
+name = "cuda_basics"
+required-features = ["cuda"]

candle-core/examples/cuda_basics.rs

@@ -7,8 +7,92 @@ extern crate intel_mkl_src;
 use anyhow::Result;
 use candle_core::{Device, Tensor};
 
+const USE_CUDA_GRAPH: bool = true;
+
+fn cuda_graph() -> Result<()> {
+    let device = Device::new_cuda_with_stream(0)?;
+    let cu_device = match &device {
+        Device::Cuda(dev) => dev,
+        _ => unreachable!(),
+    };
+    let cu_stream = cu_device.cu_stream();
+    {
+        // load_ptx cannot be called while capturing the stream so we need this to happen
+        // beforehand.
+        let u = Tensor::zeros((4096, 4096), candle_core::DType::F32, &device)?
+            .to_dtype(candle_core::DType::BF16)?;
+        let mut x = Tensor::zeros((4096, 4096), candle_core::DType::F32, &device)?
+            .to_dtype(candle_core::DType::BF16)?;
+        let v = Tensor::zeros(4096, candle_core::DType::F32, &device)?
+            .to_dtype(candle_core::DType::BF16)?;
+        let _x = x.mul(&u)?.broadcast_add(&v)?;
+        let _x = x.affine(1., 0.5)?;
+        x.slice_set(&u, 0, 0)?;
+        device.synchronize()?;
+    }
+    if USE_CUDA_GRAPH {
+        unsafe {
+            cudarc::driver::sys::lib()
+            .cuStreamBeginCapture_v2(
+                *cu_stream,
+                cudarc::driver::sys::CUstreamCaptureMode_enum::CU_STREAM_CAPTURE_MODE_THREAD_LOCAL,
+            )
+            .result()?
+        }
+    }
+    {
+        let u = Tensor::zeros((4096, 4096), candle_core::DType::F32, &device)?
+            .to_dtype(candle_core::DType::BF16)?;
+        let mut x = Tensor::zeros((4096, 4096), candle_core::DType::F32, &device)?
+            .to_dtype(candle_core::DType::BF16)?;
+        let v = Tensor::zeros((4096, 1), candle_core::DType::F32, &device)?
+            .to_dtype(candle_core::DType::BF16)?;
+        for _i in 0..100 {
+            // x.slice_set(&u, 0, 0)?;
+            // x.broadcast_add(&v)?;
+            x = x.affine(1., 0.5)?;
+            // x = (&u + &x)?;
+        }
+    }
+    if USE_CUDA_GRAPH {
+        let cu_graph: cudarc::driver::sys::CUgraph = unsafe {
+            let mut cu_graph = std::mem::MaybeUninit::uninit();
+            cudarc::driver::sys::lib()
+                .cuStreamEndCapture(*cu_stream, cu_graph.as_mut_ptr())
+                .result()?;
+            cu_graph.assume_init()
+        };
+        let cu_graph_e: cudarc::driver::sys::CUgraphExec = unsafe {
+            let mut cu_graph_e = std::mem::MaybeUninit::uninit();
+            cudarc::driver::sys::lib()
+                .cuGraphInstantiateWithFlags(cu_graph_e.as_mut_ptr(), cu_graph, 0)
+                .result()?;
+            cu_graph_e.assume_init()
+        };
+        println!("graph captured!");
+        for i in 1..100 {
+            println!("graph exec {i}");
+            unsafe {
+                cudarc::driver::sys::lib()
+                    .cuGraphLaunch(cu_graph_e, *cu_stream)
+                    .result()?
+            }
+            println!("sync");
+            if let Err(err) = device.synchronize() {
+                println!("err: {err:?}")
+            }
+            println!("done syncing");
+        }
+    } else {
+        device.synchronize()?;
+    }
+    Ok(())
+}
+
 fn main() -> Result<()> {
-    let device = Device::new_cuda(0)?;
+    cuda_graph()?;
+    return Ok(());
+    let device = Device::new_cuda_with_stream(0)?;
     let x = Tensor::randn(0f32, 1.0, (8 * 4096, 8 * 4096), &device)?
         .to_dtype(candle_core::DType::BF16)?;
     candle_core::cuda::set_gemm_reduced_precision_f32(false);